Roman Villas: A Study in Social Structure

  • 29 677 5
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview



A Study in Social Structure

J.T. Smith Drawings by A.T. Adams

London and New York

First published 1997 by Routledge 11 New Fetter Lane, London EC4P 4EE This edition published in the Taylor & Francis e-Library, 2002. Simultaneously published in the USA and Canada by Routledge 29 West 35th Street, New York, NY 10001 © 1997 J.T. Smith All rights reserved. No part of this book may be reprinted or reproduced or utilized in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging in Publication Data Smith, J.T. (John Thomas), 1922– Roman villas : a study in social structure / J.T. Smith ; drawings by A.T. Adams. p. cm. Includes bibliographical references and index. ISBN 0-415-16719-1 (hc) 1. Architecture, Domestic—Rome. 2. Architecture, Roman—Europe. 3. Architecture and society—Rome. I. Title. NA335.E85S65 1997 728 '.09376—dc21 97–248 ISBN 0-415-16719-1 (Print Edition) ISBN 0-203-00405-1 Master e-book ISBN ISBN 0-203-22010-2 (Glassbook Format)

For Heather


xiii xxvi xxix xxx xxxii

List of figures Preface Editorial notes on drawings Glossary List of abbreviations of locations PART I: AIMS AND METHODS 1

Aims and scope of the book Plan typology: some objections The terminology of social structure The history of villa classification Villa architecture What is a villa and what is its name? A representative sample?

3 4 5 6 9 10 11


Methods and assumptions What constitutes a type? Origins of plan analysis Relevance of vernacular architecture studies The principle of alternate development Room use House plans and social structure Classical canons: symmetry and axiality Matters ignored

13 13 13 14 15 15 16 18 19


Hall houses Stahl and Mayen Classification and its problems Layout and functions of the hall

23 24 26 26 vii

— Contents — The hearth/entrance relation The off-centre hearth The lower end Broad halls with more than one hearth Narrow halls Single-ended halls The inner room subdivided Double-ended halls Aisled houses Ridge-post halls Wide-nave halls Nave and aisles of equal width Hall or yard? The case of Inzigkofen Hall and porticus unified: Sinsheim and Kingsweston Divided halls Halls: a summary

29 29 30 31 32 33 35 35 36 37 40 40 41 43 43 45


Row-type houses Introduction Interconnecting rooms: Newport (I.o.W.) and Lamargelle Newport enlarged: Sparsholt Lamargelle enlarged: Downton The size and forms of units Houses entered at one end The longitudinal lobby Longitudinal lobbies and reverse symmetry Houses with two front entrances Four-cell houses The compact row type

46 46 47 49 49 51 54 56 57 59 60 62


Developed forms of row-house Kirchberg: three-room and lobby units The elaboration of units: Laufen-Müschag Bierbach: how many units in a row-type villa? Lobby types Room proportions and their implications Articulation or separate rooms? Blocks of small rooms Transformation of lobbies The significance of disparate units

65 65 67 68 70 75 76 77 78 79


Developed forms of hall houses Narrow end rooms End rooms as byres? End rooms organised around lobbies End rooms: Inzigkofen and its analogues

80 80 82 83 86


— Contents — Disproportion between the ends of the hall Larger apartments: Bocholtz-Vlengendaal Other large end apartments: Kinheim Social significance of room groupings Development of wide-nave houses

88 89 90 91 93


Problematic house types Bondorf: small house or large yard? Other villas of Bondorf type The problem of function Ranges of end rooms in broad halls Yard rather than hall? The smallest row-houses? One-room buildings: houses or what? Double-depth plans Back-to-back houses The interpretation of double-depth plans The axial corridor The social basis of axial-corridor plans Back-to-back halls

94 94 97 99 100 101 102 105 106 109 110 112 114 115


The porticus-with-pavilions: pavilions Pavilions: the classic form Where was the pavilion entered? Asymmetrical pavilions Practical asymmetry: the case of Rothselberg Asymmetry as an expression of status Small difference, significant social implication Oblong pavilions Pavilions not rectangular Minimal pavilions Pavilions in row-type houses Detached wings or quasi-pavilions One storey or two? Conclusion

117 117 119 120 120 121 123 123 124 125 126 128 128 129


The porticus-with-pavilions: porticuses Open-ended porticuses: an expression of social structure Hierarchy in porticuses open at one end The ultimate open-ended porticus: Csucshegy Tapering and splayed porticuses Two households: symmetrical entrances Two households: porticus-with-pavilions front and rear Porticus functions: recreation Recreation: ghost pavilions The porticus as living-space

130 130 132 132 134 136 137 139 140 141


— Contents — Porticuses as workplace: wide or continuous porticuses



The elements and forms of villa complexes Irregular yards Divided yards Courtyards of geometrical shapes Tapering yards Fan-shaped yards Sub-triangular courtyards Rhomboidal courtyards and the problem of Hambach 512 Long rectangular courtyards Domestic courtyards and courtyard houses Parallel ranges and detached facades Linear villas Rectangular farmyards Conclusion

144 144 149 151 152 156 157 158 159 162 166 167 171 171


Palaces, peristyle houses and luxury villas Palaces Villas as seats of lordship Peristyle villas Luxury villas Formality and luxury Lordship or joint proprietorship?

172 173 178 183 190 193 195


The villas of south-east Europe Hall-type villas Square halls Row-houses Row-house equivalents Houses with one cross-wing Houses with multiple small rooms The ways of grouping buildings L-shaped plans Forms of courtyard and farmyard Rectilinear yards Courtyard villas and peristyles Fortified villas? Conclusion

199 199 201 202 202 205 207 208 211 212 213 214 215 216


The late pre-Roman Iron Age background The evidence: limitations and problems Two-aisled houses: forms and distribution Two-aisled houses: interpretation


219 219 220 222

— Contents — Three-aisled houses Squarish and trapezoidal buildings Monospan oblong buildings Round-houses Complex houses Conclusion

226 227 229 230 231 232


Modes of Romanisation Native to Roman by easy stages: Rijswijk and St Lythans-Whitton Romanisation by luxury Houses built over boundaries Collingham and Radley: continuity or discontinuity? Built on ditches: Kaisersteinbruch and Rudston Removal to a new site Romanised courtyards and wooden buildings An implied first phase in timber? Building in stone: early halls Building in stone: round-house or round pavilion?

233 233 238 239 243 249 250 250 252 253 254


Patterns of villa development The open hall: its rise The open hall: its decline The aisled house The social development of a villa Sudeley-Spoonley Wood Change in a row-house: Great Weldon A stable villa population Prosperity without social change: Schupfart-Betberg

257 257 261 263 264 268 270 271 274


A model of development The hypothesis Before the Romans The emergence of villas The consolidation of settlements The diffusion of villa types Development of the villa system Social change in lls and yards Shrines as a unifying device Freestanding shrines and temples Why do large villas differ so much? The problem of a stable villa population Courtyards for kin or slaves? The emergence of hereditary lordship The end of the villa system Conclusion

275 275 277 278 279 282 284 285 288 291 292 293 295 300 301 301


— Contents — 303 324 330 340 360

Notes Abbreviations of periodicals Bibliography List of villas and other sites mentioned Index




Stahl and related villas: Bargen im Hegau Koerich-Goeblingen 1 [I], [II] Ludwigsburg-Pflugfelden Mayen im Brasil III–VI, VIII Saaraltdorf [I], [II] Serville Stahl [I], [II] Tiefenbach



Halls with evidence of use: Bollendorf I, III Börstingen Bruchsal-Ober Grombach A Konken Mamer-Gaschtbierg



British halls: (a) with evidence of use Farmington-Clear Cupboard I/F Somerton-Bradley Hill 1 I Somerton-Catsgore 2.1 Somerton-Catsgore 2.5 Stowey Sutton-Chew Park (b) others Byfield Langton Dwelling House I, II, III Laugharne-Cwmbrwyn North-Stainley-Castle Dykes [I], [II] Wraxall I/F



— List of Figures — 4

Halls used for stalling animals: Blieskastell-Altheim Böchweiler Crain



French hall houses: Brain-sur-Allonnes Grémecey Saint-Pierre-la-Garenne



Single- and double-ended halls: Brücken Heppenheim Overasselt Quinton II Rothselberg



Aisled buildings: (a) aisled houses Carisbrooke Denton II East Grimstead Exning-Landwade I, II Mansfield Woodhouse Petersfield – Stroud West Blatchington (b) wide-nave hall Hölstein (c) nave and aisles of equal width Winkel-Seeb A II, VI Winkel-Seeb B I, IIB (d) ridge-post hall Fishtoft


Halls open to porticus: Sigmaringen-Steinäcker Sinsheim-Sinsheimer-Wald



Divided halls: Bad Homburg Heerlen-Boventse Caumer Houthem-Vogelsang Laperrière-sur-Saône Saint-Aubin-sur-Mer Voerendaal-Ten Hove I




— List of Figures — 10

Row-houses: Ditchley I/II, III Downton Lamargelle-Versingue Newport I.o.W.



The varied positions of lobbies: Brixworth I–III Cléry-sur-Somme Ellesborough-Terrick Faversham I Mansfield Woodhouse [I], [F] St Stephens-Park Street VI Welwyn-Lockleys I, II/III



Unit-system villas: Cenero-Murias de Belono Gargrave-Kirk Sink



Row-houses with two entrances: Hemel Hempstead-Gadebridge Park III Maulévrier North Cerney-The Ditches Romegoux-La Vergnée



Virtually identical units: Boos-Le Bois Flahaut Hérouville-Lébisey



End entrances: Farningham-Manor House I Ormalingen



Longitudinal lobbies: Eaton by Tarporley Huntsham N Wellow



Reverse symmetry: Beadlam I, II High Wycombe



Compact row houses: Civray-Le Poirier Molet Hummetroth – row house equivalent Lussas-et-Nontronneau Primelles-Champ Chiron



— List of Figures — 19

Lobbies and room groupings: Anthée, detail Bierbach Cartagena-El Castillet L’Ecluse Leckbosch Küttigen-Kirchberg [I], [II] Laufen-Müschag [I], [F]



Lobbies and related forms: (a) longitudinal lobbies and elongated rooms Haccourt I Liestal-Munzach, detail Puig de Cebolla Vouneuil-sous-Biard (b) L-lobbies Fontoy-Moderwiese, aedicule North Leigh-Shakenoak B IIa/b, IIIa Stadtbergen I (c) widened lobby La Roche-Maurice



Unit variations in row-houses: Les Mesnuls Sarmentsdorf



Halls with long end rooms: Courcelles-Urville Grenchen I–III Mundelsheim Wahlen



Halls with complex end blocks: Buchten Evelette Mauren Nuth-Vaasrade Rheinbach-Flerzheim Tholey-Sotzweiler II/III



Halls with double-depth end blocks: Bocholtz-Vlengendaal [I], [F] Kinheim [F] Newel II, III, IV



Halls or yards? Alpnach-Dorf



— List of Figures — Bad Rappenau-Zimmerhof Bondorf Eckartsbrunn Messkirch Niedereschach-Fischbach 3 Remmingsheim Sigmaringen-Laiz A 26

Halls with freestanding corner posts: Bilsdorf Ludwigsburg-Hoheneck



Halls with rear columns: Hechingen-Stein I, F Inzigkofen Neckarzimmern-Stockbronner-Hof Schambach Stuttgart-Stammheim



Halls subdivided by timber partitions: Lörrach-Brombach



The smallest row houses: Bouchoir Dury L’Etoile Harbonnières Kempten-Loja 1 II Kempten-Loja 2 I/II, III Niedereschach-Fischbach 2 Plouneventer-Kerilien II Rheinfelden-Herten-Warmbach Vaux Wancourt



One-room houses: Biberist-Spitalhof Koerich-Goeblingen 2 I–V Monreal I–III Pin-Izel Sontheim an der Brenz 2, I–V



Adjoining parallel ranges and back-to-back halls: (a) parallel ranges Ambresin Basse-Wavre [I], [F] xvii


— List of Figures — Frilford Günzenheim-Staatsforst-Sulz Kernen-Rommelshausen Lendin Munzenberg-Gambach I/F Villers-Bretonneux (b) back-to-back hall Walsbetz-Hemerijk 32

Anomalous double-depth houses: Ashtead Frankfurt-Bornheim



Houses with internal corridor or yard? Bad Dürkheim-Ungstein Geislingen-Heidegger Hof Hohenfels Mézières-en-Santerre/La Croix-Saint-Jacques Neuburg a.d. Donau



Pavilion details: Bristol-Kingsweston, detail Hüfingen Stahl, detail Titelberg, aedicula Whittington



Minimal and quasi-pavilions: Langenau-Osterstetten Stratford-upon-Avon-Tiddington



Terminal rooms: Bachenau Beckingen Brighstone-Rock Broichweiden Rainecourt



Open-ended porticuses: Budapest III-Csúcshegy Démuin Mainz-Kastel, aedicule Shipham-Star II A



Tapering and splayed porticuses: Laufenburg



— List of Figures — 39

Houses with two porticuses or two porches: (a) halls Dirlewang Doische-Vodelée Herschwerler-Pettersheim Manderscheid (and detail) (b) row-house Gayton Thorpe



Ghost pavilions and porticuses as living-space: Bergen-Auf-dem-Keller Great Staughton Ovillers



Irregular and divided yards: (a) irregular Ludwigsburg-Hoheneck I/II, III Regensburg-Burgweinting (b) divided yards Bruchsal-Ober Grombach Ewhurst-Rapsley Friedberg-Pfingstweide, yard plan Friedberg-Pfingstweide 1, 2 Lauffen am Neckar Messkirch Olfermont



Yards not rectangular: (a) tapering Cachy Ecoust-Saint-Mien Niederzier-Hambach 69 Mansfield Woodhouse Mayen Le Mesge (b) fan-shaped and sub-triangular Darenth Hambleden-Yewden Manor Rockbourne II–IV, VI, VIII (c) development of yards Bignor IIA North Leigh I, IV, V



Rectangular courtyards: Athies Fliessem



— List of Figures — Hungen-Bellersheim Marchelepot Warfusée-Nord, Abancourt 44

Rectangular courtyards with hints of rebuilding: Belleuse-les Mureaux Davenescourt Marboué-Mienne Marboué-Mienne, west range North Wraxall Sudeley-Spoonley Wood Wiesbaden Höfchen



Parallel ranges and detached facades: Arradon-Lodo Leutersdorf I–III Orton Longueville-Orton Hall Farm Plomelin-Perennou Pulborough Winterton Winterton G, I–III



Linear villas: Bocholtz-Vlengendaal Halstock II Jemelle-Neufchâteau Maillen-al Sauvenière


Palaces: Budapest-Aquincum Budapest III-Aquincum, officers’ lodgings Fishbourne Palace Fishbourne Palace, officers’ lodgings Rome, Flavian Palace, administrative part Rome, Flavian Palace, general plan Woodchester



Seats of lordship: Almenara de Adaja Badajoz-la Cocosa, general plan Badajoz-la Cocosa, peristyle Bignor, detail Bignor, general plan Box North Leigh Rielves




— List of Figures — 49

Spanish peristyles: Cuevas de Soria Jumilla-Los Cipreses Santa Colomba de Somoza



Lalonquette I–V



Montmaurin I, II



Luxurious formal villas: Graz-Thalerhof Haccourt II Montrozier-Argentelle



Joint proprietorship – grandeur, not luxury: Fliessem Fliessem, detail Fliessem, interpretation Haut Clocher-Saint Ulrich



End-entrance halls: Bistrica Budakalász Gyulafirátót Majdan Telita



Square halls: Bihác-Zaloje Lisiicí II Mali Mounj Stolac 2



Row-house equivalents: Draevica I–IV Keszthely-Fenékpuszta 8 Travnik-Rankovic Sarmizgetusa Winden am See II, III



Houses essentially of two cells: Cincis Fischamend-Katharinenhof Maria Ellend-Ellender Weingärten Sarajevo-Stup Sarica



— List of Figures — 58

Houses with one cross-wing: Proboj Regelsbrunn Stolac 3 Stolac 5



Houses with multiple small rooms: Apahida Ciumafaia [II] Keszthely-Fenékpuszta 7



Ways of grouping buildings: (a) linear plans Iskar-Gara Konska Kralev Dol (b) block plans Hobita-Gradiste Izola Manerau Novi Saher (c) yards Hobita-Gradiste Kaisersteinbruch-Königshof C, D Lisiici III Ljuina Orlandovtsi Panik


Aisled buildings: (a) two-aisled or ring-post Beegden Eching-Autobahn Haps T Haunstetten Oss-Ussen Zijderveld (b) two-/three-aisled Befort-Aleburg Heuneburg Königsbrunn Verberie



Square buildings: (a) larger Braughing-Skeleton Green




— List of Figures — Danebury Hornchurch Landshut-Salmansberg (b) the smallest square houses Croft Ambrey Mailhac-Cayla Martigues-L’Arquet The Wrekin 63

Monospan halls: Babworth-Dunston’s Clump Kaalheide-Krichelberg Niederzier Hambach 59 Radley-Barton Court Farm 3



Grouping of houses: Draughton Pilsdon Pen St Michael-Gorhambury Sigean-Pech-Maho Villeneuve-St-Germain



Stages of Romanisation: Barnsley Park, I, III, V/VI, VIII Hartfield-Garden Hill II–IV Rijswijk IA–D, IIA–B, IIIA–B St Lythans-Whitton I–VI, VIII



Building over boundaries: Aylesford-Eccles I Condé-Folie I, II Kingsweston, Bristol Laugharne-Cwmbrwyn Marshfield II, IIIB Uplyme-Holcombe IIA



Building along or over ditches: Hamblain-les-Prés Hemel Hempstead-Gadebridge Park Jublains-La Boissière Kaisersteinbruch-Königshof Milton Keynes-Bancroft II, V Rudston



Problems of continuity: (a) discontinuity?



— List of Figures — Collingham-Dalton Parlours Radley-Barton Court Farm I–III (b) moving to a new site Port-le-Grand Sparsholt I–III (c) timber predecessors of stone houses Bedburg-Garsdorf Bellikon Bözen Hemel Hempstead-Boxmoor I Neumagen-Dhron-Papiermühle Osterfingen I, II (d) the significance of round-houses Manfield-Holme House II–IV Ringstead I, II 69

Rise and decline of the open hall: (a) rise Francolise-San Rocco I, IA (b) decline Bad Neuenahr-Ahrweiler [I], [F] Bristol-Kingsweston [I], alternative [II] Bruckneudorf-Parndorf [I], [II], [F] Friedberg-Fladerlach [I], [III] Frocester Court I–III, alternative [II] Frocester Court, alternative II Graux Maidenhead-Cox Green, I, II Mehring I, F Schleitheim Weitersbach I, II


The social development of villas: Blankenheim IA, IIA, IIIA Great Weldon I–IV, V/VI St Stephen-Park Street VII, VIII Sudeley-Spoonley Wood


Social stability in villas: Köln-Mungersdorf I, II/III, VI Schupfart-Betburg II, III


Second-phase enlargement: Leiwen-Bohnengarten I–III Noyers-sur-Serein [II], [F] Winkel-Seeb, Switzerland






— List of Figures — 73

Diffusion of a type: Deva [I] Schaanwald, Grabungen I Tholey-Sotzweiler I



Yards as indicators of social relations: Bondorf Chedworth I Katzenbach Koerich-Goeblingen Köngen Pforzheim-Hagenschiess St-Germain-lès-Corbeil Sontheim a.d. Brenz Vierherrenborn-Irsch



Unit-system villas of the Late Empire: Anthée, main house Aylesford-Eccles [F] Köln-Mungersdorf Milton Keynes-Bancroft V Norton Disney



Courtyards for kin or slaves? Anthée-yard Anthée, buildings 2, 3, 10, 15 Levroux-Trégonce Liédena Liestal-Munzach Noyers-sur-Serein Oberentfelden



= certain doorway = conjectured doorway = Mosaic

H he


= Hypocaust = hearth


This book is the culmination of an interest which began nearly fifty years ago when the late Sir Ian Richmond urged me, on the slender basis of attendance for two weeks at his Corbridge training excavation, to accept an invitation from the then Ministry of Works to dig the villa at Denton (Lincs.), at that time threatened by ironstone mining. Parallel interests in the history of houses, which entailed study of room function and of timber construction, soon revealed problems hardly considered hitherto by students of Roman Britain; and the potentially fruitful interaction of the three strands of my life’s work, conducted at considerable intervals and at very varying intensity, has at last produced the ideas set out in the following chapters. The approach here adopted, which is founded on the assumption that house plans reflect accurately the relations within and between the various groups or classes comprising a society, gives rise to a difficulty of presentation. Few people engaged in these studies have analysed plans in any depth and consequently few are familiar with more than a small proportion of villas in their own country – and here I refer to the major countries, Britain, France and Germany; the point applies with less force where there are fewer villas or where analytical gazetteers or summaries exist. Furthermore, hardly anyone has looked at plans from a functional standpoint, so that points of that kind made about even the few villas reproduced internationally, such as Köln-Mungersdorf, Montmaurin or Welwyn-Lockleys, require illustration for the present purpose. Because the approach is unfamiliar, all villa plans mentioned should, ideally, be illustrated, and generous aid from several bodies has enabled a considerable proportion of them to be presented. Nevertheless, every reader will deplore the omission of this or that plan, and to one and all I can only say that I have done my best. One hoped-for result of this work is the internationalisation of villa studies, something already done splendidly in a general way by John Percival’s The Roman Villa, but now in greater detail. Archaeologists presented with an unfamiliar kind of pottery or brooch will scour excavation reports and museum catalogues for the whole of the Empire until they find a parallel; it is time they did the same for villa plans. The attitude of many of them was summed up by a member of the regrettably short-lived Roman Villas Research Group who remarked to me: ‘I regard your function in the Group as that of gadfly.’ I hope this book will secure my promotion, in the eyes of that person and others like-minded, to the status of hornet. xxvi

— Preface — A few words are necessary to explain how the book was written. As the collection of villa plans proceeded, a thematic index of elements was compiled, commonly accompanied by very brief remarks scribbled – the word is used advisedly – on the photocopy itself. No separate notes were made on individual sites. When a new element was perceived, memory often brought to mind other examples whose significance had gone unnoticed, and from time to time the whole corpus of plans was combed for omissions. To interpret the many details in any one plan required repeated examination over a long time and, since the text was written directly from index and plans, constantly taking new material into account, some inconsistency arising from new ideas or changed opinions is inevitable. Even now, revision of the text with all the plans would sometimes produce a better example to make a point or permit interpretation of some overlooked feature. A book with roots as distant as this demands numerous acknowledgements, some to persons long dead. Foremost among the latter are my former history tutor and friend at Birmingham University, Philip Styles, an inspiring teacher to whom I owe an awareness of the use of architecture as historical evidence; K.D.M. Dauncey, also then of Birmingham University, for a stimulating introduction to Roman Britain; Sir Ian Richmond, who provided an opportunity far beyond reasonable expectation; Gerald Dunning, who encouraged me at Denton and was invariably helpful in Roman matters; and Edith Wightman, who sent offprints and drafts of articles and whose discussion of villa problems was enlightening. Among the living I am indebted to three people who, unlike so many scholars in the field, did not dismiss out of hand early outlines of my ideas. At the first presentation, to the Society of Antiquaries of London, Rosamond Hanworth and Richard Reece gave immediate support and have continued their encouragement. In the discussion following the Antiquaries paper Richard remarked: ‘I think we may be seeing the beginning of a new way of looking at villas.’ I hope the long-delayed result does not fall too far below that optimistic expectation. After the second presentation, delivered at a conference at Nottingham University, Malcolm Todd urged me to put into print, under his editorship, a paper which an eminent scholar had instantly dismissed as unbelievable (Todd 1978). Among others who since then have lent welcome support are Cary Carson (of Colonial Williamsburg, but with academic roots in Roman Britain), Simon James, Martin Millett, John Percival and Tim Potter. I have special reason to be grateful to Richard Reece for reading drafts of the whole of Parts II and III, and to John Wilkes for reading certain chapters. Adrian Havercroft also read and commented extensively on a few chapters, and provided valuable assistance in photocopying the whole of the text. Several other people have helped me over the years through discussion, the exchange of letters and the gift of offprints, notably Roger Agache, V.H. Baumann, Ernest Black, Wolfgang Gaitzsch, Herman Hinz, Fridolin Reutti and Franz Schubert. I am profoundly grateful to a friend and former colleague in the Royal Commission, Allan Adams, who undertook all the drawings at short notice after an initial setback. Sarah Brown and Anne Neville laboured to remove inconsistencies and obscurities in a complicated text drafted and revised over several years; I thank them warmly for their meticulous work. Any defects and mistakes that remain in text and drawings alike are my sole responsibility. xxvii

— Preface — Several foundations, institutions and libraries have aided my studies. On my retirement from the staff of the Royal Commission on the Historical Monuments of England, University College London very kindly conferred on me an Honorary Research Fellowship. The Woodrow Wilson International Center for Scholars at Washington DC granted a six months’ Fellowship to write much of the first draft. During that time occurred the only public presentation of the essential ideas, to which Professor A.G. McKay of McMaster University proved a kindly and helpful respondent. The Römisch-Germanisch Kommission at Frankfurt a.M. several times provided hospitality and access to an incomparable library of periodicals; I am especially grateful to Dr Eckehart Schubert for assistance of many kinds. At different times the British Academy awarded a personal research grant for travel and the Leverhulme Trust a Research Fellowship; and the Academy, the Marc Fitch Fund and the Robert Kiln Charitable Trust all made generous grants towards the cost of illustrations. The library of the Society of Antiquaries of London has been an invaluable resource; I thank two Librarians, John Hopkins and Bernard Nurse, and their staffs for much help. Other people too numerous to mention individually have provided information: some are thanked in notes; all, I hope, will accept this grateful acknowledgement of their help. Finally, I give most heartfelt thanks to my wife, Heather, for creating the conditions which made my work possible, for her interest in it over many years, and for helping me through discouragements; without her assistance and understanding this book would not have been finished.



Names of villas are those of the commune, parish, etc., followed as necessary by the local name and by the country or an administrative division of a country. For Germany the villa names given in RiBad-Württ, etc., are used, supplemented by reference to Müllers Grosses Deutsches Ortsbuch, 1982–3. Phases of villas are shown in Roman numerals, e.g. I, II, . . . where they follow the excavation report, and [I], [II], . . . where the phases represent the author’s opinion. [F] = final state. A phase preceding the first Roman villa is designated [pre-I] except where a succession of pre-Roman buildings follows the numerical sequence of the report. Houses on villa sites are designated by compass points or, where the original report is followed, by a number or letter, followed by a phase, e.g., Beadlam N I. Rooms retain any numbers given in the report or are numbered following the sequence of letters, hence Blankenheim 33. Drawings of houses are reproduced to a scale of 1:1000 except for a few particularly large ones which have a drawn scale. Courtyards and farmyards are generally reproduced to a scale of 1:3000. Exceptions have a drawn scale.



The following entries define only how terms are used in this book and may in some instances be additional to the senses given in OED. Aisled house: combines domestic accommodation at one end with working space or a byre in the remainder. Anteroom: a larger lobby; a room in its own right giving access to a more important one. Apartment: two or more rooms interpreted as forming a self-contained suite; an alternative to unit. Cell: a spatial unit of the plan defined by load-bearing (usually external) walls and transverse walls or partitions; the term takes no account of axial partitions. Corridor: a passage with rooms (one of which may be a porticus) on both sides. Courtyard plan: having buildings, not necessarily continuous, on three or four sides of a large open yard. Cross-wing: see Wing. Farm: used as the equivalent of the German Ackergut, Gutshof, Herrenhof, Meierhof and the French exploitation. Farmer: used as the equivalent of Bauer, Grossbauer. Galerie-facade: see Porticus-with-pavilions. Houseful: all the persons, of whatever status, inhabiting a house, which may be a single structure or comprise two or more structures. Household: a unit of consumption and reproduction corresponding to an elementary (or nuclear) family and its dependants, both relatives and servants. Hypocauston: a small hypocausted room heating adjoining rooms. Living-room: any room not serving a special purpose such as kitchen or diningroom; probably multi-purpose, providing for work, storage, eating and sleeping. Lobby: a room through which a larger one is entered; a means of approach without a major function of its own. Transverse lobby: a corridor closed at one end. Square (or small) lobby: a room about half the width of a cell in a row-house. Pavilion (tour d’angle, Eckrisalit): a squarish room at the end of a porticus. Porch: a squarish room, commonly within a porticus, giving entry to a building. Porticus (galerie, Portikus): a neutral term for a comparatively narrow roofed xxx

— Glossary — space adjoining an external elevation which in plan looks like and often is a passage or loggia but can function as a room or balcony. Porticus-with-pavilions: equivalent to Portikus-mit-Eckrisaliten, galerie-façade, maison à tours d’angles; intended to replace winged corridor. Representational room: assumed to have been used by a kin-group for feasts which reinforced group solidarity and had a religious element, as demonstrated by the presence of a shrine in the room (Newport) or in a smaller adjoining room (Kinheim); also an imposing room for ceremonies of mutual obligation, accompanied by feasting, between a grandee and his dependants. Row-type house (Reihentyp, linear-aufgereihten Haus): comprises a series of rooms, usually squarish, often interspersed with lobbies. Veranda (Laubengang): an open earthfast timber porticus. Villa: (1) the principal house(s) of a country estate or farm (= Herrensitz); (2) farm (Herrenhof, Meierhof, Gutshof). Wing: a long room or row of rooms assumed to be roofed at right-angles to and in front of the main range of a building. Cross-wing: one roofed transversely to the end of a building. Winged corridor: see Porticus-with-pavilions. Workhall (Wirtschaftshalle): a domestic hall, presumed to be for either a family forming part of a kin-group or farm servants, in which work, commonly smithing, is an important function.



Aus. Bad.-Württ. Bay. Belg. Berks. Bucks. Bulg. Cambs. Fin. Glam. Glos. Hants. Herefs. Herts. Hung. I.o.W. Lincs. Loire-Atl. Lux. Neth. Nordrh.-Westf. Northants. Notts. Oxon. Pembs. Pyr.-Atl. Pyr.-Or. Rhld-Pf. Rom. Saarld Seine-Mar. Som.

Austria Baden-Württemberg Bayern Belgium Berkshire Buckinghamshire Bulgaria Cambridgeshire Finistère Glamorgan Gloucestershire Hampshire Herefordshire Hertfordshire Hungary Isle of Wight Lincolnshire Loire-Atlantique Luxembourg Netherlands Nordrhein-Westfalen Northamptonshire Nottinghamshire Oxfordshire Pembrokeshire Pyrénées-Atlantiques Pyrénées-Orientales Rheinland-Pfalz Romania Saarland Seine-Maritime Somerset xxxii

— Abbreviations of Locations — Staffs. Switz. War. Wilts. Yorks. Yugosl.

Staffordshire Switzerland Warwickshire Wiltshire Yorkshire Yugoslavia







nyone confronted with a new book on Roman villas must ask whether anything new can be said, given that two Empire-wide studies exist, one by A.G. McKay of every kind of house and another by John Percival of the villa as a social and economic institution;1 and these are supplemented by innumerable regional surveys. Yet the question can be answered positively, for this book has an entirely different purpose and method. Its primary aim is to classify and make intelligible the innumerable and extraordinarily varied villa plans excavated over the past two hundred years. Villas form a large body of evidence which at present is either misused or totally neglected: a failure unwittingly revealed in Webster’s remark that ‘most villa excavations have been very scrappily recorded . . . short of physical re-excavation there is little that can be done significantly to increase our knowledge’.2 This unwarranted pessimism, founded in ignorance of the history of villa studies, provided a spur to prove the contrary. A second aim, to understand the structure of rural society in the Roman Empire, was, like the first, limited initially to the provinces of north-west Europe. As villa plans were collected for comparative purposes, it became apparent that sufficient resemblances existed over the whole length of the Empire from Wales to Bulgaria to make some comment on social structure possible for those areas peripheral to the main theme. The evidence is plentiful for Belgium, Britain, France, Germany, Holland and Switzerland, and in these countries sufficient sites have been dug to their lower levels to establish broadly how villas developed in that large part of the Empire. Plans are not lacking for Spain, Austria and Hungary, although phases other than the latest have rarely been explored. The sparser evidence for Romania and the former Yugoslavia, taken overall, reveals a different pattern which, nevertheless, contains a few links with that of the north-west provinces, and even Bulgarian villas have some connections. Brief enquiry into the villas of Italy and Greece suggested that they add little to the theme, although the former country will certainly be enlightening when more villas are properly dug. Throughout, villas are dealt with by modern countries because neither villa types nor their modes of development are limited to particular provinces and consequently it is less tendentious as well as less demanding of geographical knowledge of the Roman world if they are kept in a modern framework. 3

— Chapter One — The analysis of ground plans, which will be described in the following chapter, is strongly influenced by the study of medieval and early modern houses as it has developed in Britain during the past forty years. In that field, recognition of the structural complexity of old houses arising from their successive adaptation and enlargement has led to a strong emphasis on forms of plan and has produced subtle ways of determining and interpreting their development. In vernacular architecture this has resulted in the establishment of a series of types characteristic of particular periods and capable of correlation with social classes or groups; the intention is to devise a typology for the Roman Empire though without periodisation. Links of this kind enable historians to add architecture to the sources for social history rather than using it as a mere backcloth to a narrative based on documentary and literary evidence; this is proving to be true of periods for which the written sources are far more abundant than they are for the Roman Empire. Since the vast majority of villas can only be known through excavation reports, the application of the methods used in vernacular studies is limited to certain kinds of inference and particularly to the recognition of plan types. In a historical period in which hardly more than the foundations of houses remain it may be hoped that an awareness of types will lead to more sophisticated analysis of villas, not only by historians interpreting them but also by the archaeologists who dig them and who need hypotheses to test in the course of excavation.

PLAN TYPOLOGY: SOME OBJECTIONS Plan, it has been asserted, is not the only determinant of a building’s historical significance – which is self-evidently true – nor even the most important,3 which is certainly untrue. It goes without saying that style, ornament and decoration have much to tell us about the social position of the person who commissioned a house, yet archaeologists rarely realise how closely their interpretation is bound up with plan; the study of mosaics, for example, which has been so largely concerned with dating, technique and schools of craftsmen, has a strong bearing on the uses and relative importance of rooms and the understanding of how villas functioned. Little will be said about mosaics because I am not sure what the social as opposed to the iconographic significance of individual examples is, although their more obvious implications are sometimes used to elucidate a plan. An objection sometimes made to house typology as social history is that a rich man might choose to build a comparatively small house or a poor man build beyond his means, so invalidating any close correlation between size or form of house and social class. Clearly, exceptions to that rule did occur, and indeed an instance from seventeenth-century England establishes the truth of the general proposition that a man’s wealth and social position and the size of his house were closely linked. Thus the Nonconformist divine Richard Baxter records admiringly how his friend the Lord Chief Justice Sir Matthew Hale refused to conform to this social norm: ‘His garb and house and attendance so very mean and low . . . that he was herein the marvel of his age.’ Roger North put a different gloss on this situation: 4

— Aims and Scope of the Book — I shall not forgett a relation I heard the Duke of Beaufort make of the advice his country-man the Chief Justice Hales gave him when he was building at Badmanton, to have but one door to his house, and that in the ey of his ordinary dining room, or study where he past his time. This shews how all men measure things by their owne education and circumstances, and expect others should governe their actions accordingly, tho farr from the like engagements. North was in tune with his times, Baxter not, as one of Samuel Pepys’s observations confirms: ‘all men do blame him [Clarendon] for having built so great a house, till he had got a better estate’; this of the great Lord Chancellor, the dominant political figure of the early Restoration period, to whose downfall pride in what another diarist called his ‘new built Palace’ contributed much.4 Exactly the same attitudes governed building in every age prior to the late twentieth century. A house was one of three kinds of display or conspicuous consumption, the other two being dress and entertainment (Baxter’s ‘garb’ and ‘attendance’), which together provided the means whereby a man established his power and rank in the eyes of his contemporaries.

THE TERMINOLOGY OF SOCIAL STRUCTURE Whether the model of society proposed here finds acceptance or not, it can hardly be doubted that something better than the present one is needed. Most scholars who have studied villas have implicitly equated Roman society with their own as it was in the heyday of the country house or château or Schloss about a hundred years ago. It is a society dominated by landowners great and small operating a market in which land is freely bought and sold without reference to any constraint of law and custom and whose lives mirror those of late nineteenth-century country gentlemen. Beneath them is a class of farmers whose more prosperous members aspire to the higher status of gentleman landowner, whilst the others are tenants;5 and an insular aberration sees many British houses occupied by bailiffs with, below them, the ultimate absurdity of the cottager in the ‘cottage house’. One notable proposal has been made to reform this inadequate framework. A French scholar has suggested that if reference is made to such categories as country or manor house it should be systematic, using the whole range of terms describing the houses of the society selected as analogous – implicitly, that of the European Middle Ages or early modern period.6 The idea is attractive; the difficulty is to apply it. It assumes, first, that the later society is comparable in detail to its Roman predecessor, yet it cannot be taken for granted that such terms as manor house, farmhouse and cottage correspond to the social reality of Roman houses. Second and more fundamentally, it assumes that the structure of the medieval or early modern society in question has been correctly understood from purely documentary evidence, something that, for England and Wales at least, is open to question.7 Even what is often taken to be the perfectly neutral term ‘farmhouse’ has historical connotations: what kind of farmer do we envisage – a peasant farmer, a tenant analogous to a husbandman or yeoman (and if so, on what terms), or a freeholder? These are only 5

— Chapter One — the British problems; a German would interpret a farmhouse as a combined shelter for family and animals. As far as possible the word ‘farmhouse’ will not be used below and when it appears will have only the minimal sense of the dwelling place of a person engaged in agriculture. Any deductions about social structure rest on three principal categories of evidence. The first comprises classical texts: the problems of reconciling them with archaeological evidence have been encapsulated in the felicitous phrase ‘text-hindered archaeology’ and are exemplified by the endless argument about the reliability of Julius Caesar’s description of British and Gaulish society. Inscriptions form a subdivision of this category – a more reliable one but of limited scope. A second comprises the objects found in the course of excavation. Few illuminate the use of rooms, and the difficulty of interpreting them is illustrated by the contrast between fine objects, especially those connected with feasting or dress, and the simple houses in which they are sometimes found.8 The third is provided by villas or contemporary indigenous houses, all of them requiring interpretation. My own view is that those house plans where the grouping and intercommunication of rooms is known or can be inferred provide objective evidence of social organisation, showing how people actually lived and how social relations changed. It is a view based on research over the past fifty years into Welsh and English houses. First, in 1942, certain gentry seats in west Wales were found to comprise three or more independent houses instead of the one that would normally be expected. The existence of a seventeenthcentury kin-group thus implied was taken up a decade later to account for the presence in some farmyards in south-east Wales of two houses. Subsequent fieldwork established that the same phenomena are widespread in England at various social levels.9 Now joint habitation by kin-groups in these two countries at so late a date had never been suspected by historians despite intensive research, nor even by genealogists working in Wales where family and descent have been studied as keenly as anywhere. If house plans can produce surprises of this kind in a period incomparably better documented than the Roman Empire at any time, their primacy as a source of social information ought to be accepted over the uncertain weight of literary evidence; and that is the line taken throughout this book.

THE HISTORY OF VILLA CLASSIFICATION Most historians recognise that villas form a potentially important source of information; the problem is how to use the material in bulk, rather than picking out particular villas to make an impressionistic point. Classifying villas into types makes this possible, provided always that the typology has a definable social or economic significance. Thirty or more scholars in several countries have attempted classification and it might be expected that their cumulative efforts over more than a hundred years would by now have achieved a satisfactory conclusion. Far from it, as a brief review will show. Arcisse de Caumont appears to have been the first person to study villas as a distinct class of building, bringing together many French and English examples and relating them to classical texts. Many were large and, since nobody could divide 6

— Aims and Scope of the Book — them into structural phases, they were too diverse for types to be recognisable. German scholars took the first tentative steps in that direction, beginning with the energetic first director of the Trier museum, Felix Hettner, who, in 1883, in the course of a wider survey, divided villas into two classes: those with an internal courtyard, including Fliessem and most of the smaller villas; and the pleasure or luxury house (Lusthaus), such as Nennig and Oberweis. Hettner, though, like de Caumont, was hampered by treating complicated plans of large buildings – as many villas then known were – as a single building campaign.10 Towards the end of the nineteenth century the establishment of the Reichs-LimesKommission resulted in a number of villa excavations in the Roman frontier areas of Germany which produced mostly simple, little-altered plans. Using four of them, Schumacher recognised a type defined by two common elements, a gallery or portico (Halle) and, behind it (following Hettner), an open yard flanked by rooms. A few years later Haverfield, summarising current knowledge of Roman Hampshire for the Victoria County History, coined the term ‘corridor house’ to denote a range of rooms joined by a front corridor or portico. He also recognised a kind of rectangular building divided internally by two rows of columns which, a few years later, was named by Ward, not altogether happily, the basilican type. Otherwise Ward took over from Haverfield ‘the ordinary or “corridor” type’ and the courtyard type of villa built around three or four sides of a large courtyard. Indigenous houses persisting in the Roman period he called ‘cottages’.11 Meanwhile, two more German scholars addressed the problem of villa typology. Anthes, seeking the origins of villas in Germany, found the direct Italian inspiration he expected in the north-western provinces only in a peristyle ‘house’ at Caerwent – a building which subsequently elicited the qualification ‘if a private residence at all’. His almost incidental comment on the great villa of Haut Clocher-St Ulrich, that it was of three distinct parts and gradual growth, was a key to progress not otherwise used. Kropatscheck went further in the same ultimately sterile directions as Anthes, dealing with a large number of villas in neighbouring countries as well as Germany and attempting to relate them all to Italian models and the writings of Vitruvius. In a brief summary of Ward’s book he rejected the idea of the basilican building in favour of an elongated internal yard gradually built over, thereby assimilating the type to the many German villas then invariably interpreted, following Schumacher, as a courtyard with rooms on two or three sides and a gallery at the front.12 Kropatscheck’s article is the culmination of many years’ work in Germany, England and elsewhere. A decade later much of it was obsolete. The first step towards a clearer understanding came in 1918 with a remarkable book by the young Austrian art historian Karl Swoboda which proclaimed itself to be, not altogether accurately, about Roman and Romanesque palaces. This was the first study to embrace the whole of the Roman Empire and its importance for the present purpose lies, first, in its singling out the porticus-with-pavilions (Portikus-mit-Eckrisaliten) as an architectural feature common to villas everywhere; and, second, in the recognition that this element was a display feature, at once a mark of status and a proclamation of Roman values. The second and even more important step was taken by a German archaeologist in whom an incomparable knowledge of the architecture of the ancient world was applied to an understanding of the complexity of buildings and an interest in how 7

— Chapter One — they functioned. Franz Oelmann’s mind is the keenest that has ever been applied to villas and in his paper on the villa of Stahl and its analogues he achieved an astonishing feat of pure typological study, one which transformed the subject in Germany and Holland. His analytical method is considered in detail below (Chapter 3, p. 24); by it he established that what had been thought of as a yard behind the porticus was a large hall. Within six years this hypothesis was fully confirmed by his own classic excavation of the villa at Mayen.13 In these two publications Oelmann provided for the first time a method of analysing villas and an effective way of using them as documents of social history. A few years later Fremersdorf put the study of villas as agricultural enterprises on a firm footing at Köln-Mungersdorf with the total excavation of a house and its outbuildings; it was the culmination of a long-standing interest among his countrymen and marks the high point of German villa research. At the same time Paret, in comments on Remmingsheim and other villas, applied Oelmann’s ideas briefly but failed to develop them systematically.14 No progress has since been made with typology in Germany, although some brief remarks by Samesreuther suggest that the Second World War deprived us of a scholar willing to tackle the fundamental problems of villas and keen to follow the lead given by Oelmann and Fremersdorf.15 Not that scholars elsewhere have done any better. In Britain Collingwood, aware of Swoboda’s book and the Mayen report, ignored their implications in producing a purely formal classification of no historical significance. Calling the porticus a corridor and the pavilions (Eckrisaliten) wings – both names misrepresent the architectural form and function of these elements – he created two types, the bipartite corridor house with only a front corridor and the tripartite with one at the back as well. Courtyard and basilican houses he took over from Ward. Meanwhile, in France, Albert Grenier summarised Oelmann’s articles and also his earlier report on the Blankenheim villa. Awareness is one thing, understanding another: the two pages reproducing the plans of Stahl and its analogues fail to grasp their significance and the accompanying comments reflect Swoboda’s conclusions rather than Oelmann’s. Regrettably, neither Collingwood nor Grenier understood sufficiently the method so brilliantly expounded by Oelmann to apply it themselves. That was left to de Maeyer, who analysed a few hall-type villas on the lines of Mayen, something he could hardly avoid doing since the Belgian villa of Serville figured in the 1921 paper.16 Since 1945 several British archaeologists have endeavoured to improve the classification of villas, confining their attentions to insular sites; the results are deplorable. Nash-Williams’ schematic ‘evolution of the Roman villa-plan’ was futile. Hawkes, going flat contrary to Harmand’s view that modern terminology should be applied consistently, misapplied the term ‘cottage’ at much the same time as Richmond did in his revision of Collingwood. This was particularly depressing in the light of Richmond’s keen understanding and appreciation of many other kinds of architecture, not only Roman, as is his failure to develop a few tentative remarks about hall-type villas and their resemblance to those in Germany. Worse was to come. In the mid1970s Branigan stepped smartly backwards in interpreting several British villas as having what he called intra-mural yards, a point he clinched without further argument by reproducing several of the plans from which Oelmann had drawn precisely the opposite conclusion.17 No further attempt has been made to classify plans. 8

— Aims and Scope of the Book — Europe outside Germany was more fortunate in these years. Drack, in assembling all known plans of Swiss villas, coined the term ‘row type’ to describe what British archaeologists call a corridor house, thereby recognising that the significant part of a house is what lies behind the façade. He applied Oelmann’s method, unargued, to virtually all reasonably complete plans and, had the conclusions been published with a commentary, they would have had the impact they deserved and have affected for the better the course of both Swiss and German villa research. Agache, who appears to have been unaware of Oelmann’s Stahl article, analysed plans of villas in the Somme basin comparatively, the first time anyone had done this from air photographs and not the least noteworthy aspect of a very remarkable achievement. This sunny picture had a dark side. The study of Spanish villas, many of them only partially excavated, produced in typological terms one step forward and one back. Maria Cruz Fernando Castro took the step forward in putting them into more or less convincing groups. Three years earlier J.-G. Gorges had produced a pretentious classification which deserves to be widely known as an awful warning and which mars an otherwise informative and well-organised book. Lumping together into one table several typologies – those of Swoboda, Richmond, Grenier, Agache – and adding to them the notions, for they hardly amount to typologies, of E. Thomas, M. Biro and G.A. Mansuelli, then decorating the columns with schematic plans of varying degrees of improbability, Gorges produced a result best described in words once applied to a work of Latin scholarship: ‘What it most resembles is a magpie’s nest.’18 In the light of the author’s all too briefly expressed insights into the spatial organisation of peristyle villas he should have been equally independent in classifying plans. Elsewhere, Vasic failed to find types in Yugoslav villas; three authors commissioned to survey Germania and Gallia Belgica did no more than heap up materials; and with Nikolov’s opinion that ‘no meaningful classification of villas is really possible’19 Webster’s know-nothing attitude finds its continental match. I would not be so scornful of some of these performances if Oelmann’s penetration and insight had remained confined to the comparatively obscure 1921 paper, but most emphatically they did not. For many years past no archaeologist concerned with villas anywhere can have failed to become aware of Mayen and, once the report is consulted (even via Grenier), the plans there reproduced from the earlier paper establish that Oelmann’s theoretical approach had that rare attribute, the power of prediction. This was surely one of the greatest triumphs of typological reasoning and must have a powerful impact on anyone who reflects on its implications. Going back to that seminal article, which parallels the ways developed by architectural historians to interpret church and house plans alike, is an inspiration few have found.

VILLA ARCHITECTURE It may be asked why so little attention is given in this book to the villa as architecture. The reason is simple: too little is known about the matter, for all that a large number of drawn reconstructions have appeared in print. Those of Pannonian villas form what may be the largest group so far published; most are fantasy. At least one famous reconstruction, Mylius’ of Blankenheim I, defies the specific evidence of 9

— Chapter One — excavation adduced by Oelmann without, apparently, any realisation of the fact on the part of those who have reproduced the drawing subsequently.20 Hardly any reconstructions take account of the precise form of foundations, their authors taking refuge in drawings which ignore problems of variation in width and construction. As for roofs, the appearance of the cladding, whether of tile, stone slates or thatch, is the only aspect considered; how they were supported, who knows or cares? In drawn reconstructions of timber buildings archaeologists have solved the problem of continuity from Roman to medieval many times: villa houses in Sussex, FrancheComté or the Rhineland have exposed framing looking exactly like that of their respective sixteenth-century counterparts. The current vogue in Britain is for villas of two or more storeys. So strongly is this idea canvassed that in conversation one eminent Romano-British archaeologist offered the luxury villa of Oplontis on the Bay of Naples as reason for thinking that British villas too might have had upper storeys. In fact, no evidence of an upper storey has yet been found in any of the many villas of the province except over the odd room and, indeed, not a scrap of positive evidence for the necessary staircase has been found anywhere except Spain, Italy and North Africa. The structure and appearance of villas need much detailed study before generalisation for one country is possible, let alone Europe generally. For Britain some recent papers21 give hope that these matters may one day be tackled with authority, but until then I prefer to stick to a subject on which it may be possible to make some progress. If exclusion of every aspect of villas but the plan needs more justification than this, the reader should consider the crushing burdens further work would impose.22 To deal systematically with their topographical setting and the light it throws on the form of the buildings, or with their economic function – that is, to go beyond the valuable generalisations made for so many parts of the Empire by Rostovtzeff, Percival and many other writers – would require local research so detailed as to impose an impossible burden on an individual. Similarly, to relate them to particular kinds of agricultural production may be thought important. Whether plans are much affected by considerations of this kind, apart from the presence of animals in some part of the house or the ‘agricultural factory run by slaves’,23 is quite uncertain; the many corndriers alleged to have existed in British villas did not noticeably affect their planning, which can be explained on purely domestic lines. Storage of crops, notably wheat on the stalk, is a possibility suggested by early modern German and Dutch farmhouses but one that does not seem to have left recognisable archaeological traces. Only in olive-growing regions have the presses required for production of the oil-modified house plans in a distinctive way and only a handful of the villas noticed below come into this category.

WHAT IS A VILLA AND WHAT IS ITS NAME? Every effort has been made to discover villa plans. Villa in that context corresponds generally to Edith Wightman’s sensible definition: ‘all farms or country-houses built at least partly in stone’.24 In the light of recent discoveries, though, even that may not be quite wide enough. What do we call the group of rectangular timber buildings 10

— Aims and Scope of the Book — laid out on three sides of a courtyard at Druten in Holland? This pre-eminently Roman form of planning is common in French and British villas and is quite unlike native farmsteads. And what about those settlements acquiring the beginnings of Roman buildings and manners such as Harting-Garden Hill or Barnsley Park IV, where bath buildings appear among round or rectangular timber structures? They hardly qualify even under Wightman’s broad definition and, since the main house is not of stone, may be referred to, like Druten, as proto-villas. In the end it hardly matters; an historical approach must treat pre-Roman and Roman buildings and settlements as a continuum. Even in the context of provincial-Roman societies the word ‘villa’ presents difficulties, meaning at different times the house of a farm or other establishment; the house and the adjoining buildings within an enclosure or courtyard; and the entire establishment, land and buildings. Here ‘villa’ is used principally in the second sense and houses are called just that. Complete consistency of usage has certainly not been achieved because so often the house is the only thing excavated on a villa site and it is hard to avoid referring to it simply as the villa of ——. Then there is the question of nomenclature. How should a site be referred to: by its most localised name, that of the field or hill on which it stands, or by the name of the commune or parish? British archaeologists tend to prefer the first option and so the reader is confronted with names like Spoonley Wood, Wadfield, Lockleys and Dicket Mead, whose geographical location someone working at a distance – I have in mind an archaeologist working in the Danube delta – may find difficult to discover. The general rule, though, is that the city, commune or parish name prefaces the local one, hence that has been taken as the appropriate standard form.25 The advantages of this system have been demonstrated in Roger Agache’s many publications on the Somme villas. It is useful, on seeing a plan of Warfusée-Sud (or -les Terres Noires) which is about 370 m long, to be reminded by the name that another villa almost as large, Warfusée-Nord (or -le Petit Chêne et le Chaufour), is not far away. That some parts of Britain may have been almost as densely covered with villas as the Somme basin is now generally recognised, but for some British and all continental students of the subject the question will the more readily be called to mind by SudeleySpoonley Wood and Sudeley-Wadfield or Welwyn-Lockleys and Welwyn-Dicket Mead, not to mention the greater ease of finding where they are. Although the precise location of villas is not important for the present study, the occasional need to learn the whereabouts of one in, for example, Germany, has brought an appreciation of the merits of continental nomenclature and awareness of the difficulties the British way of doing things must sometimes present to scholars in other countries.

A REPRESENTATIVE SAMPLE? How firmly based is this work in terms of the numbers of villas considered? The total number of plans reviewed is roughly 1,100, including some fragmentary and uncertain examples. Many plans must have been overlooked but their involuntary omission is not likely to have affected the conclusions drawn. Probably a reasonable proportion of the villa plans in every one of the various countries has been looked at, especially 11

— Chapter One — those where a fairly recent descriptive catalogue of all known plans exists. Such are Spain and Portugal, Switzerland, Romania, Bulgaria, Hungary – more precisely, the province of Pannonia – and Holland, to which may be added a fifty-year-old work on Belgium and a more recent but unimportant catalogue for Luxembourg. For Austria and former Yugoslavia it is hoped that a reasonable proportion has been found. These countries account for about a quarter of all the plans examined. Britain, Germany and France have more or less equal shares in the remainder, probably in ascending order, although the difference – and it is only in the number of usable plans – is comparatively small. Regrettably, but hardly surprisingly in view of the magnitude of the task, no publications with plans and commentary cover these countries, although for the first two they exist in unpublished form. Britain now has a comprehensive catalogue of all published and many probable, possible and doubtful sites. For France the Carte archéologique de la Gaule is rapidly providing comprehensive departmental lists, often accompanied by plans, and for northern Gaul as a whole the situation is redeemed by a recent study dealing with the rural sites of Late Antiquity.26 Every plan has been considered carefully, most several times. At first many were unintelligible but, as understanding grew, the number put aside diminished until most of those now remaining are simply too fragmentary to make any sense of. Ideally a catalogue should list all plans examined, with a summary interpretation of each. Space prohibits that but a full list of all those mentioned, with the source of each plan, is appended (pp. 340–59). How representative are they and what proportion do they form of the total number of villas at present known to have existed? The second point could be calculated laboriously from the gazetteers accompanying more recent sheets of the International Map of the Roman Empire. Only for Britain is it easy. The latest edition of the Ordnance Survey Map of Roman Britain lists 275 villas and 285 ‘other substantial buildings’, a total of 560. Of these about 200, or rather more than a third, have plans sufficiently intelligible to be assigned to one or other category of house plan, even if sometimes the detail is obscure. It can reasonably be regarded as a representative sample of the whole. There is little point in guessing the corresponding fraction of German villas; although certainly considerably smaller, it is probably large enough to be representative. France, with its innumerable villas, presents the real problem, especially southern France, corresponding approximately to the provinces of Aquitania and Narbonnensis, to which can be added Spain and Portugal. In these countries big villas of great complexity are common and in Spain are the norm. Until several of them have been thoroughly explored down to the lowest levels the course of their development will remain obscure; consequently their treatment in this book is bound to be unsatisfactory.






n dealing with so large a body of plans, hardly any two of which are alike in their final form – and it is only the final form that is known in most instances – the first requisite is to establish types and the second is to recognise that a typology is not a series of pigeon-holes in which to place identical examples. Here experience in the vernacular field, where exactly the same problem has been faced, can offer a solution. It appears that English houses both great and small are usually the product of successive partial rebuildings, with the frequent consequence that they do not conform closely to a type but rather resemble it in the number and relative size of the various rooms: a consideration likely to apply to all domestic buildings that survive for a hundred years or more. Viewed in these terms a house type is an aggregation of functions, largely inferred, which approximates to a particular form of plan; a latitude permitting houses varying in detail to be grouped together, whether of several building phases or only one.

ORIGINS OF PLAN ANALYSIS An approach of this kind is implicit in Swoboda’s demonstration that the porticus – the architectural term he preferred to the earlier Halle and here used in its fairly general Latin sense – was a standard form of architectural display added to many kinds of house plan. Oelmann developed this line of argument to different purpose in relation to villas which have in the middle a sizeable space which was at that time regarded as a yard.1 By stripping away rooms which appear to have been added around the central space or to intrude upon it he demonstrated that the middle space must have been roofed and was therefore a hall.2 Now this is a mode of reasoning applicable to any kind of villa, because what really matters in a house is what lies behind the formal front; and it follows that the expression of Roman culture, the porticus-with-pavilions, ought not to form part of a plan typology. The pavilions (Eckrisaliten) themselves present more problems than a reading of the literature might suggest, and they will be considered separately.3 13

— Chapter Two — In his classic paper of 1921 Oelmann opened the way to a radical new approach to villas of all kinds, one in which the basic elements of facade, baths and core rooms (Kernbau) could be treated separately and more fruitfully, yet the method continued to be applied only to halls. This was certainly due to the lack of a concept capable of establishing the core of complex groupings of comparatively small rooms; and in Germany political considerations probably deterred some.4 Nevertheless, at least one attempt was made to analyse a villa in terms of its room groupings, an attempt which owes nothing overtly to Swoboda’s or Oelmann’s ideas. This is Koethe’s analysis of Oberweis, in the course of which he pointed out that like groupings occur on both sides of the central axis of this large villa, which thus combined a measure of symmetry in plan with that of the elevation.5

RELEVANCE OF VERNACULAR ARCHITECTURE STUDIES With these observations Koethe provided a foretaste of a kind of analysis not undertaken again until the late 1970s, when attempts were made to discover recurrent distinctive room groups in row-type villas (defined in chapter 4), of which Oberweis may be regarded as a very grand example. They rested on a different basis, that of vernacular architecture. Two papers by the present author made a modest beginning to the enterprise; they were succeeded by Drury’s bolder demonstration of how this might be done, although he failed to find clinching arguments in favour of the various possibilities.6 Recurrence of distinctive room groupings is one key to the social analysis of villas. A second is the relation of principal elements of an open hall or of a room to one another. By this is meant, for example, how the open hearth in a hall or the fireplace in a room is positioned relative to the entrance, and how it relates to the whole floor space; is it in the middle or near one end, and if the latter, what are the implications for the use of that end which receives little warmth? And where the position of fireplace or hearth is known, its position relative to any inner room and particularly to the doorway leading into it are important. This way of looking at plans comes directly from the study of medieval and early modern houses, in which characteristic lines of movement can be discerned. To the corresponding patterns in villas can be added the evidence of domestic religion, which can reveal, through the position of shrines, the ways in which rooms were used and approached. To many people those remarks will appear excessively optimistic, for the most serious difficulty in studying villa plans is ignorance of the positions of doorways and fireplaces or hearths. It goes without saying that intercommunication between rooms is the single most important key to the way they were used, a point recognised unhelpfully in reconstructions bespattered with doorways arbitrarily, without supporting argument, such as Basse-Wavre (Belg.). Students of villas, faced with no more than footings, have tended to despair of ever knowing how they functioned. This, though, is a matter on which comparative study can throw light, because some villas in France, Germany and Switzerland, of types found also in other parts of Europe, remained to a sufficient height to reveal a few doorways and in a few cases nearly all of them. 14

— Methods and Assumptions —

THE PRINCIPLE OF ALTERNATE DEVELOPMENT All villas occupied for any considerable length of time were altered and enlarged, often several times. Archaeologists take this for granted and indeed the process may be regarded as a natural response to changing social needs, so that a little-altered villa of long duration presents a serious problem of historical explanation. Nevertheless, the implications of rebuilding or addition merit more discussion than they commonly get, and again a concept drawn from the study of vernacular houses is useful. Farmhouses of any considerable age are almost invariably of two or more building phases, as has been observed in several European countries. One process by which such change takes place is called ‘alternate development’, meaning that only part of an existing house is rebuilt at any one time. This often results from a change of function in one part or end of the house, especially where a part devoted to farm purposes is converted to domestic use.7 Romano-British aisled houses, which appear to have begun as a form of long-house – a combined shelter for family, stock and crops – are a generally recognised instance of such change. Alternate development can also occur when, through the addition of a wing or other form of extension, the older part declines in importance, after a while appears old-fashioned, and is then altered if not rebuilt. This phenomenon is observable in the smaller country houses where rebuilding begins with the domestic or upper end and only later is the lower, service end rebuilt – sometimes not at all. The division of medieval open halls into smaller rooms is also well known; it can be paralleled in Roman times, though not with the insertion of an upper floor such as is usual in early modern England, and so can the total rebuilding of the hall. It is not that changes of these kinds have gone unnoticed by archaeologists, simply that no underlying principle has been recognised and commonly no explanation is offered. In drawing on parallels from late medieval and early modern England it is not the particular historical period that matters but rather the architectural forms, the ways in which they were used and were changed, and their social implications. It is not intended to suggest that English society was closely comparable to Roman Britain in its social structure, though it may be added that the two had more in common than Roman Britain has with the all too frequent model of Victorian England.

ROOM USE Buildings in the Roman period were often enlarged without undergoing substantial alteration to the core; this is a process observable in English country houses and farmhouses too, and in both cases houses often grew to a very large size. What were all the extra rooms used for? Two related concepts may sharpen the analysis of houses in an era for which documentation is confined to a few inscriptions. One is the observation made in connection with the English gentry in the late seventeenth and eighteenth centuries, that they needed an irreducible minimum of living-space comprising an entrancehall, four living-rooms, bedrooms and accommodation for servants; any more rooms than these were intended for the hospitality and display essential to maintaining social position among peers – ‘parade’ in the vocabulary of the day.8 15

— Chapter Two — This notion that only a certain number of rooms are necessary to maintain different facets of a way of life can be applied to villas to see if anything of the kind can be inferred in the very different circumstances of provincial Roman life. The second concept is that a social hierarchy is reflected not only in the size of buildings, from palace to cottage, but also within the buildings themselves, whether by a hierarchy of rooms graduated by size and appointments or by the disposition of fittings and furnishings within rooms. This may appear self-evident but the corollary, that a society whose buildings do not display marked differences is more egalitarian, is rarely recognised. Taken in conjunction with the notion of minimum room needs for any given level of society at any particular period, it may help to distinguish between changes which correspond to those in the social structure and those reflecting increased prosperity.

HOUSE PLANS AND SOCIAL STRUCTURE Running through all that has been said so far is the assumption that house plans, provided they are interpreted correctly, reflect the reality of social structure, irrespective of whether this goes counter to contemporary laws or descriptions. The point, adumbrated in chapter 1, can be amplified from early modern Wales, where the laws imposed after the Acts of Union of 1536 and 1542 abolished partible inheritance, otherwise known as ‘gavelkind’. Theoretically, the equal division of estate and goods among heirs ceased, yet the houses suggest that sharing continued without the parcelling out of the land into ever smaller and less economically viable farms that made the practice so ruinous. Two examples show the conflict between the letter of the law and the custom of the country. Thus in north Wales, at Park, Llanfrothen, where documentary evidence suggested an orthodox gentry family headed by sheriffs and justices of the peace, something like a kin-group existed in the late seventeenth and early eighteenth centuries, occupying both the great house and several small ones close by.9 The second example is from south Wales: For a long time the whereabouts of Edwinsford-uchaf had puzzled students of Carmarthenshire documents. The wretched house simply could not be found . . . a . . . survey of Edwinsford . . . revealed there were two houses on the site embedded in a mass of Victorian accretions which enclosed them both. Edwinsford-uchaf was in fact next door to Edwinsford-isaf; at one corner the two houses touched.10 These examples of apparent conflict between the written and the architectural record are particularly significant for Roman villas because parallels exist in the Roman provinces for both the architectural solutions adopted to tackle the inheritance problem – the problem of conflict between fairness towards all descendants – and the overriding need to preserve the economic viability of the family land. That does not preclude other interpretations of the same architectural phenomena, provided equally informative historical circumstances occur in conjunction with them. Yet if so fundamental an aspect of family organisation remained undetected in the written 16

— Methods and Assumptions — sources, and indeed remains so even now, as a structure in the Braudelian sense – something taken for granted, not needing comment or explanation – how much more likely that it and others are undetectable in the scanty information we have for the Roman provinces. So far I have suggested that sixteenth- and seventeenth-century house plans in England and Wales or, more specifically, elements of plan, may help to explain analogous (but not identical) features in Roman villas. To this it has been objected that comparative interpretations of this kind ought to be tied down reasonably closely to one period of time: that a plan of part of a medieval hall-house means something quite different about social structure from a similar plan which is part of an eighteenthcentury stately home. This raises the spectre of uniformitarian assumptions about which people really must be more explicit.11 It is, of course, true that a medieval hall, open to the roof, heated by an open hearth, and forming the focus of household life as well as the point of entrance, was very different in function and social significance from the similar-sized and equally lofty space, heated by a wall fireplace and with an elaborate ceiling, which likewise served as an entrance and point of reception for strangers in an eighteenthcentury ducal mansion; but the relation between the two rooms and the rest of the respective houses is so utterly different that confusion of function, in any reasonably subtle analysis, is impossible. With some plan elements such as lobbies or inner rooms, the problem does not arise; they reveal their function by their relation to larger rooms or entrances, even where lines of movement can only be discerned in a general way. Moreover, if sufficient examples can be found, one or more are likely to have archaeological evidence of function which, used judiciously, can throw light on the others. In fact, as indicated earlier, no overall comparison between provincial-Roman and early modern Anglo-Welsh society is intended. Given that house plans at both periods are essentially a series of rectangular blocks subdivided in various ways, the occurrence of similar features in a comparable relation to one another at both periods suggests a similar mode of use. Two front entrances where one would normally be expected is an example; and, although that may be thought too obvious to be ignored, archaeologists nevertheless managed to avoid discussing its implications at Maulévrier (Normandy), Sinsheim (Bad.-Württ.) and Chedworth (Glos.) for all the years since they were published in 1836, 1846 and 1868 respectively, and architectural historians have been equally blind.12 The occurrence of the same kind of architectural change taking place in the Roman provinces and early modern England argues that similar forms of change took place in both societies. In fifteenth-century England a house incorporated a principal room or hall heated by an open hearth and open to the roof. During the sixteenth and seventeenth centuries the principal room of a new house had a ceiling and a chimney-stack, and old houses were either replaced or altered to conform to the new pattern; only a few survived for hundreds of years in their original form. As will be shown below, a comparable process is observable in the Roman world. 17

— Chapter Two — Some villas which began with open halls were later subdivided into several comparatively small rooms; others were demolished to be replaced by larger houses of different plan; a few persisted for a long time beside those of more advanced type. These changes were by no means identical with what happened in early modern England but are close enough to provide clues to the nature of the social changes represented by the transformations of house plan. All these general points about the interpretation of plans have to be related to the situation in the Roman provinces, where indigenous populations, after the initial conquest, gradually adopted the architectural trappings of the conquerors; this they did because, as had been the case with Roman imports before the conquest, such forms carried prestigious associations with political power, both the Roman conquerors’ and their own, and also with social status. Luxury spending – conspicuous consumption – was an essential element, as is borne out by sites like Hartfield-Garden Hill or Barnsley Park, where the first element of Roman architecture to be built was, or included, a bath house. On other sites the first consideration was to build a recognisably Roman house, more in the sense that it was different from any indigenous building than that it was truly a piece of classical architecture. The many finds of painted and sometimes moulded plasterwork from the demolished earliest phases of villas suggest that ornament and decoration were as important as the form of the building itself in proclaiming that its owners had made the social transition to the ways of the new masters. Only slowly did the canons of classical architecture gain ground; axiality and symmetry had to be adapted to the reality of social life. How far, though, are those classical canons bound up with public architecture, and to what extent were they modified or ignored in private houses?

CLASSICAL CANONS: SYMMETRY AND AXIALITY In the grandest houses, those of a size approaching that of public buildings, symmetry and axiality were combined in a genuinely classical composition. Nennig, which has one of the most frequently reproduced plans, and Oberweis, which is not as well known as it deserves to be, are among the best examples. Even there, though, not every function could be accommodated in a single unified composition; a bath house at Nennig was placed at what seems like an inordinate distance away, perhaps in order not to detract from the house, perhaps for deeper social reasons; and at Oberweis a long porticus linked the house to an otherwise detached minor block. These examples show that total symmetry, embracing all the parts of a house large enough to lend itself to the application of classical canons on the grand scale, was rarely achieved and may well not often have been sought. Unified British courtyard houses, however big, lacked such monumental classicism; symmetry was much more in the eye of the beholder. Many villas were like some eighteenth-century English country houses in having a perfectly symmetrical front elevation flanked by very different-looking service quarters. From the latter, an unavoidable blemish, the architectural critic averted his eye. This consideration makes symmetry an ambiguous quality, requiring fairly subtle evaluation of its representational significance. To take two examples, both on a 18

— Methods and Assumptions — large scale; Echternach in Luxembourg is as symmetrical as can possibly be, both house and subsidiary buildings, whereas Winkel–Seeb, even larger overall, displays perfect symmetry in the subsidiary buildings but not, surprisingly, in the great house. Much smaller houses present the same kind of contrast. Most have a generally symmetrical front formed by a porticus and its two balancing pavilions, yet the latter are often of slightly unequal size. Similarly with house plans: often what were evidently intended to be balancing rooms or pairs of rooms around a central axis differ slightly in size. Such partial or imperfect symmetry will form a recurring theme of the book. A caution is necessary here. Late nineteenth-century architects of the Beaux Arts school sought perfect symmetry in neo-classical and classical buildings alike, and this led Mylius, for example, in his well-known reconstruction of Blankenheim, to disregard the archaeological evidence which showed the building was asymmetrical. Acceptance of this turn-of-the-century outlook and the restorations it gave rise to has led many archaeologists to complete partially excavated houses with perfectly symmetrical plans, as was done for Rivenhall, Cenero-Murias de Belono and Sarmizegetusa.

MATTERS IGNORED Any attempt to found an account of social structure on a typology of plans runs into a problem of chronology. It has often been said that a typology does not necessarily have chronological implications, yet it can be disconcerting to find a house plan recurring a century or more after its first appearance. In fact it implies no more than the continued usefulness of a particular kind of plan for purposes which changed little over a long period, or for social groups of similar composition. The phenomenon does, nevertheless, require a specific and not merely a general explanation, and one is offered below (chapter 16). Many villas, including some of those excavated with proper attention to stratigraphy in the 1930s and even later, present problems concerning the chronology of their various phases, while for most of those dug before then the dates of their beginning and end are very uncertain. The nature of the problem is revealed by a reassessment of the Welwyn-Lockleys villa, in the course of which the commencement of the Romanised stone building was revised from c. AD 60–70 to c. AD 300 on the basis, principally, of pottery, of which ‘not a single piece is illustrated of the first and second centuries’ apart from an early group and a few samian fragments. This sounds convincing, yet Ward-Perkins thought the new dating ill founded because he had caused many of the early sherds then considered uninformative to be thrown away.13 Just as archaeologists in the 1930s improved greatly on their predecessors’ study of finds, so a later generation made a further advance, and in doing so tended to forget that strong continuities of excavation technique masked their own different and far less selective attitude to pottery and other small finds. Thus, despite more refined dating of pottery forms which permits building phases to be reassessed, uncertainty remains about how reliable the surviving sample is, whether only drawn and published or available for re-examination in a museum. 19

— Chapter Two — For many villas the most that is to be had is a coin list, commonly extending over two or three hundred years, with a considerable number (in Britain) of fourthcentury coins. While a comparative profusion of coins can be taken to indicate activity on the site, the old assumption that the earlier coins indicated when the villa began cannot. Often a thin scatter of late first- and second-century coins has been taken to date the early occupation of a villa but greater awareness of monetary history, following the work of Richard Reece, has made such simple deductions unacceptable without corroboration from pottery.14 No account has been taken of these changed views; it would have been impossible to do so systematically and in any case the theoretical basis of this book sidesteps the need for a precise chronology. Nowhere does the main thrust of the argument stand or fall by dates. Finally, no account is taken of the surroundings of villas apart from the occasional mention of that most characteristic siting on a hill slope overlooking a river or stream. This is because the aim of the book is to establish the significance of types of house and of variations of those types, not to elucidate why a given villa developed in a particular way. To embark on the latter course would have been an impossible task. If the argument about types has any validity, others can apply it to the villas of a locality, bringing in revised chronologies, topography, field systems, pollen and bone analyses and much else besides. Agricultural buildings are hardly mentioned. To understand their architectural simplicity needs specialist knowledge, preferably of the practical kind which Applebaum has brought to the subject. An author who lacks that understanding had better leave it alone.







hen the inhabitants of a native community, of whatever size or social composition, decided to adopt a Romanised way of life in an appropriate architectural setting, what kind of house did they choose to build? Essentially they had a choice between what may be called from their plan forms the hall type – the simpler of the two – and the row type. It was not necessary to begin with the simplest version of the simpler type; a more developed one might correspond better to the needs of a particular community. The implications of this will be dealt with later but the first step is to establish the existence of the two types and their variants, beginning with halls. The simplest kind of house to be built wholly or partly of stone comprised a single large room open from ground to roof and heated by a central hearth. It must have had a general resemblance to the medieval open halls which survive – most of them much altered – in considerable numbers in England, so that when Oelmann wanted a comparison for the villa he excavated at Mayen, the first of its kind to be recognised as such in the course of excavation, he chose Penshurst Place in Kent. England was the only country where so elementary a form of purely domestic building could still be found, and although no hall survived with its open hearth in use, whereas not a few north German aisled farmhouses of the same structural type did, the house with the higher social status – rather too high, as Oelmann implied – was thought more appropriate. ‘Simple’ and ‘elementary’ here refer only to the plan, since buildings of this kind were commonly quite wide and consequently their roof construction and its supporting walls presented greater technical problems than those of smaller structures of more complicated plan; some, especially in Germany, must have had sophisticated forms of roof construction. Commonly such halls had, at one end or both, subsidiary rooms which, in most cases, were under one continuous roof. In the course of time the number and size of the end rooms tended to grow but provided the hall remains the principal room and is big enough to dominate the rest, the term ‘hall house’ remains appropriate. These developed examples were all that Oelmann had to work with when he first established the existence of the hall type, and it is necessary at this point to set out his method more fully. 23

— Chapter Three —

STAHL AND MAYEN In 1921 the prevailing opinion was that many villas had an open courtyard in the middle. In seeking to controvert it Oelmann1 took as his prime example the villa of Stahl (Rhld-Pf.) which had been carefully dug and recorded some forty years earlier (II, Fig. 1). He observed, first, that the facade corresponds to the type of porticusand-pavilions lately identified by Swoboda and that it is not joined organically to the building behind; they form two distinct parts, and whether of one build or two is immaterial to identifying the core elements of the house. Second, the middle space contained a hearth which formed the only source of warmth in the building, and cellar steps which required protection from the weather; and the hearth in particular argues for an enclosed room, not a mere shelter. Third, on the east side five rooms form a bath suite which, like so many others of the kind, has the appearance of having been added to an existing building, as do the rooms on the north side and the internal porch which are hardly separable from the bath block; these are all additions. On these grounds Oelmann concluded that a yard interpretation was ‘as good as impossible’ and that the middle space was a large hall open to the roof, with a smaller room partitioned off it ([I], Fig. 1). The latter, he thought, might have had a low upper storey reached by a staircase rising above the one to the cellar – almost the only sensible suggestion ever to have been made (without prejudice as to its correctness) about access to the many upper storeys alleged to have existed, and not the least notable aspect of this brilliant demonstration. Analysis of eight analogous villas, of which two at least, Serville (Belg.) (Fig. 1) and Saaraltdorf (Moselle) (Fig. 1), were built as simple halls with only a porticus-and-pavilions additionally, confirmed this conclusion. This logical demonstration began to change the interpretation of German and Dutch villas immediately but the argument was clinched by Oelmann’s own excavation of what was undeniably an open hall at Mayen (Rhld-Pf.) (Fig. 1), one which had developed in exactly the ways he had envisaged. A theory with the power of prediction is irresistible; henceforth what had gained currency as the Stahl type became the Mayen type and the high quality of the excavation has caused its conclusions to be widely quoted, to the neglect of the feat of deduction behind it. Subsequent excavations and re-examination of other long-known villas have provided more evidence of halls. Ludwigsburg-Pflugfelden (Bad.-Württ.) (Fig. 1) and Konz-Lummelwies (Rhld-Pf.) are recent discoveries. Bargen im Hegau (Bad.Württ.) (Fig. 1), dug in 1925 and published three years later, has a ‘yard’ in which the only two rooms had straight joints to the outer wall and which is fronted by porticusand-pavilions; it must be a hall. But the closest parallel for Stahl [I] may be KoerichGoeblingen 1 [I] (Fig. 1), which is somewhat more advanced in detail but illustrates almost the same stage of hall development. So common are halls like these that German archaeologists can refer to them as villas of the normal type.


— Hall Houses —

Figure 1 Stahl and related villas


— Chapter Three —

CLASSIFICATION AND ITS PROBLEMS Hall houses can be divided into categories based on several criteria: proportions and dimensions, notably breadth; forms of roof support; whether shelter is provided for animals as well as a family; the position of the entrance; and the forms which development of the basic type took. Two classes of plan stand out by their proportions. The first and by far the more numerous is commonly 9 m or more wide and often has proportions of length to breadth ranging between about 5:3 and square, although they can be 2:1; this is the broad hall type. The other is often no more than 6 m wide and commonly has more elongated proportions, often 2:1, sometimes even 3:1; this is the narrow hall type. It goes without saying that this classification is bound to be somewhat arbitrary. For the purposes of social interpretation aisled houses provide a useful category which is based partly on the form of roof support and partly on the combination of house and byre. Forms of roof support provide two more types, the ridge-post hall and the wide-nave hall. The various types have characteristic possibilities and limitations of development and in these derivatives additional rooms can increase until they finally dominate the hall. With the chosen method of analysis it is not difficult to discover broad halls in old publications, though harder with narrow. To understand the halls themselves is much harder, even from recent reports of high quality, because the vast majority have undergone changes which add to or subtract from the functions for which they were originally designed, and it is difficult to be certain how far the internal dispositions correspond to the original ones. Halls surviving with a little-changed plan tend to be the humbler ones, hence their internal arrangements may not be altogether typical of the type as a whole, and those found under houses of different type have usually lost all internal detail.

LAYOUT AND FUNCTIONS OF THE HALL What was a hall like? What functions, domestic, craft or agricultural, were carried on within it? No single example is sufficiently well preserved to answer these questions; the evidence has to be assembled from several, including those in the poorer settlements which were little altered and which are commonly referred to as cottages or craftsmen’s workshops. Nevertheless, they belong to the same generic type and throw light on their grander cousins. A rudimentary and well-recorded British example, Somerton-Bradley Hill 1 I (Som.) (Fig. 3), adds detail to the impression drawn from Mayen III. It was a small hall house entered in the middle of the south side, and facing the doorway was a cluster of three hearths, only two of which, probably, were in use at any one time. The two biggest were very close, one of them being the equivalent of the central hearth at Serville. The other, almost abutting it, was fired from an elongated pit on the north side and resembles what is sometimes described as a bat-shaped ‘oven’. A third hearth was placed a little way to the left of the entrance, corresponding, perhaps, to what is called in medieval halls the upper 26

— Hall Houses — or socially superior end, so that at some time the hall may have had two focal points rather than one. Otherwise there was little indication of use except for two enigmatic shallow pits, one of them quite large (over 2 m long), at the putative lower end, and there was no drainage or other provision for cattle. Houses like this must have been common but are hard to recover on sites which persisted for centuries. Chastres (Belg.) and Dragonby are structurally superior forms of the Bradley Hill house. For halls at any social level, one of Oelmann’s observations about Mayen is relevant: that throughout the villa’s existence the open hearth was used for cooking as well as heating,2 the nearby drain serving for the disposal of waste water. This raises the question of just what the ‘ovens’ at Bradley Hill I/1 and elsewhere were for; in fact they may have served essentially the same purposes – warmth, cooking – as the open hearth at Mayen. In some of the humblest British halls a social differentiation between the upper and lower ends is apparent, as was certainly true in the much grander Köln-Braunsfeld,3 whereas others with three hearths or ovens, none of which is of special importance, imply a different pattern of living. Generally the British and continental halls reflect similar ways of life.

Figure 2 Halls with evidence of use


— Chapter Three —

Figure 3 British halls


— Hall Houses —

THE HEARTH/ENTRANCE RELATION Broad halls provide evidence for the relation between the two most fundamental elements, the principal or domestic entrance and the principal or only hearth. Most fall into two groups with many minor variations. In the first a central hearth stands more or less opposite an entrance in the middle of one of the long walls, so that the hearth was the first thing seen by anyone coming in; Mayen, Serville and Courcellessur-Nied-Urville are like that. This position for a hearth, where its warmth had most chance of being dissipated by draughts and without any screen to alleviate them, must be determined socially, and the likely explanation is that the hearth was shared without sharp distinction between upper and lower ends. Villas of the Mayen/Serville type, a hall with porticus and pavilions, can be paralleled with minor differences at Neuss [I], Tiefenbach (Fig. 1) and many other places. Most probably had a central hearth. Not infrequently, though, the hearth was slightly towards the rear of the hall and nearer one end than the other – often the left-hand end – so that the person entering was able to orientate himself and go to the upper end, where the hearth was, or the lower end, whichever befitted his station. This disposition of hearth relative to entrance suggests, by analogy with medieval English halls, some distinction between the part nearer the doorway and that beyond the hearth, the place of honour being at the point furthest from the entrance. Persons entering would see, on the other side of the fire, the place reserved for those on whom authority or seniority conferred privilege, and they themselves would be noticed and their deportment observed. A pattern like this can be inferred in a hall as simple as that at Lavans-les-Dole (Jura), where one corner is shielded by the hearth from the rest of the hall to create a privileged area equivalent to the medieval dais.

THE OFF-CENTRE HEARTH In the second group the fire is near the end of the hall and often in the far corner to left of the entrance. Sometimes a well-built stone fireplace is set against a wall near the corner, as at Stahl (Fig. 1) and Bollendorf (Rhld-Pf.) (Fig. 2); sometimes a hearth is actually in the corner, as at Börstingen (Bad.-Württ.) (Fig. 2) or, after a doorway into the adjoining pavilion was blocked, at Neumagen-Dhron-Papiermühle [II] (Fig. 68). Since these were the only hearths discovered, they were presumably just as much the place of privilege as a central hearth, yet they imply somewhat different social relations: the persons around a corner hearth were not well placed to observe those entering, nor did the hearth itself form a visual and social barrier between the entrance and an inner, private room. Evidently it no longer mattered: this category of hearth/ entrance relation implies a different order of importance as between the hall and other rooms. Some difference of function is apparent within the hall, so that at Stahl, for instance, entrance to an inner room was quite divorced from the hearth, and the removal of the latter from a dominant central position may be bound up with the transfer of the representational function which went with privilege to one of the unusually large pavilions. 29

— Chapter Three —

THE LOWER END These tentative interpretations of the socially superior part of the hall leave the lower end to be explained. Where the hearth is central, domestic or agriculturerelated work may have been carried on anywhere within the hall except around the hearth and dais. If the household were large, a division can be envisaged in which the right-hand side on entering was occupied by servants both indoor and outdoor, or by dependants of some kind. Steiner made the point about Bollendorf that ‘The lack of work-rooms permits the belief that the building – at least in its last manifestation – was less a farm than a villa in our modern sense’;4 yet in this particular house the off-centre hearth makes craft-related use of the rest of the hall likely. Various kinds of domestic, non-agricultural work are conceivable: spinning, weaving, making wooden tools and equipment for farm use – any or all of these occupations could have been carried on in the lower end of the hall. An obvious alternative function suggested by the north German aisled farmhouse of the sixteenth to twentieth centuries and its Iron Age counterparts at Ezinge (Neth.) and Feddersen Wierde (NiederSachsen) is the stalling of cattle in winter. Mayen provides a widely accepted interpretation on these lines, although one not applicable to the whole of the lower end. It had in III and IV a doorway in the gable-end which must have provided access from the farmyard to the working end of the hall and to the narrow partitioned-off L-shaped room running round two sides. Whether this oddly shaped space was for cattle, as was suggested by the architect Mylius, is not absolutely certain. In accepting the idea Oelmann was influenced by north German farmhouses, yet the absence of any drain at the rear of the hall, which was built into the hillside, is surprising. Moreover, the space in question, only 2 m wide, is not enough for fully grown beasts, only small cattle or sheep. 5 Nevertheless the gable-end doorway points to a close connection between the lower end and the farmyard and it is hard to think of it as anything other than a byre. Just such a doorway at Crain (Yonne) (Fig. 4) leads to a quite narrow undrained space marked off by a light partition, and here, too, the farmyard access implies a byre.6 At Blieskastell-Altheim (Saarld) (Fig. 4), where no doorways were traceable, interpretation of a lightly partitionedoff room at the east end as a byre is bolstered by the presence of a fodder trough. The absence of drainage is nevertheless remarkable but is explicable if enough

Figure 4 Halls used for stalling animals


— Hall Houses — straw were available to absorb the urine of stalled cattle.7 Despite doubts on this score, it appears that some simple halls, lacking any other domestic room, sheltered cattle as well as the owning family.

BROAD HALLS WITH MORE THAN ONE HEARTH Larger halls frequently have two or three hearths, one of them conspicuously more important than the others. Thus Mamer-Gaschtbierg (Lux.) (Fig. 2) has a well-built hearth at the west end of the hall close to the inner room (later divided), whilst in the corners at the inferior end are two large areas of burning. This is a social differentiation between the hearth around which the privileged congregated and two used by other people for the dual purposes of work – including, probably, blacksmithing and cooking – and warmth. In past ages the practice of living away from one’s workplace did not apply, and those who worked at the hearths are likely to have eaten and slept near them.8 At the villa of Bruchsal-Ober Grombach (Bad.-Württ.) one of two or perhaps three halls (1; Fig. 2) had two certain hearths and a third which is neither located on the published plan nor described. One was in the middle of the hall – in 1911, a yard – in a position comparable to that at Serville. A second, of tile and stone, lay nearer the west end and backed on to what is described as a Halle, that is to say an opensided room with a ceiling or roof carried on posts and interpreted as an entrance-hall within the ‘yard’. Although no details of the post structure were published, the relation of hearth to upper end suggests it resembled the dais of a medieval house, even to having a ceiling or canopy; and something comparable appears to have existed at Konken (Rhld-Pf.) (Fig. 2). So the Ober-Grombach hall had two principal focal points and a third less important one somewhere in the east end or near the north-east corner, an arrangement reflecting a more stratified houseful than those at Mayen and Serville, or Mamer-Gaschtbierg with its functional distinction. Outside Germany broad halls are not very common. France has a scatter represented by Grémecey (Moselle) (Fig. 5), Saint-Pierre-la-Garenne (Eure) (Fig. 5) and Clinchamps (Calvados); also the enigmatic Brain-sur-Allonnes (Main-et-Loire) (Fig. 5) which looks like a villa but had three bronze cauldrons containing votive deposits.9 Hall derivatives found at Izernore (Ain) and Lalonquette (Pyr.-Atl.) (Fig. 50) show that halls may come to light anywhere. Broad halls are rare in Britain. Byfield (Northants.) (Fig. 3) resembles continental halls as closely as any but lacks all internal detail. Its true classification, though, is uncertain following the excavation of a villa similar in size and general appearance at Stratford-upon-Avon-Tiddington (War.), which proved to have an unusual kind of aisled construction (Fig. 35). Swindon-Okus (Wilts.) (19×11.2 m) is very broad for a British hall yet lacks any sign of being aisled. Farmington-Clear Cupboard (Glos.) (Fig. 3) is smaller (21×8.5 m), and has, like Byfield, the characteristic elongated British proportions and preserved informative detail. Here the principal ‘hearth/ oven’ stood nearly opposite to and just a little to the left of the entrance, and beyond it, further into the hall and next to the inner room, was a paved area equivalent to the ‘open-sided room’ of Ober-Grombach A; though not raised like a dais above the 31

— Chapter Three —

Figure 5 French hall houses

general floor level, its treatment demonstrated superiority in a less literal way. Beyond the paving was an inner room. At this end ‘shallow fire pits’ near the hearth were perhaps places where the fire had once been. To the right of the entrance and near the south-east corner was a second hearth, thought to be a working area. Whatever the truth of this – and the building of a bath suite must have distorted the original arrangements to some degree – there was less paving hereabouts, so the hall provided for two social foci of unequal importance. How many people this implies is quite uncertain, but at all events the space available was intended for considerably more than a conjugal or even a stem family. A second hall of comparable size to Farmington may be discernible in the incompletely recovered plan of North Stainley-Castle Dykes [II] (Fig. 3).

NARROW HALLS Stowey Sutton-Chew Park (Avon, formerly Som.) (Fig. 3) is a good example of the narrow-hall type. The hall area is 114 sq. m, about the same as Stahl and bigger than Serville (101 sq. m); Mayen is not much bigger (135 sq. m), Bollendorf and Grémecey (both 220 sq. m) are twice as big. As with Stahl, the idea that the central space ‘was probably not roofed’10 can be dismissed. It sheltered three ovens, one at each end and one near the middle of the rear wall; and there was also a ‘rough fireplace’ or hearth, probably secondary and in an unusual place, tucked away in the near lefthand corner on entering.11 This seems usually to be the upper end of a hall, and this was where the best-built oven – ‘really more in the nature of a furnace’ – stood. Since no animal bones were found in the vicinity ‘it is not quite certain . . . whether [the oven] was in fact for cooking’,12 yet there was no evidence of industrial activity either. That point applies to all three ‘ovens’ none of which showed traces of any


— Hall Houses — activity other than generating heat; could that have been their principal purpose? The lack of a fireplace in the middle of the hall and of a dais are significant in the light of Farmington and other villas. Presumably three groups of people or households used the hearths in a communal hall – it is quite big enough. The dominance of one household, if it existed to any considerable degree, must have been expressed in terms of access to the inner room or pavilions. Another British hall interpreted by its excavators as a yard is Wraxall (Som.) (Fig. 3), which appears to have had a flagged stone floor throughout. Its hall had the proportions of Stowey Sutton and about the same overall area as Mayen IV. A few barely Romanised halls were even more elongated, one of the most rudimentary – not a villa – being building 2.5 in the Somerton-Catsgore settlement (Fig. 3). Its proportions were nearly 4.5:1 and although described as a barn it had two built hearths/ovens and two other hearths. A building so provided cannot have been solely for storage;13 it must be a house in which ancillary farming activities were carried on. It is not surprising that the proportions of the remote and barely Romanised house at Laugharne-Cwmbrwyn (Fig. 3) in west Wales (Pembs., now Dyfed), the one most distant from an urban centre, should resemble those of the hardly more Roman hall Somerton-Catsgore 2.5. A much smaller hall, the ‘Dwelling House’ at the well-known villa of Langton (Yorks.), began in I (Fig. 3) as a simple hall, to which, as at Cwmbrwyn, a small room was added at one end. A few narrow halls are known outside Britain. The most notable is Schimmert-op den Billich, commonly called Ravensbosch (Neth.), which is part of a villa or perhaps a small town (vicus).14 It has proportions of 4:1 and porticuses front and rear; that at the front has pavilions, one being recessed into the hall; that at the rear is wide, as if for some work purpose. A broad hall like Mayen IV and a narrow one like Stowey Sutton, despite their comparatively small difference of size, carry quite different social connotations. The single large central hearth so characteristic of the broad hall implies a unity within the houseful not demonstrated by the three ‘ovens’ of the narrow one or the two hearths of the elongated broad hall at Farmington. If, as these examples show, elongated broad halls have much in common with narrow halls, their proportions ensure different possibilities of development. Neither class shows much sign of stalling cattle; Blieskastell-Altheim and Mayen III are exceptions and the building at Böckweiler (Saarld) (Fig. 4) called the Wirtschaftshof or work-yard may be another. Most have no definite evidence of accommodation for cattle.

SINGLE-ENDED HALLS Improvement of the basic hall could take several forms. A room of equal breadth or two small rooms might be added; or a porticus-with-pavilions; or a small room at the rear; or a cellar; or a combination of these. The simplest development was to add a room at one end of the hall. Parallels from medieval England are called single-ended halls, an illogical but convenient expression which will be adopted here. The new room may serve some completely new requirement or accommodate a function previously performed in the body of 33

— Chapter Three — the hall and, since evidence on which to form an opinion is usually lacking, the question remains open. Maidenhead-Cox Green I (Berks.) (Fig. 69) was a narrow hall with an inner room entered, it is reasonable to suppose, from the hall. The corresponding room at Farmington had a hearth and well-made floor and provided space enough for a family. Some indication of the social relations between the inhabitants of the villa is provided by the position of the doorway between hall and inner room. Sometimes this can be inferred from the limitations imposed by other features, as at Farmington, but in most cases, as at Stahl or Doische-Vodelée (Belg.) (Fig. 39) it is guesswork. At UplymeHolcombe IIA (Fig. 66), a house of the same plan-type as Maidenhead, the threshold between the two rooms was recovered; it was in the middle of the partition and implies a different relationship between the rooms from that at Farmington. Instead of the inner room being reached by a doorway behind the hearth, access to it was less removed from the body of the hall and more open to general view. Both houses are assumed to have had the inner room at the upper end. This may not always have been the case; just as single-ended medieval halls could have the extra room at the lower end so, perhaps, might provincial Roman halls such as Stowey Sutton, and as Somerton-Catsgore 2.1. certainly did (Fig. 3). Another form of improvement was to build a cellar in the hall. Sometimes it was in a corner, as at Dreieich-Götzenhain (Hessen), Stein (Neth.) and VoerendaalUbachsberg (Neth.), sometimes clear of one, as at Overasselt (Neth.) (Fig. 6). Whether, in such cases, a room stood above the cellar is impossible to know from the published evidence.

Figure 6 Single- and double-ended halls


— Hall Houses —

THE INNER ROOM SUBDIVIDED An alternative way of providing special-purpose space was to have a pair of rooms at one end of the hall, as at Rothselberg (Rhld-Pf.) (Fig. 6), Frankfurt-Bergen-Enkheim [I] (Fig. 40) and Michelstadt-Steinbach (both Hessen). In all three one room was larger than the other, with a cellar beneath the smaller. The location of the cellar entrance at Rothselberg does not preclude the possibility of a doorway from hall to smaller room but may indicate that this room was an inner, private room entered through the larger one and overseeing access to the cellar. At Bergen-Enkheim the high status of this part of the house is underlined by the hypocaust in the larger room.15 The best British parallel, a much smaller narrow hall at Quinton (Northants.) (Fig. 6), had a hearth nearer the west end of the hall where there are two rooms, one of which was enlarged at the expense of the other. A doorway opened from the hall into the larger room. In the hall a raised and better-surfaced floor was marked off by a simple kerb on the side nearest the doorway – a dais. Friendship-Taylor, who excavated the site, considered a light partition might have stood on the line of the kerb, rather like that around the dais at Ober-Grombach A. Whether the end rooms formed a suite, the smaller reached through the larger, or were entered separately, is unknown, but the whole conveys the same impression as the upper end of a medieval domestic hall. The larger hall house at Overasselt (Fig. 6), with proportions of about 3:1, has, besides the cellar already mentioned, two rooms at the south end of not quite equal size, one of which had a hypocaust. The presence of important domestic rooms at both ends of this building supports the negative evidence of Quinton, that cattle were not stalled within the hall. In broad halls the paired rooms are sometimes separated by a space that looks like a through passage leading out of doors; but it often became so through alteration and was not necessarily so originally. In its simplest form this disposition appears in Mehring I (Fig. 69) and at the east end of Grémecey, and raises the question of why the rooms should be separated. Implicitly they were not connected functionally, as might be expected with two service rooms or two rooms of a suite; possibly the ‘passage’ was a lobby with two doorways shielded from the hall.

DOUBLE-ENDED HALLS A double-ended hall (another term from vernacular architecture) has rooms at both ends, like the German halls of Brücken (Rhld-Pf.) (Fig. 6) and Heppenheim (Hessen) (Fig. 6), which have two small rooms at one end and an undivided fullwidth room at the other. Great Staughton (Cambs.) (Fig. 40) has only one room at each end. In houses of 10 m span and upwards such a room might be very narrow in proportion to its length. Larger groupings of rooms occur, differing very much in detail and complicated by the presence of the pavilions at each end of the porticus. In trying to understand how the hall itself functioned it has been possible, so far, to ignore pavilions, something the more easily done when the pavilions stand largely or wholly beyond the ends of 35

— Chapter Three — the hall, as at Serville; but in analysing the more highly developed houses, where the pavilions are integrated architecturally with other rooms at the ends of the hall, the possibility that they intercommunicated with them cannot be ignored. Two Dutch villas, Nuth-Vaasrade and Buchten (Fig. 23), illustrate the point, which will be taken up later (chapter 8).

AISLED HOUSES Broad halls may be divided by roof-supporting posts set either axially as ridge-posts and forming two aisles, or in two rows forming, in the customary English terms, nave and aisles.16 In Britain one large group in the latter category has a distinctive bipartite character consistent with its accommodating, in its early forms, a family and livestock under one roof; originally called the basilican house,17 it has been given the name aisled house to distinguish it from aisled buildings devoted almost entirely to agricultural or industrial purposes, and from the aisleless house-and-byre.18 Contrary to expectation, the enormous amount of villa excavation and aerial photography during the past thirty years has produced only a handful of possible examples in mainland Europe.19 In its simplest form the aisled house is a long timber building with two internal rows of earthfast posts and no structural partitions from end to end of the middle span. Functional divisions were marked by the position of the hearth or by no more than different kinds of floor – hard-rammed and swept clay in the domestic part, a rougher cattle-trodden surface in the rest. Many such have been excavated in the coastal regions of Holland and north-west Germany and within the Empire it was the dominant type at Rijswijk (Fig. 65). In Britain, Exning-Landwade I (Suffolk) (Fig. 7), Denton I (Lincs.) and Lechlade-Claydon Pike (Oxon.) may have been like this, for although positive evidence of cattle is hard to find, an agricultural use for one end is established with some probability by a doorway in the gable-end wall wide enough for animals or carts. In the aisles, though, especially at the family end where alteration is always more destructive, light partitions may have been entirely lost; for whereas in the nave the division between living and work parts seems to have been customary rather than structural, elsewhere some separation of family space from cattle is likely. With aisled houses as with monospan halls, original form and function have to be inferred from such evidence as remains after extensive alteration. It seems that half the length of the nave or a little less was living-space. Denton had a large hearth, the only one in the building, about half-way down its eight-bay length. To one side of it, after partial rebuilding in stone (Fig. 7), were domestic rooms, and it can reasonably be assumed they perpetuated earlier usage. How much of the nave beyond the fire, if any, was living-space? Since the hearth is the only cooking place, the area around it must have been domestic, and similarly, when the day’s work was done, people will have congregated around it. A much bigger aisled house, Winterton D I (Lincs.) (Fig. 45), had a hearth in the nave at a comparable point, about where the aisle rooms ended; the large open space beyond did not have a wide doorway at the end and revealed no trace of cattle, only a hearth, five trench-furnaces and some paving slabs. A smaller building B I was similar; both acquired rooms at the end of the nave. 36

— Hall Houses — Other aisled houses hint at a similar division. West Blatchington (Sussex) (Fig. 7) looks as if it was entered, by analogy with Denton III (Fig 7) and halls such as Serville, through a doorway opposite a hearth, and so into the domestic part which was cut off from the rest in [II].20 Carisbrooke (I.o.W.), Norton Disney (Lincs.) (Fig. 75), Petersfield-Stroud (Glos.) (Fig. 7) and West Dean (Wilts.) are similarly proportioned. Exning II, East Grimstead (Wilts.), Mansfield Woodhouse (Notts.) (all Fig. 7) and Weyhill-Clanville (Hants.) all acquired domestic rooms at both ends and in much of the aisles, leaving a large irregular-shaped space in the middle: essentially the same process of development as occurred in many a German broad hall. Clearly many aisled houses, especially the large ones built with stone footings, never sheltered cattle, whereas Denton and Stroud, which retained their wide end entrances throughout, may possibly have done so, although some doubt exists.21 Only those which were minor elements of a villa, such as the one outside the courtyard at Sudeley-Spoonley Wood (Fig. 70), are likely to have been part house, part byre. Moreover, Winterton D and the aisled house at Darenth (Kent) (Fig. 42) were both preceded by aisleless buildings lacking any positive evidence of stalling cattle, so that the apparently obvious link between aisled houses in Britain and those at Rijswijk and the Friesian coast must now be questioned. Phosphate analysis is needed on a suitable site. An aisled house at Odiham-Lodge Farm (Hants.), which seems from its plan to have developed in much the same way as Winterton D, has been interpreted in a more detailed way than any other villa through the artefacts found in the rooms.22 Certain rooms are assigned to men, others to women, a small one to both. A wing at what is conventionally regarded as the upper end of a house of this type comprises two ox-byres and a stable and perhaps a second small stable. This west part of the house did not communicate directly with the rest. A raised granary occupying one aisle adjoined that part of the nave used for threshing. How much weight should be attached to these attributions? Some aspects are very odd: the animal stalls did not communicate with the house, yet those who looked after valuable livestock surely slept close to it so as to respond quickly if a beast fell sick; there is no wide doorway for waggons to deliver corn on the stalk for threshing and take away grain; and the removal of one of the two major functions of this house type to the opposite end, confusing upper and lower ends, is hard to parallel in any house type – indeed, hardly conceivable. Odiham needs reassessment from the excavation archive and from an architectural standpoint. All the examples quoted so far are British. A building likely to be of this type at Villeneuve-sur-Cher/les Augerets (Cher) is exceptionally long, the aisles have stone footings throughout, and the ends are less differentiated than is usual, but it has a much closer resemblance to a British aisled house than most in France.

RIDGE-POST HALLS A second category classified by structure comprises halls having, on the long axis, a row of posts rising to the apex to support the ridge-piece, to which the rafters are affixed. 37

— Chapter Three —


— Hall Houses —

Figure 7 Aisled buildings

Two buildings of this type were excavated at Bedburg-Garsdorf (Nordrh.Westf.) (Fig. 68). The smaller, the principal house, with proportions of 5:2, combined in II stone footings for the hall and end room with earthfast ridgeposts. On one side of the yard was a larger building of the same type, a workhall with four hearths set slightly off the long axis, probably because it was easier to contrive some kind of smoke vent to one side of the ridgepiece than across it. Nor surprisingly, major villa buildings of this primitive type are rare and it is to be expected that derivatives in which some of the inconvenient middle posts have been replaced by monospan trusses will be found; Saaraltdorf (Fig. 1) and Heidelsheim (Bad.-Württ.), each with a single axial post-hole, may be two such. Rudston 3/II, of the same structural type, occupied the same position relative to the main house and no doubt served the same purpose; Fishtoft may be a minor British example (Fig. 7). A less fully reported example at Verneuilen-Halatte II (Oise) is likely, from its position in a large courtyard, to have been a workhall, for all that the main house of that phase was not discovered. Less certain is the function of a well-built structure at Bocholtz-Vlengendaal (Neth.) (Fig. 46), the middle one of three sizeable buildings set out in line. Neither of the suggested uses, barn or storehouse (possibly of two storeys), is compelling. Halls of this kind are related to the wide-nave type. 39

— Chapter Three —

WIDE-NAVE HALLS This category of hall is more instructive than the fewness of examples might suggest. The only one excavated to modern standards, Hölstein (Switz.) (Fig. 7) has a nave to aisles proportion of approximately 3:1 instead of the 2:1 of the aisled house. Structurally the type is linked to the previous one by the use of one or two ridge-posts or some equivalent support; most pairs of posts carried a king-post construction on tie beams. Hölstein began as a hall open from end to end and with a porticus, perhaps entered at the end, along most of the north side. In II, two rooms at the upper (east) end, occupying two bays of the colonnades, were partitioned off, one being provided with a hypocaust, and a small square room was created at that end of the porticus; these rooms may have replaced a similarly used but structurally undivided space in I. A bath suite was added at the opposite end in [III]. No evidence of the stalling of cattle was found at the lower end of the nave, only agricultural and cooking equipment and horse-trappings. Hölstein evidently developed and was used in much the same way as Winterton D. Kaisersteinbruch (Aus.) (Fig. 67) has comparable nave–aisle proportions. Lacking ridge-posts, it must have had king-post trusses throughout except at the north end, where lengthwise stability for the roof appears to have been provided by an axial wall. Rooms were partitioned off, probably in [II], to make a unit or apartment, and near it the only entrance discovered may, by analogy, have been opposite the principal hearth. It has frequently been claimed that these two buildings belong to the class of ‘basilican’ or aisled houses, whereas their proportions speak against any direct connection. Others like them have a thin but wide distribution: they include NeerharenRekem (Belg.), Wollersheim-Am Hostert (Rhld-Pf.), a building at Liestal-Munzach (Switz.) (Fig. 76) and possibly the strange building at Ash (Kent), to judge solely by its proportions; and from aerial photographs, Saint-Aubin le Mazaret E. Perhaps the alleged peristyle villa at Bennwil (Switz.) belongs to this class.

NAVE AND AISLES OF EQUAL WIDTH Differences in nave–aisle proportions are primarily structural matters, yet they also govern the conditions of use. Winkel-Seeb (Switz.) raises questions of this kind more acutely than any other. The principal building (A, Herrenhaus; Fig. 7) incorporates a hall 31 m long but only 9 m wide and open to the roof which was divided up by two rows of thirteen columns into three spans of about 3 m. What were they intended to support? An upper storey carrying heavy loads is the only reason that comes to mind for so many columns so close together, yet few loads except bulk grain needed that much support. Furthermore, in VI–VIII, when only the middle third of the hall retained columns, an upper storey is scarcely possible. In VI a ‘roundish-concave’ wall about 8 m long and occupying the whole of the south ‘aisle’ was built, and in front of it was ‘exceptionally heavily burnt earth’: an installation described as a ‘chimney-like hearth with a roundish 40

— Hall Houses — heat-radiating wall’.23 It is hard to envisage a chimney or smoke hood of the requisite size rising above this through an upper storey; an open hearth is more likely. On the other hand, columns so closely spaced are unnecessary for any kind of roof. A second building at Winkel-Seeb presents the same problem. In the inner courtyard, B IIB (Fig. 7) has two rows of poorly aligned columns forming three more or less equal aisles in an overall breadth of 8 m. It is described simply as a hall and has open hearths, yet is reconstructed with an upper storey because, as in A, there are so many columns;24 but no explanation is offered for their number or spacing. In fact, both A and B present such contradictions that no solution can be offered. It is an odd situation because building E (Fig. 72) has more usual proportions like those of Hölstein. A few other buildings with three equal aisles are known. A large one occupies the north side of the domestic peristyle at Liédena (Spain) and a much smaller British example is East Dean-Holbury (Hants.). In France there are two at SaintAubin le Mazaret and two subsidiary agricultural (?) buildings at Maulévrier and Châtillon-sur-Seiche.25 Although the Maulévrier hall is so much less important than the Seeb buildings, it is over twice as wide (19 m) and can perhaps be compared to a granary at Rome-Via Gabina with equal aisles and nave (also 19 m). The SaintAubin buildings are wider still – one about 28 m – so that they may have had three parallel roofs. And in that unsatisfactory state the problem of Seeb A and its analogues must be left.

HALL OR YARD? THE CASE OF INZIGKOFEN For half a century Oelmann’s interpretation of Stahl and its analogues held undisputed sway until, in the 1970s and 1980s, a number of archaeologists working in central and southern Germany have asserted, without challenging Oelmann’s conclusion formally, that this or that villa had an open yard more or less surrounded by rooms, while ignoring the problems thus created. Discussion will be divided between probable halls which are dealt with here and more problematic examples which are reserved for chapter 7. The argument, insofar as any is advanced, is that the width of the central space is too great for a monospan roof, therefore it must have been a yard, and the occasional presence of a hearth or a staircase to a cellar is taken to imply partial roofing. And despite the demonstration that the plans of Stahl and Mayen make good sense as a number of rooms subordinate to and opening off a hall, nobody has attempted to show the relation between the rows of rooms on two or more sides of a yard; how they functioned as a house, that is. Nor, since the idea that the porticus-with-pavilions is a characteristic of provincial Roman culture is not specifically rejected, is it explained why so distinctive a facade should have been thought appropriate in front of an open yard. And whereas a hall makes the central space intelligible, the purpose of a comparatively small, enclosed and usually undrained yard requires explanation not so far offered. Inzigkofen (Bad.-Württ.) (Fig. 27) is a key site in this debate. Reim, arguing that the space in the middle (16.60 m overall; 13.50 m clear) is too wide for a roof 41

— Chapter Three — without posts, and that the foundations of the inner porticus wall are not only too slight to support the weight of tiles but are slightly narrower than the outer wall, concluded that the inner space can only be a yard. This justification of a courtyard, in a footnote listing hall interpretations, makes a veiled challenge to orthodoxy;26 it stands or falls by the roof arguments, which will now be assessed. First, the foundations. They were of limestone fragments, only partly held together by mortar.27 It is hard to know just what ‘partly’ signifies, but provided the rear wall of the porticus, of which nothing remained, was well built, its 0.90 m thickness need not preclude a roof of wide span. Given a roof which did not exert a strong outward thrust and was not subject to heavy lateral wind pressure – and a Roman roof of low pitch with king-posts in tension would meet both conditions – the support afforded by a masonry wall would suffice. Moreover, the wall could have been of mixed construction, incorporating posts at the truss ends. As for the span, the row of bases on the west side are larger than needed for a lean-to;28 freestanding posts supporting roof trusses would reduce the span to 13.5 m. Oelmann anticipated the objection about wide roof spans and his demonstration that Roman technology was capable of roofing impressively wide spans need not be recapitulated. It is sufficient to add a few examples from villas. Blankenheim I (Fig. 70) had a hall 12 m wide, Bruckneudorf (Aus.) [I] (Fig. 69) one of 14.5 m, while that at Laufenburg (Bad.-Württ.) (Fig. 38) was yet wider at 17 m.29 Clearly wide spans were perfectly feasible in villas of high status. How far such technical feats can be envisaged lower down the scale is an open question, but spans of 12–13 m were not exceptional and can be assumed in all the villas in Rheinland-Pfalz in respect of which doubts have been raised, few of which exceed 12 m in the clear. If 13.50 m is thought impossible, that has to be argued. That leaves unanswered the point that cellar steps need a roof.30 Unfortunately, nothing is said about the surface of the interior space; whether yard or hall, surely some part of it was paved? If a yard, were there drains? Nor is it explained what activities took place within it. A kitchen garden is a possibility, an ornamental garden more appropriate between the domestic ranges, yet no paths were found. We are not told what was found in this space; the evidence is not given; the question seems not to have been asked. Inzigkofen prompts another question: how did the two rows of rooms flanking the yard function as a house, and why are they set out so differently from those in other villas with a Romanised frontispiece? At only some 4 m wide they are narrow compared with British row-houses, which themselves tend to be narrower than their European counterparts. They have no room of particular importance. More important, no veranda or gallery links them with the porticus and one another, yet intercommunication, producing a series of passage-rooms, is inconceivable. As rooms opening off a large open hall they are credible; regarded as two separate blocks of rooms (houses? apartments?) fronting a yard, the way they functioned needs explanation. In the light of these remarks the uncertainty expressed about the 9.5 m wide interior space of Hirschberg-GrossSachsen I (Bad.-Württ.)31 is unnecessary, and indeed for any span less than 13 m. It may even make intelligible such oddities as LörrachBrombach (Fig. 28). 42

— Hall Houses —

HALL AND PORTICUS UNIFIED: SINSHEIM AND KINGSWESTON A rare plan-form occurs at Sinsheim-Wald (Bad.-Württ.) (Fig. 8). It has a four-bay colonnade familiar in southern Germany and another, quite differently spaced, forming the rear wall of the porticus. All the column bases both front and rear are 0.60 m square and have in the top surface a mortise intended to receive the tenon at the foot of a timber post of large cross-section. Now the rear of the porticus – the front wall of the hall – comprises several closespaced posts interrupted by a length of wall and was thus well capable of supporting a very considerable roof weight. So was the rear colonnade, which is of four bays varying in length between 3 and 5 m. Given posts of dimensions appropriate to the bases, a roof is perfectly feasible;32 and such large bases and posts would be unnecessary for lean-to roofs. Sinsheim is without parallel in Germany, although Sigmaringen-Steinäcker (Bad.Württ.) (Fig. 8) resembled it if any of the four bays of the colonnade or arcade facing the porticus were open. A British villa, Bristol-Kingsweston, provides a more certain parallel with its open arcade separating hall from porticus. The underlying intention was to obtain, in effect, a wider hall than the customary British proportions allowed by adding to it the partly separated-off porticus; it follows that the porticus was enclosed, not arcaded, at the front.

DIVIDED HALLS A few villas in which the hall is divided structurally into two parts tend to bear out the idea that such houses could be occupied by more than one household. At Voerendaal-Ten Hove I (Neth.) (Fig. 9), a transverse wall divides the hall into two equal parts, each about 13.5×8 m. A greater number of additional rooms on one side is balanced by larger ones on the other, resulting in approximate equality; and an oblong room is shared unequally. Heerlen-Boventse Caumer (Neth.) [I] (Fig. 9) may have been a bipartite hall with porticus-and-pavilions. At Vesqueville (Belg.) the hall

Figure 8 Halls open to porticus


— Chapter Three — appears to have proclaimed division very simply by opposite pilasters forming an intangible boundary, one respected by all that could be crossed as occasion warranted; and that idea is supported by the presence of a hearth in one corner. Vesqueville represents a half-way stage between an open and a part-structurally divided hall.33 The simplest bipartite division occurs at Laperrière-sur-Saône (Côte-d’Or) (Fig. 9) and Bad Homburg (Hessen) (Fig. 9), and at the latter is matched by an oblique wall dividing the yard.34 Such walls can occur indoors too, as at Saint-Aubin-sur-Mer (Fig. 9), where the hall is divided by a wall running obliquely across it into two not quite equal parts. Division is not always equal. In several houses a square hall adjoins a smaller room of the same depth which is bigger than any ordinary end or inner room; both are under one roof. One such is Nünschweiler (Rhld-Pf.), where the hall (13×10 m) is larger than the second room, which is approximately square, but the latter is so big that it must be a second hall. Both were subdivided, the principal hall structurally by a cellar with – perhaps – a room above, the smaller hall functionally by an L-shaped paving of rubble stone.35 One of the analogues for Stahl, Neckarrems (Bad.-Württ.), has two only partially separated hall-like rooms; a wide opening (4.5 m) in the wall between them permitted intercommunication, so that here, as at Vesqueville, the hypothetical two households were closely linked in their daily lives. A final example

Figure 9 Divided halls


— Hall Houses — is Houthem-Vogelsang (Neth.) (Fig. 9): phase [I] is a large bipartite block fronted by porticus and pavilions; the two rooms in the south hall may belong to [II].

HALLS: A SUMMARY Halls with a remarkable variety of structure and plan are found over a wide area of Europe and in more regions than those mentioned. The categories used above may make the diversity intelligible by using the fragmentary evidence of individual houses and a multitude of examples to establish a coherent type having many variants. Viewed as a whole, they have much in common. All are good-sized buildings and some are very large; all are capable of accommodating more than a nuclear family. To suppose such a family living at one end of, for example, Farmington, and working at the other would be an inconceivably wasteful use of so large a space. If an owning family and its servants, both indoor and outdoor, be thought of as occupying the upper and lower ends respectively, there is little sign of the spatial differentiation that might be expected in a situation where there is great disparity of wealth; for Farmington, and still more some of the bigger broad halls, must have been quite costly to build. In a very few instances there is definite evidence of dual occupation, the most striking being Sinsheim, which has two entrances into the porticus. That villa, like so many, has lesser rooms at both ends of the hall, and these are sometimes the product of a single building campaign, sometimes of gradual growth; the tendency is towards duality, even if it is only manifested in the two pavilions at the ends of a porticus. Then there is the question of the extent to which cattle were stalled in a hall. Even aisled houses in their Romanised forms are doubtful as combined shelters for family and animals. As for other halls, the claims made for many are dubious and can often be dismissed out of hand. The indoor crafts necessary to keep agrarian production going are a much more likely use of the space in most instances. More will be said of this in discussing how hall houses developed (chapter 6) but before that the second fundamental type, the row-house, has to be considered.






cattered throughout the northern provinces are many houses which lack a dominant room in the sense that the open hall at Mayen dominates the whole house throughout its history. They take the form of a row of two to five rooms – larger houses will be treated separately as far as possible – which are interspersed with smaller squarish or corridor-like rooms, and will be designated the row type. The underlying concept was first used by Drack in classifying Swiss houses, and was devised independently, though not given that name, to meet the somewhat different houses in Britain.1 At the outset other problems of terminology have to be tackled. The first arises from the presence in the row of rooms, especially where there are five or more, of one somewhat larger than the rest: what to call it? Room use in villas is mostly unknown and the designations given to rooms by Romans are uncertain in their provincial application. In medieval and early modern architectural history nomenclature is just as much a problem, the same name being used by contemporaries for an enormous single-room building open to the roof such as Westminster Hall, for the 5 m-square principal room, lofted over, of a sixteenth-century two-room peasant house, and for the splendid point of entry of an eighteenth-century country house; the continuity of the old name reflects the gradualness of change in room use and the consequent lack of urgency at any given time to devise a new and appropriate one.2 Not only are the uses to which such a principal room was put uncertain, they can often be inferred to change even where the room stays the same size. In this situation ‘hall’ must often serve for want of a better term but consistency is unattainable; where a more definite use can be inferred, living-room or representational room will be employed. A second problem emerges from considering the implications of row-houses for social structure. The recognition that they may provide accommodation for more than one household has led to the borrowing from vernacular architecture studies of the term ‘unit system’; hence ‘unit’ refers to one component element of two or more in a row-type house, each of which corresponds to a household.3 The neutral term is useful in making comparisons, although alternatives will be used, but the real problem is to decide the number of rooms in a unit; different interpretations group 46

— Row-Type Houses — rooms differently.4 A further complication stems from the marked variations of room size – essentially, depth – which can be found in this kind of house.

INTERCONNECTING ROOMS: NEWPORT (I.o.W.) AND LAMARGELLE The first task is to establish the forms that these domestic units might take. Newport (I.o.W.) (Fig. 10) is unusual in retaining evidence of doorway positions and is one of the few row-houses where intercommunication between rooms can be known almost completely and their uses, to a lesser extent, inferred. If porticus-and-pavilions are set aside, it comprises a range of rooms that can be matched at many other sites.5 In the middle the largest room is entered by a wide opening (2.4 m), just inside which, buried in the floor, were two boulders. The larger had a flattened top and the smaller one a hole for a metal pin; both probably had something to do with a shrine.6 At the back of the room, which did not intercommunicate with those on either side, was a hearth. It was hardly a living-room but rather a place, open to view from the porticus, where the household gods were venerated and ceremonial feasts held: a representational room, a place where those symbolic acts were performed, no doubt often reciprocally, which were designed to bind the whole household together. On either side of this, the principal room by size but not ornament, is a suite of two rooms

Figure 10 Row houses


— Chapter Four — entered from a lobby, each room in the suite to the left being smaller than its counterpart to the right and probably better finished. From the first the house had a porticus and it is possible that the bath block and right-hand pavilion are both part of the original design, whereas an identical plan at Gargrave-Kirk Sink A (Fig. 12) lacks the baths. Yet another simple version of the Newport plan appears at Lamargelle-Versingue 2 (Côte-d’Or) (Fig. 10). A square middle room, having only a doorway to the porticus – not found, though it certainly existed – is flanked by simpler suites, each comprising a passage leading to a doorway near its blind end, from which an adjoining square room was reached. These passages,7 members of a numerous class, had no other function and will therefore be called transverse lobbies. Newport has rooms which perform a similar function: one is only half the depth of the range, hence small lobby; its elongated counterpart will be called a longitudinal lobby; and obviously these two may vary in their proportions. A lobby of whatever kind ensures privacy in the room it serves by shielding the entrance from the gaze of the curious, whether strangers or servants whose business lay elsewhere. It thus appears that at both Newport and Lamargelle the flanking suites were completely separate; one, perhaps both, had access to a pavilion, and in both villas the underlying concept of a middle room between and independent of two units of habitation, and implicitly in common use, is clear. At Lamargelle the flanking rooms are the same size, at Newport not, so the two suites could be equal or slightly differentiated. Which was the more important, the larger (to the east) or the smaller, is not quite certain at Newport because total destruction of floors in the larger one precludes comparison with the mosaics of the smaller; yet the complete absence of tesserae in the east part of the house argues that there never were any mosaic floors there and that the smaller rooms were the better finished. This contrast between size and finish recurs in other villas, Chedworth, for example. Such sets of rooms may be described by the seventeenth-century English term ‘apartments’, that is to say they provided for two persons or, far more likely here, two households (and their personal servants),8 leading separate lives in many respects but sharing some common activities. Among the latter were some carried on in the large middle room, which, if the shrine is conjectured correctly, was the focus of domestic religious observances and, no doubt, of the feasts accompanying them. This room, entered by a wide opening which was not necessarily closed by doors, and open to the view not only of the households but also, being the first part seen by them, of strangers, contrasts strongly with the seclusion of the apartments. Here all made offering to the household gods before receiving hospitality or transacting business. An alternative to the occupation of each apartment by a household is by an individual – husband in one, wife in the other, each with at least one personal servant. No evidence of such a division of the sexes has been advanced in a RomanoBritish context, so henceforth it will be assumed that an apartment corresponds to a household. In that case Lamargelle and Newport show the number of rooms appropriate to a nuclear family of a certain social position at the time they were built: fewer in the French villa than the English, only one large room and access (presumably) to the nearer pavilion. Since pavilions are, typologically, additions to 48

— Row-Type Houses — the basic plan, it appears that the minimal requirement of a household may have been only a single sizeable room, with access to other facilities.

NEWPORT ENLARGED: SPARSHOLT This form of interpretation can be applied to other row-type villas, most of which have not revealed doorways. Sparsholt (Hants.) (Fig. 68), though, has several. The five middle cells of the principal house [I] are identical to Newport save that they are not a mirror-image pair flanking the middle room (as Newport nearly is); the righthand lobby and large room are interchanged. This illustrates an important principle governing the planning of row-houses, that the rooms of the individual units comprising them can be arranged in any order. But Sparsholt is larger than Newport, having at each end a narrow room of the same width as the adjoining small pavilion, itself no wider than the porticus. These rooms may, indeed, be additions to the original core and are shown as such in the hypothetical development. Fortunately, doorways were preserved in this part. Entrance to the new room at the south end was from the porticus through what may have been in formal architectural terms a pavilion and this doubtless applied at the north end too. A new unit was thus added at each end, doubling the number of households. A feature of Sparsholt requiring explanation, and it is one found in most rowhouses, is the variation of room size. The principal rooms of the two original apartments differ slightly, the one to the south being bigger than its fellow to the north, yet its lobby and adjoining room are smaller: an apparent inconsistency in making the relative importance of the two units clear and one not found at Newport, where all three rooms of the better finished apartment to the left are smaller than their counterparts to the right. In fact, differences of size between units are common, indeed normal, in row-houses, and are observable in the biggest of the kind. This asymmetrical disposition of the two apartments at Sparsholt no doubt had a social basis, a desire to remove the entrance to the inferior, workaday apartment away from the front entrance, and this in turn had an architectural consequence; the three doorways facing the porticus could not be symmetrically placed, as they often were, and as reconstructions of villas almost invariably show them.

LAMARGELLE ENLARGED: DOWNTON A seven-cell plan of comparable size to Sparsholt and with equally instructive detail occurs at Downton (Wilts.) (Fig. 10). A middle room with mosaic pavement is flanked by two equal-sized blocks, each with what looks like a wide transverse lobby between two square rooms. The greater than usual width probably denotes some common use, not simply circulation; hence it can be assumed that the two rooms it serves were used by the same household. This two-unit villa was built with a porticus. By comparison with Newport or Sparsholt the most striking element is a small architectural feature blocking direct access from porticus to middle room; it stands at the back of an internal porch, with 49

— Chapter Four — doorways on either side, through which the two apartments had to be reached, and was probably a water shrine of some kind, the modest counterpart of the wellknown and more elaborate structure at Darenth (Kent) (below, p. 156). If the small middle cell of three is viewed as a transverse lobby, each apartment is a more spacious version of the major apartments in the previous villas, yet the better-preserved part of the porticus to the south reveals a complication. At the end is a pavilion-like room serving the very humble and unpavilion-like function of furnace room, so that it may never have had an internal doorway. Between it and the porch is a small room that appears to have functioned as an anteroom to the lobby proper, as if to shield the porticus from the disagreeable sight of a menial task. That hardly affects the general sense of a plan comprising two apartments, each having access to a middle room of common use which was furnished with a mosaic pavement and opened on to a little nymphaeum. Classification is further simplified by recognising that the small lobby, though a refinement of the transverse lobby, performs fundamentally the same function and is interchangeable with it. These considerations applied to the core of the plan make intelligible other houses with a big middle room and symmetrical or near-symmetrical flanking suites. The best known is Ditchley (Oxon.) (Fig. 10). It must have been an imposing example of its kind, having two good-sized wing rooms in addition to the main range. Little Milton (Oxon.), Hemel Hempstead-Boxmoor II (Herts.) and MaidstoneLoose Road (Kent) have exactly the same core rooms as Ditchley although the small and transverse lobbies change places; whilst, at Walton on the Hill-The Heath, a small lobby leading to one of the two apartments stands at the end of the building instead of flanking the middle room. In houses without a large middle room such as Faversham I (Kent) (Fig. 11) and Gargrave-Kirk Sink D (Fig. 12) transverse and square lobbies stand back to back. These variations reinforce the point made earlier, that the rooms of the individual units comprising a house can be arranged in any order. The principal house at Mansfield Woodhouse (Notts.) (Fig. 11) differs only in that the supposed transverse lobby is wide enough to be virtually a room in its own right; Cobham Park (Kent), destroyed at one end, is certainly of this class; PangbourneMaidenhatch (Berks.) and Deanshanger (Northants.) are variants of Lamargelle. In short, five-cell plans of several variant forms are common in Britain. No conclusive evidence has yet come to light to confirm the idea that houses of this kind were for two households but one tantalising hint that the idea may be correct appeared at North Cerney-The Ditches (Glos.) (Fig. 13). There partial excavation of the first stone building revealed what looks like half of a five-cell plan – a simple straight range – and it has, leading up to the one transverse lobby that was uncovered, a stone-lined path. It may be that a second lobby and path are waiting to be found, analogous to the two parallel approaches found in some much larger villas such as Romegoux-La Vergnée (Charente-Mar.) (Fig. 13) and Winkel-Seeb (Fig. 72). If this suggestion is correct, it confirms the impression given by lobbies that unitsystem villas were the norm in Britain.


— Row-Type Houses —

Figure 11 The varied positions of lobbies

THE SIZE AND FORMS OF UNITS Row-houses could consist of several virtually identical units: Hérouville-Lébisey (Calvados) (Fig. 14), Boos-Le Bois Flahaut [I] (Seine-Mar.) (Fig. 14) and Rockbourne W Bldg III (Hants.) (Fig. 42) all have three rooms, each with a transverse lobby. Bramdean [I] (Hants.), where one of three lobbies is small and square, shows the beginnings of differentiation by form; Newton St Loe W [I] carries this further in having three lobbies, all different. Analysis is easy where each room has an obvious lobby, but in many villas at least one living-room stands next to what is either a wide 51

— Chapter Four —

Figure 12 Unit-system villas


— Row-Type Houses —

Figure 13 Row houses with two entrances

Figure 14 Virtually identical units


— Chapter Four — lobby or a narrow room, and the difficulty is the greater where no large middle room divides the range into apartments. It is hard, in all but the simplest houses, to decide how many rooms constitute a unit – changes over time are likely – and how many units (or households) comprised the houseful. Romegoux (Fig. 13) is an interesting case. It has a large courtyard screened by a porticus, in front of which are two square porches and, between them in the middle, a little open-fronted building that can only be a shrine. Across the courtyard two paths ran to two more porches within a porticus; and it is observable that, whereas the outer porches are placed perfectly symmetrically, the inner righthand porch is slightly nearer the east side than its counterpart. In a villa as rectilinear as this an apparently trifling difference cannot be dismissed as carelessness in setting out; in fact it conforms to the different importance of the rooms – entrance-hall and living-room combined, probably – into which the porches lead, the west one being twice the size of the east. That may imply four units each comprising one room and a transverse lobby – a larger Hérouville-Lébisey: four units of living-space grouped in pairs of not quite equal importance beside a common living-room/entrance.

HOUSES ENTERED AT ONE END Romegoux and Downton show that no simple correlation exists between villa size and unit size, and Farningham-Manor House I bears that out (Fig. 15). As built, it was a series of units on an L-shaped plan, entered at the south end of the wing and so along the porticus; only in II did it acquire an impressively wide entrance in the middle. In the way it was approached Farningham resembles Blankenheim I (Fig. 70), and it has, at the end of the wing, a wide doorway opening into a shallow room which may have an equivalent in the shallow unexplained structure in the German villa. It is inconceivable that this was the entrance9 because, unless it were itself a passage-room with other such behind it, it can only have led to one room. How many units there were is open to argument, as in most other villas of this size. Two parts of the house can be interpreted with reasonable confidence. The wing is a hall-like room 13 with a lobby 12 and was probably a workhall like the

Figure 15 End entrances


— Row-Type Houses — comparable room in Blankenheim I. At the other end of the porticus, furthest from the entrance, is the largest and most private apartment, comprising a transverse lobby and two rooms 1, 3,10 like the Downton units or Ewhurst-Rapsley 6 V (Fig. 41). Probably the rest comprised three units. Two units which were, functionally, a mirrorimage pair with square lobbies, had intercommunicating living rooms 6 and 7, the doorway between them being well established, and a third, in the corner, had a transverse lobby. If the suggestion of intercommunication between units be thought to invalidate the whole mode of interpretation, early modern houses show this need not be so. Where units form two detached houses they could be joined corner to corner to permit intercommunication, and where they form a single range a doorway may be found linking different units, exactly as at Farningham.11 Whatever the truth of this, an important aspect of Farningham is that the accommodation it provides might elsewhere have been divided between several buildings, as at St Lythans-Whitton (Fig. 65) or Gargrave-Kirk Sink. Upchurch-Boxted (Kent) has points of resemblance to Farningham-Manor House but the report has too little detail to allow the comparison to be pursued. Another member of the same class may be Alresford (Essex); in [I] it appears to have been entered at the end of the left-hand wing and so along the porticus, and in [II] acquired a second wing, like Farningham. Reach (Cambs.) and Ridgewell (Essex) may have developed in a similar way. Once variation in the number of cells in units within the same house is recognised, the way is open to interpret other villas, particularly those entered along the porticus rather than at right angles to it. Aylesford-Eccles I (Kent) (Fig. 66) becomes more intelligible if the unusual open-ended room at the south end, like that at Farningham I or its simpler version at Blankenheim I, is seen as a shrine, and to give some idea of what it may have looked like, the shrine (aedicula) from the fort at Mainz (Fig. 37)12 has been modified with a king-post roof (Fig. 66); and the ornament would be simpler, achieved with plasterwork and paint. There, the person entering paid his respects before proceeding towards the two transverse-lobby units and the larger one with superior floors beyond. But did all five rooms with tessellated floors belong to that unit? The transverse lobby and adjoining rooms certainly, by analogy: the narrow north end room presumably; the biggest room in the house, between the superior unit and the rest, perhaps not. That may have been common to all the units because a number of villas seem to have one room additional to those which adjoin lobbies, as at Box (Fig. 48), Compton Dando-Littleton (Som.) and St MichaelGorhambury. Even if the five end rooms prove to have been wrongly divided up, the unit system offers the only reasonable interpretation of so long a range. Continental examples are less easy to find. Generally, end-entrance plans have more complex forms of lobby and will be discussed later (chapter 5). One particularly relevant house which deserves mention here is Ormalingen (Switz.) (Fig. 15), which was recognised from the first as having been entered at the south end of the porticus. It differs from the British houses in having, at about a third of its length from the entrance, a wall in which stood a doorway of large dimensions; tufa voussoirs of its round-arch head were found.13 Its purpose is clear, to form a social barrier between the room at the south end – a workhall? – and the three apartments, although its 55

— Chapter Four — placing hints that the south apartment may have been of slightly lower status than the other two.

THE LONGITUDINAL LOBBY A lobby of one form or another is an essential component of all but the simplest row-houses. Besides the two already mentioned, a third kind lies parallel to the front wall of a house and provides access from the entrance – usually the porticus – to a principal room at each end, while directly in front of the entrance is a third inferior room – or sometimes two – that may have been reached either from the lobby or from the major rooms. Entry into the house in this way is so unimpressive by comparison with entry into a hall or even into any good-sized room that it must have connoted some important aspect of social structure – taken here to be duality of power or ownership. Houses of this type demonstrate as clearly as any the joint occupation implicit in a particular architectural concept, despite the small size and restricted development of many of the examples. In its simplest form the type occurs at St Lythans-Whitton VIII W (Glam.) (Fig. 65), in a house comprising an elongated lobby serving two rooms, that to the left being larger than the one to the right. How the room behind the lobby was used, whether in common or in connection with only one of the large rooms, is unknowable. In the light of Huntsham N (Herefs.) (Fig. 16) – the misnamed ‘cottage-house’ – where the corresponding space is divided into two unequal rooms, the former may be the more likely. Huntsham, though, has its own problem, for it was not divided simply into major and minor units; instead, the larger major room (1) adjoins the smaller minor room (3). As with other small distinctions of this kind, the social nuances underlying them must have been of considerable importance to be worth bothering about. This is precisely the kind of counterpoise of rooms found in quite different circumstances, whether in the conjunction of the larger pavilion with the smaller end chamber of a hall or, in a row-type house, that of the larger lobby with the smaller main room. Although difficult to understand, this very common balancing of component parts of a house must correspond to some aspect of provincial Roman social structure that required concrete expression; and it provides a contrast to those houses where the larger major and minor rooms go together, such as Gargrave-Kirk Sink C [I] and (excluding the wings) Eaton-by-Tarporley (Cheshire) (Fig. 16). It is

Figure 16 Longitudinal lobbies


— Row-Type Houses — precisely the smallness, variation and repetition of these architectural differences that shows how important they were. Some subtler model of society is needed to explain them than the current ones of a quasi-manorial lord or capitalist farmer or even a farmer operating with little more than his family’s labour. Where the room behind the lobby was structurally undivided, it may have been a common workhall, as the Dwelling House II at Langton (Yorks) (Fig. 3) implies. It began as a superior version of St Lythans-Whitton W, the big room to the left having a hypocaust and its fellow to the right another occupying only half of the room. In the narrow room behind the lobby were a small plunge bath – presumably used in common – and a furnace serving the right-hand room. The use of the rear space seems to have varied between a workaday service room and two high-quality rooms, but probably it was always shared between the apartments. The most ambitious application of the longitudinal lobby in a British villa is at Wellow (Som.) (Fig. 16), where, despite the lack of doorway positions, the implied pattern of circulation is clear. An unusually large middle room with a mosaic floor is flanked by transverse lobbies which may have connected the porticuses; more importantly, each led to a longitudinal lobby at the rear, from which opened two square rooms. The purpose of this was to ensure the upmost privacy for the tworoom apartments. A further point of interest is that, in a large and well-appointed house, the accommodation deemed necessary for the personal use of those persons occupying the most important suites was no more than two rooms – exactly the same as in Downton and many others. It exemplifies the truth of Laurence Stone’s observation that a certain minimum number of rooms were required for daily living by members of a given social class or order, and this deserves to be kept in mind when alternatives to the unit system are put forward. In a few instances the longitudinal lobby is wider than appears necessary for its primary function. Cenero-Murias de Belono (Oviedo) N (Fig. 12) is one such. It is as though the lobby were used as work- or living-space for the family occupying this unit. Similar conjunctions of three rooms are widespread, occurring in Belgium and Germany, for example.

LONGITUDINAL LOBBIES AND REVERSE SYMMETRY Architectural duality is a constant theme in the houses of the Roman provinces and takes a great variety of forms. One of them, very obvious once remarked, is reverse symmetry, that is to say the provision of two like rooms or sets of accommodation facing in opposite directions, a peculiarity which has been noted occasionally in individual buildings14 but has never been examined as a recurrent phenomenon. Of the several forms it takes, only those including a longitudinal lobby will be considered here and of these the most easily explicable is Beadlam N I (Yorks.) (Fig. 17). It had three doorways at the front, an imposing one 2.3 m wide in the middle and one near each end, sited as near symmetrically as the plan allows. Their placing and associated features demonstrate how closely architecture mirrors social relations. In I, access is principally by the two side doorways. The left-hand one opens into a longitudinal lobby 5 and so to the west unit, comprising rooms 7 and 8 and also a 57

— Chapter Four —

Figure 17 Reverse symmetry

little workhall (6) in which was the furnace for 7 and two sets of iron tools. At the east end, where the porticus terminated,15 the doorway opened into a general-purpose living-room heated by an ‘oven’, and just inside is a doorway into what became in II the best-appointed room in the whole house, embellished with a tessellated floor; thus the external doorway and the best-room doorway stand in exactly the same relation as does the doorway from the lobby and that into the pavilion (8). The two apartments shared a common middle room like the one at Newport and, like it, were entered by a wide doorway. A shrine might have been expected about here in I, something a little more elaborate than the pot (for a votive offering) just beside the west doorway. The plan is in effect a rearrangement of a row-house with a middle room flanked by two two-room units, into neither of which does an additional room (6) fit easily: a situation resembling that in several other villas except that in them the extra room either always was, or developed into, something far superior to a workhall. Beadlam N suggests that the middle room (4) – a modest representational room – and the workhall (6) were held in common by the two households. The two rooms were easy of access for both, and the reason for cutting off the workhall by a doorway may be because it was in the inferior part of the house rather than for any utilitarian considerations of the smoke and smells arising from its use. Beadlam is the most intelligible example of a mode of planning found in much of the Empire and over several centuries, the reverse symmetry of the side-by-side halls at Francolise-San Rocco [I] (Italy) (Fig. 69) being the earliest example. However close the formal likeness of the two sets of rooms, they were rarely identical, a condition paralleling the small differences between the units of row-houses. High Wycombe (Fig. 17) is another British house organised on this principle. As at Beadlam, the longitudinal lobby at the front is cut off by a wall and doorway from the middle room and the one at the rear is open to it, and since no menial or dirty task was performed here the difference must have a social explanation. More difficult to understand 58

— Row-Type Houses — is the function of such elongated and at the same time so extraordinarily narrow (6×2.5 m) rooms. The proportions can be matched in continental row-houses but there the greater width of the main range, usually 8–10 m, makes it earlier to envisage a use.16 Bedrooms were the excavator’s suggestion, but whether rooms were specialised to that extent is open to doubt, and if they were, why were they the closest to the public space of the entrancehall? Whatever they were for, this house is one of the few which do not have the principal rooms in the main range. But of its bipartite character there can be no doubt.

HOUSES WITH TWO FRONT ENTRANCES The idea of joint occupancy aroused instant hostility when first mooted. Two defences were erected against such unwelcome thoughts.17 Thus Ferdière dismissed the hypothesis of collective property as the result of (among other things) ‘erroneous interpretations of plans of the living-quarters of villas’, and ‘adventurous hypotheses on the symmetry of villa plans’; yet he, like his compatriots, has never explained the plan of the little villa at Maulévrier (Fig. 13), first published in 1836 and republished in 1934 by the man he acknowledges as ‘the great specialist on Roman Gaul’, Albert Grenier,18 or even that of Romegoux, well excavated in the late 1930s. Almost everyone takes it for granted that a country house or a farmhouse has a conspicuous main entrance, with a minor one at the back for servants or farm-hands and, probably, a doorway into the garden: only archaeologists and architectural historians can take two front doorways for granted. Yet two doorways can only be intended for two sets of persons – who must surely correspond to two households, not family and servants – to enter separately; and Maulévrier can be interpreted in these terms without straining the evidence. It is divided internally by a cross-wall midway between the two front entrances into two parts of equal size but with a different disposition of rooms. The wider entrance leads to a unit like the equivalent part of Sparsholt, the narrower to a Downton-type unit, and the different kinds of lobby show that each carried a social meaning. Romegoux (Fig. 13) tells a similar story on a larger scale; it is important for the informative detail lacking at Maulévrier. Two porches in front of the outer porticus match two within the inner porticus; parallel paths like those at Winkel-Seeb connected them. The house comprised, on the west side, what appears to have been a common hall, the largest room in the building, which was flanked by two units, each with a transverse lobby and living-room. A third such unit lay to east, followed by a somewhat enigmatic room in which was a sunken structure of squared masonry; this contained a considerable quantity of cinders and put the excavators in mind of a vasarium or wash-house.19 Last came a unit identical to the three preceding ones but superior to them in having, in part of the lobby, a hypocauston20 giving indirect heat to the larger room. In all this, only the alleged wash-house causes surprise, its lowly function sitting uneasily with the porch in front of it. A more likely analogy for a room so approached is Downton, with its nymphaeum in front of the common middle room. Viewed in this light Romegoux has two rooms of common use – a representational room and a larger hall (workhall?) – each standing between two domestic units. It is a grander cousin of Ormalingen (Switz.) or Aylesford-Eccles (Kent). 59

— Chapter Four — Hemel Hempstead-Gadebridge Park (Herts.) (Fig. 13) is more easily understood in a general way than in detail. On the north side, from which the villa must have been approached in III and IV and probably throughout, are two small projecting porches21 proclaiming two points of entry and two apartments behind them. The west porch stands opposite what looks like the central lobby of a two-room unit, the east porch does not relate to anything in particular, and all that can be said is that the five rooms broken by two lobbies constitute a row-house to which two entrances would be appropriate.

FOUR-CELL HOUSES Houses comprising four cells are common and typologically they precede the fivecell category with a middle representational room. One group has back-to-back lobbies, usually one transverse and one small, for example Faversham I.22 The different kinds of lobby reveal that some social distinction, some difference in status and authority, existed in the kin-group. How big a change this represents from the earlier situation, where villas like Hérouville-Lébisey [I] or Boos [I] had three units differentiated only by size, is hard to say, but probably the effect on those who entered such houses was as great as must have been the contrast in the storeyed houses of post-medieval England between a cross- (or through-) passage and a lobby entrance.23 Faversham I (Fig. 11) was entered from the south by a timber veranda not covering the full length of the building. Verandas coupled with differently placed lobbies occur in other four-cell row-houses such as Welwyn-Lockleys I (Herts.), Brixworth I (Northants.) (both Fig. 11) and Chilgrove 2/II (Sussex) and there too they do not extend the whole length.24 The reason must be that it was not needed; a utilitarian shelter went only as far as the final point of entrance, that being the extent to which protection from dripping eaves was needed. In that case the north unit of Faversham I had a doorway either into the transverse lobby (for otherwise the lobby had no conceivable purpose) or, as the logic of the veranda dictates, into the hall-like livingroom (1), where it was opposite a hearth, as in many hall houses. Brixworth I (Fig. 11) lacks the transverse lobby but is otherwise like Faversham I; the point where the veranda ends suggests that the doorway into the unit further from the house entrance was in room 2. The different placing of the square lobby in Welwyn-Lockleys I makes it impossible, strictly speaking, to be sure that it comprised two units. Only its general likeness to the others suggests that is so, in which case the veranda terminated exactly as at Brixworth, at the entrance to the bigger room (4) of the far unit. If that were so, the doorway faced a hearth, and there may have been another doorway into the end room (5) placed comparably to that at Beadlam N I into room 2. The general type, irrespective of the position of lobbies, is common. At Gargrave D [I], Cléry-sur-Somme (Fig. 11) and possibly Hartlip W [I] (Kent) the lobbies are back to back, at Ellesborough-Terrick (Bucks.) (Fig. 11) they are at the ends, and Combe St NicholasWadeford W (Som.) is probably the same but with two transverse lobbies. As far as their primary purpose is concerned, the two kinds of lobby are interchangeable.25 60

— Row-Type Houses — Four-cell houses with a single lobby are thought to comprise two units because those with two lobbies do, but it is conceivable that the two rooms extra to the obvious unit were a common hall and a one-room apartment. One corollary of the row-house style of life, that entry into a living-room should be through another and usually smaller one, makes it unlikely that a unit of whatever size would be reached through a common hall. Thus consideration argues in favour of a second two-room unit at Welwyn, and in other adequately excavated and fully published examples the same conclusion is probably justified, but one French villa with a comparable plan certainly incorporated a common hall. Civray-Le Poirier Molet (Cher) (Fig. 18) comprises three square cells and a transverse lobby at the end, and, since the middle cell of the three is open to the porticus, that part must have been public space common to all. It was roofed over, no doubt, so that apart from its having a front opening of full width instead of a wide doorway, this room is the counterpart of the middle room at Newport. One further example, one of the most frequently quoted British villas, deserves discussion – St Stephens-Park Street (Fig. 11). In VI (the first Roman phase) it was a four-cell building, including what was either a narrow room or a wide transverse

Figure 18 Compact row houses


— Chapter Four — lobby (2.4 m) over a cellar.26 It had a porticus on the west side which is likely to have been continuous around the south end; and if, as is generally assumed, the house was approached in VI from the east, as it certainly was later, a porticus existed there too. That assumption is compatible with the longitudinal lobby (2) on that side serving the big room (1), and from it opened a second lobby leading to a small room with a hearth (3). The whole is the equivalent of one of the apartments at Newport, which leaves the putative narrow room over the cellar looking like a lobby to the second large room.

THE COMPACT ROW TYPE In central and south-west France many houses, most of them known from aerial photographs, have a more compact plan than those so far described. From five to seven rooms are arranged within a rectangle having proportions of about 5:2 or 5:3 and at first sight these houses, which are mostly of symmetrical plan, can have absolutely no relation to the row type. On closer examination the rooms vary in size in much the same way as do row-houses, so that comparison between the two categories is enlightening. This kind of plan is obviously unrelated to the hall type, and because its elements correspond to those of the row type and could just as well have been strung out in a line, it will here be called the compact row type. The relation between the orthodox type in its several variations and the compact version can be demonstrated by some examples, all in the department of Cher. It may be appropriate to begin with an unquestionably orthodox villa. La GuercheLe Grand Chausseroi W comprises a porticus and three cells, the middle one with a wide opening to the middle room, Newport-style. The right-hand cell is like one of those at Maulévrier, the left-hand one an undivided elongated room; each is assumed to be a unit or apartment. A kind of plan not diverging much from this appears in the villas of Primelles-Champ Chiron (Fig. 18) and Morthommiers-Le Crot; both have a porticus, a middle space completely open to it, and a room on either side. But the middle space, which must have been roofed, is subdivided, and here is the only difference between the two houses. Morthommiers has, partitioned off at the back, a narrow room looking like a lobby giving access to one or other flanking room. Primelles has in the corresponding position a chamber wide enough to be either an anteroom serving the same purpose or an inner room to the hall/porticus; similarly transformed, it, too, would not look amiss. A still simpler combination of the basic three rooms is found at Lazenay-Les Sales. There the porticus as an architectural entity is suppressed, to be replaced, for the two rooms beside the hall, by front lobbies like the one on the right-hand side of Maulévrier; the person entering presumably came into the hall by a doorway in the middle, to be received there or to proceed through a lobby to one of the apartments. This is not so different from Lamargelle except that entry at Lazenay is into a public space which is also living space. And with Vallenay-Patureau Fourneau the concept of transformation within the row type is complete: it is like a poorer version of Boos or Hérouville-Lébisey with each unit turned on its side so that the lobby is at the front, thus eliminating the need for an orthodox porticus. In such a case the middle 62

— Row-Type Houses — lobby, being at the front, may rather have been a short porticus giving access to the adjoining lobbies. Again a comparison with orthodox row-type houses comes to mind between this arrangement and the quasi-pavilions at Sparsholt through which the end rooms were presumably entered; the difference lies in the proportions of porticus to lobbies. Sparsholt has a bearing on another house, one known through aerial photography, at Fromentières (Mayenne). The core comprises three square cells separated by two transverse lobbies; but the middle cell, which has a transverse lobby-like room at the rear, is open to the porticus, which terminates, like the British villa, in two small square rooms. Only one British house, at Brislington (Avon), probably, but not definitely, had a middle room completely open to the porticus. Both these villas are minor variants of the row type. Two general observations can be made about the compact row type. The first concerns the porticus which, where it formed an undivided space with the common middle room, must have been part of the common living-space, and to that extent resembles the porticus of other villas where the porticus was in all probability enclosed, such as Great Staughton (Cambs.) (Fig. 40). That leaves unresolved the purpose of the narrow room at the rear of the hall and, since it appears in four houses of the small sample available, it must be important. Its lobby-like appearance may be misleading; the presence at Vallenay of such a room, in addition to the three at the front, argues for some other function than an aid to circulation. One possibility, suggested by the position of this small room, is that it accommodated a shrine. Uncertainties of this kind, though, may be held to matter less than the overall resemblance of the variant forms to the row type proper; if so, a general likeness permits a linkage with more complicated versions of these houses which can be found in other parts of France. At what point do divergences from the concept of a row-house demand the creation of a new type? The question arises with a villa in Aquitaine, Lussas-etNontronneau (Fig. 18), where the principal house, though composed entirely of features found in row-type villas, looks very different from any other so far discussed. It is entered from a colonnaded porticus through a quite large porch – large enough to provide a waiting space for servants – and so into the rear living-quarters. Beside the porch to the left is the exact counterpart of the middle room at Newport, with its front walls reduced to pilasters to give an opening the full width of the room – like Brislington, perhaps. It no doubt served as a representational room and perhaps contained a shrine. The smaller room on the other side of the porch, which also has a wide doorway, may have been a small workhall/kitchen. In the rear part of the house the most notable feature is the unequal size and reverse symmetry of the two apartments to the left and right; both preserve a fireplace backing on to the partition between room and lobby and situated in a characteristic relation to the doorway. Between them is what is described as a court, although it must have been roofed. These dispositions afforded maximum privacy for the apartments, divorcing them from the public spaces of the house to a degree impossible in all but the largest rowhouses such as Bierbach (chapter 5 and Fig. 19). The reverse symmetry of Lussas-et-Nontronneau finds not a parallel but an echo in the mirror-image pair of rooms of the only German example of the compact row 63

— Chapter Four — type. Hummetroth-Haselburg (Hessen) (Fig. 18) has, in its two principal rooms, the thickened walls found in Jemelle-Neufchâteau (Fig. 46) or Bierbach; they mark off the most important inner part of the room and are the equivalent of a dais in a hall or the apsidal termination of some grand rooms in other villas. The pair of rooms on each side of the middle hall-like room and the little apsidal room could be strung out in line like an orthodox row-house. Looking at Lussas-et-Nontronneau and Hummetroth as row-type derivatives may be thought to strain classification to its limits. Nevertheless, it is useful to regard them thus on the grounds that the transformation of component parts, of which these villas are such striking examples, is merely done in a more radical way than in most forms of the type, while the whole retains the underlying concept. That is important for social history.





any houses comprise a string of rooms of various sizes that do not relate one to another in any clearly discernible way. They have two features in common, one being the absence of anything that could be described as a dominant hall, the other that a number of usually squarish rooms are separated by cells divided into anything from two to five rooms. That is enough to put them into the row type, and, as before, only the main range, the essential core of the building, will be considered. At the outset, though, it has to be admitted that many of the larger villas can be interpreted in more than one way in detail; the forms which lobbies take vary greatly and only the additive principle underlying the planning of this kind of house is insisted on. Two comparatively uncomplicated examples will show a line of development from the row-houses described earlier.

KIRCHBERG: THREE-ROOM AND LOBBY UNITS Küttigen-Kirchberg [I] and [II] (Switz.) (Fig. 19) began as a house of seven cells, the ends of the first build being marked, as Drack observed,1 by thicker walls. Being sited on the edge of a fairly steep slope to a river, it was necessarily entered by a rear porticus, very little of which remained. A middle room (5) has on each side two formally identical units, each a more advanced form of those at Downton (Wilts.) and virtually identical with the domestic end of Schupfart II (Switz.) (Fig. 71) or Ferpicloz [I] (Switz.).2 The result can only be described as symmetrical, yet the two smaller rooms (3 and 4) in the west unit are very slightly narrower than the corresponding ones (6 and 7) in the east, and this, in a building set out with characteristic Roman accuracy, is certainly intended to express some difference. So slight a difference of size must be a matter of status rather than function; it can hardly have been perceptible even to those living there, although it may have been emphasised by equally small differences in architectural ornament or decoration. It is not easy to appreciate the significance of such small distinctions between one part of a house and another from a plan alone, yet they must have been thought very important to be so carefully made. Again, a comparison with seventeenth-century English houses may help. A minor manor house in Hertfordshire, Aston Bury, has two staircases, to all intents and 65

— Chapter Five —

Figure 19 Lobbies and room groupings


— Developed Forms of Row-House — purposes identical, serving a not very large straight range of rooms. The details are not exactly the same but only someone very familiar with both could distinguish one from the other in photographs. Yet the staircase projection at the upper end is very slightly wider than its fellow at the lower end and projects slightly more – something that is only appreciated at the rear of the house, and more apparent to the modern eye after looking at a plan, so used are we to accepting near-symmetry as true symmetry. So, too, in a farmhouse at Pirton, Walnut Tree Farm, which has two formally identical porches: yet the one at the upper end is just a little wider than the other at the lower end.3 To people lacking the printed material through which most knowledge is acquired today, these small differences no doubt leapt to the eye and conveyed an instant social meaning, as they certainly did in Roman Britain. To return to Kirchberg, like many continental row-houses it has, not the squarish rooms that were the rule in Britain, but oblong ones: 1 and 10 are half as long again as they are wide. If they were entered at one end, the first third of the length must have been, to some extent, a passage-room; if near the middle, the room was effectively divided into two parts. In either case the superior end was probably at the display front of the house (south). Further differentiation is provided by the front porticus, embellished with pilasters to give the house an imposing appearance when seen from the valley below. It is wider at the west end where the very slightly inferior unit is; and that the narrowing is intended to proclaim the superiority of that end is confirmed by the nature of the additions in [II]: service rooms, including a kitchen, to west, a suite of rooms with hypocausts and mosaics to east. Kirchberg, larger and more complicated than Downton or Newport (I.o.W.) shows how architectural features apparently without significance are given meaning when interpreted in a social sense. A further point is that the two smaller rooms reached off each of the transverse lobbies, though formally similar to the square lobby and adjoining room of Newport, in fact perform a different function. These considerations are observable in other villas like Kirchberg. Another Swiss villa, Vicques [I], though more pretentious, with the porticus extending forwards to pavilions totally unconnected with the main domestic activities, is otherwise almost identical in the body of the house except for the different sizes of the smaller rooms in the two units; and here the existence of a rear porticus, which Kirchberg [I] must have had, is well established. A larger version appears at Landen-Betzveld [I] (Belg.). It has the same plan as Kirchberg [I] with the addition at each end of a smaller unit comprising a large room entered from a transverse lobby.4

THE ELABORATION OF UNITS: LAUFEN-MÜSCHAG As provincial Roman society developed, more varied and complex kinds of household required the elaboration of the lobby-and-three-rooms type. It is easier to point to the new forms than to explain them, as Laufen-Müschag (Switz.) (Fig. 19) shows. There phase [I], excluding the porticus-and-pavilions, resembles Kirchberg [I] in a general way, with two units, each with a transverse lobby, flanking a middle room (6),5 yet on closer examination much is different. 67

— Chapter Five — The north unit has a room (2) twice as long as wide, so it must have been used somewhat differently from the corresponding room in Kirchberg and still more so from that at Landen. The lobby (3) is awkwardly shaped because the other rooms off it are of unequal width. It must have been important to make room 5 wider and to enter it at the far end, not at what could have been the easiest point architecturally, where it adjoins 4; and from these assumptions it follows that the important part of 5 is likely to have been at the east end.6 Doorways into the lobbies and middle room would give the symmetry normally sought.7 An oblong shape for the middle room (6), assuming it had a common representational function, is appropriate. At the far end, no doubt, was a shrine or seat of honour to be used on formal occasions, including ceremonial dining, and approached with the degree of respect the occasion required. The south unit comprised, on one side of the lobby (9), two rooms (7 and 8), much like those at Kirchberg, and on the other the largest room of all (10/11), half as big again as its counterpart (2) and looking like a domestic hall. The large pavilion (15) which adjoins it has three hearths of various periods and is probably the workhall.8 This analysis shows that the two units of Vicques had acquired different functions and that the house had lost the near equality of the two parts of Kirchberg. Clearly we are dealing with a different kind of houseful. Rooms 3–5 are important indicators of a new style of life not so far observed in row-houses and present a conjunction of shapes reminiscent of the block of rooms in the middle of Köln-Mungersdorf (Nordrh.Westf.) (Fig. 71). Laufen appears to represent rather different styles of life in the two households, one of them markedly superior. It is still a row-house but has lost the ethos of earlier ones. One other Swiss villa deserves mention in this connection. Zofingen has two cells subdivided in the same way as rooms 3–5 at Laufen, but they are extraordinarily elongated and in inverse symmetry. I do not accept Drack’s view that this arrangement and that at Laufen are the result of alteration.9

BIERBACH: HOW MANY UNITS IN A ROW-TYPE VILLA? Row-houses develop in size as well as in the variety of their lobbies and some are so large that they must consist of more than two units; thus, even if the interrelation of lobbies and rooms is hard to explain in detail, their general significance may be clear enough. At Bierbach (Saarld) (Fig. 19) the essentials of the latest phase are simple: excluding the middle room, five large rooms are served by four lobbies.10 The front elevation of this stately house was dominated by a wide portal leading into the porticus and on to the wide doorway of a very grand hall. At the ends of the porticus two lesser entrances appear originally to have led directly to lobbies: to the left a transverse one (between 18 and 20), to the right a half-depth one (between 2 and 5) for although no doorway from porticus to 3 was found, the major room 2 is unlikely to have been reached only through two lobbies and another large room. The different forms of lobby, paralleled in so many row-type houses, evidently connoted a social difference in the eyes of anyone entering the house; the smaller lobby led to the most important private room and the most important household. 68

— Developed Forms of Row-House — To demonstrate the importance of rooms 2 and 20 each has, at the end furthest from the doorway, a thickening of the walls, whose purpose was probably to support a vault or canopy of timber over the socially important end of the room, on which the narrowing would focus attention; and standing in opposite relation to the porticus, these rooms provide an instance of reverse symmetry. At Bierbach this arrangement appears to be secondary if, as the report argued on the basis of masonry joints, the west end room (20) originally intercommunicated with the adjoining pavilion;11 but that leaves the wall thickenings incomprehensible. A large and imposing villa with two social focal points so carefully balanced in importance cannot be the residence of a single proprietor and his immediate family: it must correspond to some larger kin-group. There are two more lobbies. As Kolling, reporting on excavations of thirty years earlier, noted, the block of four rooms 6–9 to the west of the narrow corridor connecting the porticus to a rear doorway stand in close relation to one another; in fact they and 5 look remarkably like St Stephen-Park Street VI (Fig. 11). In this little block Bierbach reveals another manifestation of reverse symmetry: one of the two large rooms, 5, is entered at the south end, the other, 9, from the north, a point which may have a bearing on the British house. At the other end of Bierbach the relation of the remaining rooms to one another is less clear. The large room 16, a workhall and kitchen, was entered from the porticus through a square lobby (14). Opening off it were two rooms (15 and 17) which were probably both service rooms and places where those responsible for food and cooking slept. The function and point of entry of the remaining room (13) are unknown. What emerges from this analysis is that a villa as large and architecturally impressive as Bierbach was far from the conventional interpretation offered for it as the house of a great proprietor – a wealthy man who built, near the country town of Schwarzenäcker, a summer seat appropriate to his style of life.12 The wealth is all right, and quite likely the town connection too, but it was as much the seat of a kingroup as of the person who was its head and representative for the time being; certainly not just a summer residence. It can be interpreted as a houseful comprising two major households centred on 2 and 20; two lesser ones in 5 and 9; and a sizeable fifth one in the service block (12–17) which is centred on 16. In this last group of rooms, 13 may have been ancillary to the great hall, perhaps a shrine room like the comparable room at Kinheim (Fig. 24).13 That leaves rooms 18–20, which do not intercommunicate with 15–17, as a separate household. It is all a far cry from a luxury villa like Mayen-Allenz (Rhld-Pf.) or even the sophisticated Blankenheim III (Fig. 70) with its dining-room and baths. The same problems appear, commonly with far inferior evidence on which to base discussion, in many other elongated houses and also in a few unusually complex row-houses. A remarkable example of the latter kind is Cartagena (or Cabo de Palos)-El Castillet (Murcia) (Fig. 19), which is divided by two transverse lobbies into three units. Those to the west and east are nearly identical, with three rooms and a longitudinal lobby; the middle one is simpler, with four rooms; all are somewhat more developed forms of the Bierbach units. The key to understanding such villas is to break up hitherto unintelligible ranges by describing and classifying the several types of lobby used to articulate them. 69

— Chapter Five —

LOBBY TYPES The Longitudinal Lobby: a Variant Form This distinctive room grouping, comprising a room or rooms opening off a lobby parallel to the porticus, was described earlier: Wellow (Som.) is a good example (Fig. 16). Functionally it is equivalent to a transverse lobby off which two rooms are reached, as at Kirchberg. It has a variant form in which the two rooms opening off the corridor are of unequal size, so that the smaller one has the appearance of one lobby opening off another; in effect, an L-shaped lobby, the smaller part of which is separated by a wall, in which (it is here assumed) was a doorway from the larger. St Stephen-Park Street VI (Fig. 11) is an example, where the smaller lobby may have provided a way through to the rear (west) porticus without the need for an outside doorway in the adjoining large room. It has a French parallel. Hombleux (Somme) has the more elongated rooms characteristic of so many continental villas and an unusually wide porticus facing away from the courtyard. If entry be assumed on that side, the plan is exactly like Park Street. Sometimes longitudinal lobbies appear in pairs, set in reverse symmetry on either side of a transverse lobby from which, it is reasonable to deduce, they were reached; but that deduction awaits confirmation from the discovery of doorways. Vouneuilsous-Biard (Fig. 20) has a group of rooms like this on the north side of the main block, so that the longitudinal lobby further from the porticus formed, with the transverse lobby, an L-shaped approach to the room it served. At Sainte-Solange (Cher) a similar grouping incorporates two principal rooms of unequal size. In such a plan the narrowness of the spaces in question tends to confirm them as being lobbies, since any other use, given their proportions, is hard to imagine apart from fantasies about staircases. Haccourt I (Belg.) (Fig. 20) has this kind of room disposition in a modified form, the longitudinal lobby (5) being wider and more like an ante-room. In the light of earlier comments about lobbies as expressions of social difference it is worth noting that Haccourt has three kinds of lobby and that the south end (rooms 7–11) resembles Faversham I or Ridgewell (Essex) adapted to the greater width of a continental house. The L-Lobby A variant of the quasi-L-shaped lobby has no doorway between the two parts and so wraps unbrokenly round two sides of a room. To understand how such an arrangement functioned requires positive evidence of where the doorways were and, equally important, negative evidence showing where they could not have been, because the several positions possible imply different ways of use: whether the doorways were at the ends of the lobby or near the corner of the room matters a good deal. Unfortunately such evidence is scarce. A striking example of the L-lobby occurs at North Leigh-Shakenoak B IIA (Oxon.) (Fig. 20) where, assuming the porticus to have an entrance in the middle, another doorway opposite it would have led straight into the lobby. What was the 70

— Developed Forms of Row-House —

Figure 20 Lobbies and related forms


— Chapter Five — purpose of guiding those entering the house into a minor space rather than, as in many villas, an important room? Two factors may have produced this situation. The first is that the house is essentially like Maulévrier in being divided into two not quite equal parts, but differs from it in that the two units have been moved sideways to accommodate a lobby; and the second is an unusual degree of concern for the privacy of the superior unit in a house of this size. The lobby may well have given access to the other room of the middle range which had a hearth and appears to have been given over to workaday activities including, perhaps, cooking. That leaves the porticus providing access to the cellar and what were structurally wings, not pavilions; the general appearance of the house is conveyed by the model from Fontoy-Moderwiese (Lorraine) (Fig. 20). Lobbies like that at Shakenoak occur at North Wraxall (Fig. 44) and in the little house at Hales (Staffs.); both lead to the principal room, whereas a similar lobby at Folkestone B does not. The L-lobby could be used to achieve somewhat different purposes, as appears from the two ways it could be situated. At Shakenoak and its like the room approached from the L-lobby is at the rear of the house, whereas at Latimer (Bucks.),14 Piddington (Northants.) and Selongey (Côte d’Or) it was at the front; the former looked out over a garden or fields, the latter over the farmyard. Selongey is different again; the room reached from the L-lobby may in [I] have been the principal of two units. This difference of detail is important and may be compared to the way the two longitudinal lobbies at Beadlam N (Yorks.) (Fig. 17) and High Wycombe (Bucks.) (Fig. 17) are approached. In all three instances it has to be presumed that one lobby was entered by a normal-sized doorway and the other by a much larger opening of the same size as the structural members framing it; and it was itself framed, perhaps, by pilasters of wood or plaster. This was one of the many ways the variations in the status of the units accommodating households were distinguished. Many houses have two lobbies but comparatively few have two of the same type, so that, in Britain, a transverse lobby in one unit is commonly matched by a square one in the other. L-shaped lobbies are much less common but Upchurch-Boxted (Kent) – a villa remarkable for the variety of ways in which rooms are articulated – has two, albeit with unusual proportions because the lobbies are so wide that one of the rooms is reduced almost to insignificance. Piddington resembles Boxted to the extent that it combines an L-shaped lobby15 with a quasi-L-shaped one, that is, a longitudinal lobby opening off a transverse one. L-shaped lobbies, on present showing, are rarer in continental villas, although discoveries in recent years suggest that more will appear. One such, Selongey, has already been mentioned and Hamblain-les-Prés (Pas-de-Calais) (Fig. 67) appears to be another. It has, in the largest structurally undivided cell, an unexplained L-shaped space which encloses and, so far as the plan indicates, is not partitioned off from what is called a floor (aire); this must be an L-lobby.16 Champagnole (Jura), in the part excavated, has one cell similarly subdivided. In Germany at Stadtbergen I (Bay.) (Fig. 20), an L-lobby may have existed from the first or it may have been added shortly afterwards. In making comparisons between continental and British villas their respective sizes should be remembered; Champagnole has, adjoining the L-lobby, a hall-like 72

— Developed Forms of Row-House — room some 15×8 m which is not far different in size from the open hall of an English villa. The formal approach adopted in this book takes little account of such matters, important as they are, because they cannot be considered systematically until the groundwork for comparison has been laid.

Lobby, Room or Both? Lobbies vary in width relative to the rooms they serve or enclose. Sometimes what is formally a lobby is too wide to be a mere passage and in a few cases is clearly as important as the room itself. At Zofingen two rather narrow elongated cells are described as having ‘small enclosures’17 within them, or a room within a room; yet, that said, the inner rooms are so small and so elongated – one is 4.8×2.4 m, the other 6×2.1 m – that they must surely be service rooms, but to what? Puzzling though these two subdivided cells are, they must have been of some importance because they are in reverse symmetry. Fremersdorf, discussing the Köln-Mungersdorf villa, drew attention to a comparable, though not identical, grouping of three rooms (13/14, 15, 16) at the middle of the house (Fig. 71); the undivided space 13/14 may be called a widened lobby. He noted the lack of a doorway between 13 and 14 as something unusual but certain, there being no foundation for a threshold between them,18 and explained the rooms thus: entrance by 14 from the front porticus led to the passage (room) (13) and on either to the rear porticus or through the service room (15) to the kitchen (17);19 the structural block or bay is completed by the dining-room (16). This is the only explanation that has been offered of how such a lobby/passage might have functioned and virtually the only attempt to emulate Oelmann on Blankenheim. It is not satisfactory insofar as it involves two passage rooms (13 and 15), and 17 is an unlikely kitchen, but Fremersdorf deserves credit for this all too rare essay in interpretation. Blankenheim in phase IIIA (Fig. 70) acquired a group of rooms closely resembling those at Mungersdorf. A new and much smaller principal room (35) was created and, because of its wide doorway, was interpreted as having the same representational role as its much larger predecessor (33/38) but also as the oecus or dining-room. No explanation of the adjoining rooms, a corridor (34) widening at the rear into 33 from which 36 – a ?service room – was reached, appears in Oelmann’s report. The space at the end of the corridor (33) may have been what was known in English country houses of the seventeenth and eighteenth centuries as a waiting-place, where servants congregated after a task was done to await their next instructions. In those houses another requirement is recorded that may have had a counterpart in the better houses of the Roman provinces, that of a surveying-place where a supervisor could monitor movements between kitchen, food stores and dining-room in order to ensure efficient service and, no doubt, deter theft.20 That cannot have applied at this particular spot but such a function should be borne in mind. An Italian ancestry might be expected for an element of the plan far removed from the comparatively simple, though subtly expressed, requirements of earlier villas in the north-western provinces. Although no systematic search has been made, 73

— Chapter Five — the villa of Orazio has an arrangement comparable with that at Mungersdorf, in the cell to the east of the large middle reception room.

La Roche-Maurice Widened lobbies of any kind are not common. They appear in two very much smaller French villas, La Roche-Maurice/Valy-Cloistre (Fin.) (Fig. 20), which is the only one fully published, and Saint Acheul W (Somme). Its plan, which is said to be ‘perfectly logical’,21 has been equated, in a misuse of terminology remarkable even by the customary standards of archaeologists, with the British ‘cottage-house’. In fact, La Roche-Maurice is instructive about why L-shaped and widened lobbies are as they are, and about the close link between their shape and the ways in which rooms were distinguished to make their relative importance clear. Furthermore, because doorway positions are so rarely found, the plan deserves discussion for the light it throws on much larger row-houses. As in so many houses, the middle room has a wide entrance; here 5, unusual by its shallowness and smallness, is dominated by a 4 m-wide opening to the porticus and must have served some representational purpose. The doorways into the lobbies 4 and 7 were close to the wide middle opening and carefully differentiated by width – 1.2 m and 0.8 m respectively – so that anyone approaching knew at a glance which one it was appropriate to use. The narrower of the two led into the lobby (7) and so through another wide doorway (2.2 m) into the hall (8), which was a kitchen and may have had something of the character of a workhall. Facing its entrance was the next widest doorway into a smaller room (6) which was no doubt the most important in the house from a social standpoint. Beside the doorway is space enough for servants to stand or sit as they awaited summons. On the less important side of the house the lobby (4) apparently led only to the narrowest of the internal doorways and the smaller of two rooms, 3. Restrictions on excavation precluded discovery of the doorway into the remaining room (2), which was assumed to have had an external entrance in the south wall.22 If this is correct it has functional consequences; it suggests some use connected with the farm, rather as one pavilion at Marshfield (Glos) was used as a smithy. Nothing was found to show how room 2 was used and it may have been a workhall having the same relation to 3 as the hall (8) had to the innermost room (6). The villa of La Roche-Maurice was interpreted very differently by its excavators. In their opinion the open-fronted middle room was where a landed proprietor periodically sat behind a table to collect rents in kind from tenants waiting in the corridor, who then made their way out by a kind of ‘one-way system’:23 an astonishingly explicit equation of Roman villa and nineteenth-century country or manor house, and at the same time a confusion of functions as between owner and bailiff. But when far-fetched comparisons are set aside, what was the composition of the houseful? It was obviously far removed from any hypothetical kin-group comprising more or less equal households. Fundamental to any interpretation is the way the lobbies divide the house into two parts, leaving the shallow room 5 with its inordinately wide opening 74

— Developed Forms of Row-House — as part shrine to the household gods, part representational room, in common use. Beyond that, much depends on whether the rooms are envisaged as being either living-rooms or bedrooms or as having a dual purpose. Room 8 is the easiest to interpret. To judge by its size and open hearth – in both respects it is much like the hall of the Frocester Court (Glos.) villa (Fig. 69) – it is more than a living-room for the occupants of the principal private room opposite; rather, it can be envisaged as a hall for a household living and working there. Across the lobby, room 6 provided for a somewhat superior household which may well have shared the hall. At the lower end are, again, two rooms, 3 being for a household of lower status and 2 given over entirely to work except, probably, for housing the persons whose work it was; and if this room had only an external entrance, theirs was a menial position. All that can be said about Saint-Acheul (Somme), given an inadequate report and small-scale plan, is that it appears to have at the south end two rooms and a lobby resembling rooms 5-7 of La Roche-Maurice, and, forming a near-mirror-image to it, three more rooms.

ROOM PROPORTIONS AND THEIR IMPLICATIONS Lobbies in their various forms occur throughout the western provinces and the houses in which they appear are recognisably of the same generic type. Nevertheless, as between British and other row-houses, a difference is apparent: in the former, rooms are mostly squarish; in the latter they are commonly oblong and sometimes almost passage-like, as in Haccourt I and Puig de Cebolla (Valencia) (Fig. 20). Rooms so different in shape cannot have been used in the same way. An oblong room entered in the middle of a long side is thereby divided to some degree into two parts; entry at one end automatically creates lower and upper ends, the former being to some degree a passage space to reach the latter, the more private inner part. These considerations are less true of a square room. Entry near a corner can be aided by the position of a hearth or fireplace, by furniture or by a mosaic to create an upper end or superior part, but such effect does not stem naturally from the shape of the room; in an oblong room it is unavoidable. Where an oblong room appears to be a representational room, like that at the middle of Laufen-Müschhag (Switz.), it can reasonably be assumed that the far end from the porticus was the point of importance where there was a place of honour for the household deities or the head of the houseful. It is not easy to see how a room measuring 4×9 m could be used for feasting and entertainment of a rather formal kind, although its proportions are not very different from the middle room at Hummetroth. Conceivably some non-structural distinction of function existed, as, for example, between an area where offerings were made and the remainder which was used for ceremonial dining on religious or social occasions. Dual use is possible and to some degree likely in any oblong room. The shapes and conjunctions of rooms will have conformed to their intended use, even where, as must often have been the case when buildings were altered, architectural considerations limited the possibilities. To understand room use at all several assumptions have to be made, the most important of which is that each unit 75

— Chapter Five — or apartment was occupied by a separate household. A household, which varied in numbers according to the life cycle, is here taken to be a group of persons occupying an apartment by virtue of status within a kin-group; it might be a conjugal family or an elderly couple or even a single person, according to circumstances. A further assumption is that at least one personal servant slept in each apartment. In provincialRoman society under-employment is likely to have been endemic and servants correspondingly plentiful; and before the introduction of bells operated from a distance, servants necessarily had to be close at hand at all times. In many, perhaps most, households, both master and mistress will have had valet or maidservant in attendance at all times and one or two more servants available on call. Room and lobby shapes all had to provide for families in the early modern sense of the word, which included servants.

ARTICULATION OR SEPARATE ROOMS? The subdivided cells punctuating the rows of large ones are of very varied kinds and it can be hard to tell room from lobby. One guiding principle is that passage-rooms are unlikely to have found favour. A particular pitfall in the present approach is how to distinguish a narrow space leading to a room – a lobby – from a not much wider one which served some other purpose in addition to intercommunication, or was not a passage at all but a narrow inner room. In villas of considerable depth from front to rear the division of a cell into two compartments by a longitudinal wall need not imply that the smaller one was a lobby. The point is illustrated by Liestal-Munzach (Switz.), where unusually detailed evidence is available (Fig. 20). A transverse mosaic-floored lobby (4) leads to a room with a hypocaust and mosaic (1) – a well-appointed and comfortable room, approached in suitably dignified fashion and shielded from the curious at the front of the house. Between it and the porticus was, not a lobby, as might be argued from a less detailed plan, but a small room entered only from the porticus. To some extent proportions provide a safeguard against wrong interpretation, and only where the small room is quite narrow is it virtually certain to have been a lobby. Even then, the alternative of a passage from one flanking cell to another remains a possibility, provided some plausible need for it can be suggested. This, incidentally, is the objection to the common interpretation of transverse lobbies as passages, in houses with no rear porticus like Frocester Court I: what is the point of an enclosed way from front to back? Where the rear space of a subdivided cell is wider, it may not have been limited to transit but be best described by the term ‘anteroom’, a place where people wait before entering a superior’s room or where some business is first transacted; but since those purposes cannot be discovered archaeologically the word will rarely be used. These considerations explain why a simple form of lobby may be hard to identify, as some examples show. In the north half of Anthée (Belg.) (Fig. 19) the cell comprising rooms 40/41 looks exactly like a room and longitudinal lobby, yet the doorway positions show this to be wrong. Room 41 is entered from the west porticus, and doorways into 40 were not found, so that even if the two rooms intercommunicated they cannot have been 76

— Developed Forms of Row-House — related as lobby and inner room. Nor is 41, as its position might suggest, a superior room reached from the transverse corridor and so placed for privacy, like the one at Liestal. On the other side of the corridor, room 46 looks as if, with 40 and 41, it formed a unit comparable to the domestic end of Schupfart I; yet it was entered at both ends and cannot have been used in the same way. Although villas of the size and, particularly, the complexity of Anthée present unusual difficulties, the articulation of simpler plans only one room in depth is not always easy to understand, as Stadtbergen I (Fig. 20) shows. It had two rooms, 5 and 11, which might be considered longtudinal lobbies to the adjoining larger compartments (6 and 12) of their respective cells and which were approached by the transverse lobbies 7 and 13; and, although 7 is uncomfortably wide for a lobby, the two blocks of rooms can be seen as two units or apartments. A third comprises the lobby (3) and room 4. That leaves 8, 9, 10 in the middle, bearing some resemblance to the core of North Leigh-Shakenoak IIA, to account for. But an alternative interpretation is possible if Reutti’s opinion, that 7, 8, 9 were originally one room, be accepted. That provides a hall-like room fronted by a portico of four columns in the middle of the house, and two threecell apartments entered at the ends by 3 and 13.24 So far, an appropriately classical, symmetrical exterior; to go beyond that and interpret the apartments presents difficulties. The north end is easy: entrance by lobby 3 into the living-room 4 at the west end and thence to the lobby 5 and inner room 6. This pattern is not repeated to south, where the lobby 13 leads to the longitudinal lobby 11 and so to the inner room 12, but leaving the large living-room 10 approached by 11, which is something it would be hard to parallel. Neither solution is wholly satisfactory, and only further evidence of doorway positions will bring certainty. Stadtbergen is an instance of the problem posed by Drury with reference to simpler plans,25 and Farningham-Manor House I (Kent) would be another if no doorways were known.

BLOCKS OF SMALL ROOMS A number of villas incorporate a group of four small rooms of different sizes at one end or, occasionally, both ends. Köln-Mungersdorf and Newel (Fig. 24) are examples in hall derivatives. A rare instance where such a configuration produced evidence of doorways is Les Mesnuls (Yvelines) (Fig. 21), where a block of four rooms (1–4) was quite separate from three larger rooms, all linked by doorways, which form the rest of the house. A strange feature of the plan is that the sole room with a hypocaust is only 3.5×2.8 m, so small that it ought to be what Black has distinguished as a hypocauston – a small room giving indirect heat to the adjoining rooms;26 yet its warmth benefited only two lobbies and the porticus. This oddity apart, the four rooms look like an apartment for a household of superior status to whoever occupied the ?representational hall 5 and the unit comprising 6 and the workhall 7.27 Similar groupings of rooms, bigger than those at Les Mesnuls but small by comparison with the three very large principal rooms, occur at both ends of the Swiss villa of Kulm. At the west end are four intercommunicating rooms reached from a transverse lobby. A group of five rooms at the east end is better articulated by a 77

— Chapter Five —

Figure 21 Unit variations in row-houses

transverse lobby giving independent access to four of the five rooms; perhaps one did not belong. In this group only two rooms intercommunicate. Although these two blocks are not easy to explain in detail, the differences between them, despite their similar size, and their siting at opposite ends of the house, must correspond to the existence of two households of high status. Kulm is a highly developed row-house on the grand scale, suitably enriched with mosaic floors and marble panelling. For a villa published as long ago as 1761 we know a good deal about it.28 The end blocks at Kulm are unequal. Other villas have somewhat simpler variations on the same theme, notably Olfermont (Haut-Rhin) (Fig 41), where four small rooms are grouped together at what looks like the east end of the original main range.

TRANSFORMATION OF LOBBIES The idea of transformation has already been used to create the compact variant of the row-type house; it is applicable to lobbies too, as has already been implied with a comparison of Wellow and Kirchberg (above, p. 70). A very simple case is the lobby off which opens a second, shorter lobby from which the room within it is reached. The first lobby can be either transverse or longitudinal; assuming that the room is entered by a doorway at the far end of the combined lobby and that the focus of attention is in a constant relation to it, the only difference made by turning the lobbies through 90 degrees is to the lighting of the room. In a room of some importance, that will affect how those entering will see anyone occupying a place of authority, whether in direct light from a window or against a small amount of indirect light from the porticus. Something comparable can be envisaged with the small square lobby of the type found at Newport. There it and its adjacent small room are set transversely to the main body of the house – they had to be, because the range in which they stand is narrow. In a larger continental villa it was possible to turn the lobby-and-room around to lie parallel with the porticus, although no villa so far has revealed where the doorways were. But such a transformation would make a small part of Anthée (Fig. 19; rooms 55, 56, 59) intelligible, and also Saint-Julien and Aubigny (Somme), known only from aerial photographs, may be susceptible of the same explanation. Caution is necessary in applying this kind of interpretation to pairs of small rooms at the rear of a cell, at the far end from the porticus. Only if there is also a rear porticus, 78

— Developed Forms of Row-House — or a transverse lobby gives access to it, can one of the rear rooms be regarded as a lobby; in Switzerland, Sarmentsdorf (Fig. 21) is an example. It is not obvious that the lobby-and-room are secondary, as Drack supposes,29 but even if they are, the interpretation may stand. Otherwise, and this is the usual situation, they must be a pair of inner rooms at the end of a comparatively large living-room facing either a porticus, as at Trouey (Cher), or a peristyle, as at La Roquebrousanne (Var). In that position the two little rooms stand in the same formal relation to the larger room as do those at Quinton (Northants.) or Rothselberg (Rhld-Pf.) to large open halls.

THE SIGNIFICANCE OF DISPARATE UNITS Row-houses comprising two units with identical lobbies occur fairly frequently, Boos, Downton I, Lamargelle (Côte-d’Or) and Kirchberg [I] being examples.30 Many, though, are like Faversham I or Ditchley I in having two lobbies of different form, placed in series or back to back. In houses with a large middle room, like Ditchley, there is comparable variation. And size makes a difference: in the bigger continental villas a British square lobby and its accompanying room could be placed parallel to the porticus, not at right angles, while functioning in exactly the same way. If that is the situation in a simple villa it is hardly surprising that the lesser rooms in a large elongated one should show the variety of size and shape they do, nor that they are not symmetrical. Moreover, if the preceding classification of lobbies and their association with other rooms as household units is anything like correct, it breaks up long rows of rooms into intelligible groups and provides a reason why they invariably differ. That general explanation by no means accounts for every big row-house fully, for in each case there are likely to be uncertainties concerning the size of units and the purpose of particular rooms – whether lobbies, anterooms, waiting-rooms or what were called in the eighteenth century closets, which were small rooms opening off larger ones and adaptable for many purposes. It also leaves open a question about certain rooms slightly bigger than the normal cell which are not necessarily linked to lobbies and have something of the appearance of a workhall. In the kin-group hypothesis adopted here to explain the additive nature of villa planning, attention has several times been drawn to slight variations in size of the principal rooms. Different forms of lobby were another expression of those variations, one which is likely to have been enhanced by treating the doorways facing the porticus somewhat differently. These ways of displaying status were, in all likelihood, primarily for the benefit of members of the group. Strangers must occasionally have visited a villa and when they did they are likely to have been shown to the appropriate room or apartment by an inferior member of the houseful; then they, too, will have understood the code of ornament and position which expressed the relative status of the two or more households. A plan such as that of Bierbach permits the appreciation of all these points except for the ornament.31 A few long villas show little distinction between the superior and inferior ends, for example L’Ecluse-Leckbosch (Belg.) (Fig. 19) and some fairly large villas such as Fitten (Saarld), which lack any clear articulation by lobbies, are incomprehensible in detail yet certainly belong to the category of row-type villas. 79




all houses grew from simple beginnings in ways which reduced the importance of the hall itself to the point where the typological connection is almost lost. In taking up the story from the elementary forms of addition mentioned earlier, it has to be recognised that, as with row-houses, the many variables make a tidy account of their further development impossible. A hall house could gain additional accommodation by the addition of rooms at both ends of the hall, at the rear, in wings or by encroachment within it, and all of these might be accompanied by removal of a function to another building. An almost invariable addition is the porticus-with-pavilions, combining architectural display with extra rooms whose relation to the adjoining rooms in the main range is usually uncertain; and it needs to be emphasised that addition is used in a typological sense and does not refer exclusively to a later phase of construction. End additions could take the form of a single narrow room; a number of rooms, from two to four, of uniform width, sometimes with a passage-like room as well; two or three rooms of different width, sometimes separated into two blocks by a passagelike room or entrance; three or four rooms, uniform or not, similarly separated; and all these may be present at either end of the hall in almost any combination. Virtually every villa excavation report treats them simply as a collection of rooms without meaning or significance beyond their number and the quality of their decoration. Since little attempt has been made hitherto to understand them in social or functional terms, several houses will be examined in order to reveal common elements underlying their diversity. They will be treated, not in typological order from simple to complex, but by discussion of their possible purpose.

NARROW END ROOMS The simplest kind of addition is the narrow undivided room extending the full depth of the hall from front to back. Few such rooms have produced evidence of flooring or fittings. Köngen I (ex-A) (Bad.-Württ.) (Fig. 74), exceptionally, revealed what was thought to be a stone cistern, 2.5 m in internal diameter and 3 m deep.1 Although its purpose is unknown, it implies some utilitarian or service purpose. Evidently this 80

— Developed Forms of Hall House — was the lower end of the hall, a point confirmed if we assume that the cellar was in its customary position near the upper end. Weitersbach (Bad.-Württ.) I (Fig. 69) is another house with a narrow room at each end. The one to south is entered from the hall, that to north perhaps in the same way2 but with doorways also into both a pavilion and the rear porticus; it can scarcely have been more than a wide passage. As usual, the cellar was at the upper end of the house, being placed beneath a small room which opened off the hall and did not communicate with the porticus.3 Although there can be little doubt that one end room in Weitersbach I served as a passage, the need for the function is not obvious. As will appear later, long narrow rooms are so common in German villas that other purposes must be envisaged, however difficult it is to discover them. At Manderscheid (Bad.-Württ.) (Fig. 39) the narrow room (8) could only have linked two pavilions and it was certainly not a porticus of any kind, so storage related to activities in the hall appears to be its most likely use. Similar considerations apply to Weitersbach I (Fig. 69) and Heppenheim (Fig. 6). Mundelsheim (Bad.-Württ.) (Fig. 22) has a plan resembling Manderscheid. Paret,

Figure 22 Halls with long end rooms


— Chapter Six — who dug it, was clearly troubled by the long room 5 and sought to explain it as being ‘probably divided by two cross-walls into three rooms (bedrooms?) accessible from the middle hall’;4 but no evidence of such division is recorded and parallels argue the contrary. Moreover, he does not mention an element of the plan which may relate to the end room, namely, that the pavilion 6 does not relate to the end of the hall in the usual way, like 7, but is reached by a small oblong space. It looks like a lobby, but why there? Not simply to shield the doorway into the pavilion; that would normally be achieved by the little square projection from the porticus found in some houses, as at Gayton Thorpe [F] (Fig. 39). This lobby may have been entered from the porticus by a doorway on the side nearer the hall, then a turn to the right led to the end and the pavilion doorway and one immediately to the left into the narrow room. Whatever the true situation, it suggests a fairly close connection between porticus and end room but one stopping short of the intercommunication found at Weitersbach I; and if that be accepted, a domestic or service use is far more likely than the agricultural ones sometimes proposed. Two Swiss houses have a long narrow room at both ends of the hall.5 Grenchen (Fig. 22) has a room (3) at the upper end which is about 12×4 m and one at the lower end (7) which is marginally narrower, both being devoid of floors or finds. A little privy entered only from 7 may imply a domestic use for the latter.6 The end rooms at Wahlen (Fig. 22) have similar proportions, and their description as ‘porticuslike annexes’ or ‘side-porticuses’7 does not advance understanding; it argues a disproportion between living space and circulation space that requires explanation – a problem recurring in villas like Hemel Hempstead-Gadebridge Park III (Herts.) (Fig. 13). A further description of these extraordinarily proportioned rooms as ‘decorated living-rooms’8 is reinforced by the way the house was entered; a rear doorway is diagonally opposite another into the north room, so that they stand in much the same relation to one another as the hall and inner-room doorways at, for example, Frocester Court (Glos.) (Fig. 69). This likeness is the stronger for the absence, first, of a middle doorway into the hall from the porticus and, second, of a hearth, for although one was looked for in the centre of the hall and not found, that may be because it stood nearer the north (or inner) room doorway. On the available evidence the long rooms in both villas appear to have been domestic and that must influence interpretation elsewhere.

END ROOMS AS BYRES? The longest of these end rooms, at the end of the hall-house at Betzingen (Bad.Württ.), is five times as long as it is wide (20×4 m) and was regarded from the first as a cattle-byre, ‘not only on account of its elongated form’ but also because its position conforms to Vitruvian precept.9 Since it has been presented by Applebaum in the standard history of English and Welsh agriculture as one of a number of houses incorporating rooms which are claimed, largely on the evidence of proportions and position, as byres,10 his argument requires discussion, irrespective of how closely other examples conform to Betzingen. Applebaum’s interpretation is founded on an ‘especially close correspondence’


— Developed Forms of Hall House — of plan detected by de Maeyer11 between an eighteenth-century farmhouse at Limerlé (Belg.) and the hall-type villa of Sauvenière (Belg.) which had been interpreted in a largely agricultural sense on its publication in 1900.12 De Maeyer, remarking that the ‘porticus-type’ did not wholly go out of fashion in later times – in other words, that it survived for a millennium and a half – assimilated the Limerlé house to the Roman one and accepted nearly all the room uses first proposed for the latter except that the central yard, in deference to Oelmann, became a barn flanked by stables.13 Applebaum accepted this, balking only at the idea that the porticus sheltered carts and farm implements. Chance likenesses between whole plans of houses built some fifteen hundred years apart are not likely to add to understanding, nor is another of Applebaum’s suggested byres, at Gerpinnes, convincing. There an apartment is said to be divided by ‘internal masonry counterforts’ into six stalls, one of which had painted plaster walls and which the excavators thought of as stores or lodgings for domestic servants.14 Then follows a string of British villas – Cherington, Colerne, Rodmarton, Titsey – with long compartments at the end of the hall: ‘their dimensions as well as their positions correspond so closely to the examples at Gerpinnes and Betzingen that their identity as accommodation for livestock can hardly be doubted’.15 In the end, all that remains of this particular argument is Sontheimer’s belief that the long room he excavated at Betzingen in 1905 was a byre. So it may be, for it is hard to think of any more likely use, but it has to be recognised as the conjecture, unsupported by evidence, that it is. Nor do parallels help, since they too show no clear signs of cattle being stalled in them: for although drains are not essential to byres16 it might be expected that one or two would have them, or a mucking-out hole as in Welsh long-houses, or two or three doorways to avoid the inconvenience and waste of space caused by cattle moving along them. Among the parallels for this feature of Betzingen is Neckarzimmern-Stockbronner Hof (Bad.-Württ.)17 (Fig. 27), where a long narrow room (7) on the south side of the hall was entered by a full-width opening next to a probable doorway in the west end of the south wall. The width of the opening into the room implies some nondomestic function, perhaps a cart-shed.18 A similar room (8) seems to have existed on the other side of the hall. Another example at Niederzier-Hambach 403 (Nordrh.West.), though producing no evidence of use, faces the working side of the farmyard and away from the secondary domestic building. Given the uncertainty of the evidence, it might be wise not to interpret the end rooms cited by Applebaum and others like them such as Courcelles-Urville (Fig. 22)19 as cattle-stalls until some positive indication appears.

END ROOMS ORGANISED AROUND LOBBIES Halls having two rooms at one end, or sometimes two rooms flanking a lobby, have already been touched on (chapter 3). There are more complex groupings. In hall houses, as in row-houses, any articulation of rooms is achieved by lobbies of one form or another. Thus Mayen VI (Fig. 1) included a rebuilding of the east end to provide a passage from the lower end of the hall to the garden(?), off which 83

— Chapter Six — opened an oblong space possibly divided into two rooms by a timber-framed(?) wall of which there were only scanty remains.20 No mention is made in the report of the two doorways shown on the plan which, since they would have proved the rooms’ existence, are presumably an interpretation rather than a factual record of the excavated remains. Oelmann’s judgement was not influenced by any comparisons, yet now Evelette (Belg.) (Fig. 23), a house where the hall is less dominant than at Mayen, has exactly the same combination of lobby and two narrow rooms at the lower end; a combination also seen in a different context at High Wycombe (Bucks.) (Fig. 17) in the two pairs of rooms of unusual proportions, each reached from a lobby or passage. A more striking likeness between halls and row-houses appears in the domestic end of Schupfart (Fig. 71) and the apartments of Küttigen-Kirchberg [I] (Fig. 19) or Vicques [I] (all Switz.), all of them having groups of three rooms organised around a transverse lobby; and it has been suggested that the incompletely explored east end of Tholey-Sotzweiler (Saarld) (Fig. 23) was like this.21 This kind of plan was developed at Newel (Rhld-Pf.) (Fig. 24) by replacing the largest room with two squarish ones (5 and 6) and having a short lobby (8) branching off the main one (4).22 Schupfart III (Fig. 71) turns this arrangement through 90 degrees, reduces the lobbies

Figure 23 Halls with complex end blocks

— Developed Forms of Hall House —

Figure 24 Halls with double-depth end blocks

to an L-plan and adds a small fifth room, a hypocauston, providing indirect heat to the adjoining rooms. Those who lived in the end blocks of Schupfart and Newel must have organised family relations in much the same way as their counterparts at Kirchberg and Vicques. Yet another hall house with rooms looking like a row-house is Koerich-Goeblingen 1 I (Lux.) (Fig. 1), where at the east end, a lobby like that at Newport (Fig. 10) is flanked by two square rooms. The result looks exactly like the principal part of the little house at Drax (Yorks.) and is no different in principle from the more elongated Langton Dwelling House II (Yorks.) (Fig. 3).


— Chapter Six —

END ROOMS: INZIGKOFEN AND ITS ANALOGUES A cursory glance at German hall house plans reveals that few have the same number or size of rooms at each end. One explanation could be that the two ends were given over to different functions, the obvious analogy being the parlour and service quarters at opposite ends of the medieval English hall – not that even that distinction is clear in every instance. Another is that their occupants differed in status: family and servants, parents and children, are possible divisions of the household, or functional and social components could be combined in different ways in the two ends. These considerations underlie the following analysis. The most elementary division is between houses where the two ends are more or less equal and those where they are conspicuously unequal. In the first category, if the simpler ones with only two rooms such as Leutersdorf [I] (Rhld-Pf.) (Fig. 45) be excluded, Inzigkofen (Bad.-Württ.) (Fig. 27) is the best example. It was not entered in the same way as the similar house at Stuttgart-Stammheim (Fig. 27), at the end of the rear aisle; rooms line the hall from front to back. At the west end the narrow room 5 is probably an annexe to 6 and entered from it, because there is insufficient clearance to allow a ‘bridge’ like that which appears to have existed at Stammheim (below, pp. 100–1). At the south end the square room 11 and the smaller one 10 may correspond to the room and annexe 6, 5 opposite, while the third room (9) is of quite different proportions, more than twice as long as wide. A feature these rooms have in common with those of row-houses is that hardly any two are the same size, and the difference between them is so small that some social purpose must underlie it; and the exception, 9, is so much bigger than any other that it must have been used for common activities, forming public rather than private space. Since this room cannot have been entered at the porticus end, and an entrance in the middle would be uncomfortably close to the cellar staircase, the doorway is likely to have been at the west end and, like that to 6, facing the middle of the hall. Which was the superior end of the hall, north or south, is not obvious from the plan alone, and the unusual feature of two ramps implies either a partition dividing the exceptionally large cellar into two parts, which was not found, or common use of the whole. Using the analogy of row-houses, Inzigkofen provided for at least two households, north and south, but more likely three with a further one of unknown size, and of lower status, in the hall. A measure of equality between two sets of rooms is expressed differently at the end-entrance hall of Mauren (Bay.) (Fig. 23): a hall, that is, with porticus-and-pavilions on one of the shorter sides. The hall is flanked by two ranges of rooms identical in depth but of markedly different width. Presumably each set fulfilled very similar purposes for different numbers of people or, more likely, for people whose different status was expressed in terms of space and comfort. Entrance to the house proper was by three openings of graduated size between columns, the widest into the hall, a narrower one to the right-hand rooms23 and the narrowest – the most restrictive and exclusive – to the bigger left-hand rooms. The right-hand range is like its counterpart at Inzigkofen, first a lobby-and-room unit, then one of two rooms, a pattern repeated opposite. Whatever functions were performed in room 9 at Inzigkofen were either carried out in the hall at Mauren or not at all. The differentiation so 86

— Developed Forms of Hall House — carefully observed in the rest of the building did not, curiously enough, extend to the pavilions, which, unlike those in many villas, are of equal size. Despite this, Mauren reinforces the impression of duality so commonly found in villas. No two houses of any size are really alike. The most that can be expected is that they show resemblances, as Schambach [I] (Bay.) (Fig. 27) does to those just discussed. One feature which sets it apart from similar broad German villas is the slope from rear to front, so sharp that it is said to have been terraced – something which is very hard to envisage.24 It has a probable row of three columns and two responds at the rear of the hall, their positions being marked by gaps in a stone kerb or foundation, and at each end is a row of rooms, four to north, three bigger ones to south. The latter are interrupted by a passage from the south porticus by which the house was entered; the west porticus probably stood over an undercroft, like Stammheim, or a crypto-porticus like that at Buchs (Switz.). Surprisingly, the west side lacks pavilions; perhaps a doorway can be envisaged to separate the two porticuses on the lines of the arched opening at Ormalingen. Even though Schambach is different from Mauren in several ways, they have a family resemblance in the width and number of rooms in their respective ranges, and in their large halls. Whatever kind of ownership be envisaged for the inhabitants, it can hardly be the conventional landed proprietorship proposed here as at other villas as unlike it as La RocheMaurice (Fin.) or Aylesford-Eccles. The social implications of these houses can be carried further. Inzigkofen, Mauren and Schambach all have two sets of rooms which can properly be called apartments and are separated by a room of common use so large that they are appendages to it. This resembles the situation in row-houses with a large middle room except that there the work element is transferred to a separate workhall. There is nothing to suggest that either row in Inzigkofen and its like was devoted solely to service or storage or work purposes unless the relative width of those at Schambach be interpreted that way. At Mauren inequality of the two sets of living accommodation is obvious enough in a plan yet in the building itself only the graduated entrances revealed it; within the hall they are likely to have presented, if not a symmetrical appearance, at least a uniform spacing of openings. The dual activities of the occupying group, however shared, were so important as to find expression in the cellars: two staircases at Inzigkofen, a division into two parts at Schambach. Although dual use is apparent in all of them, the three houses reveal the direction of change, from near-equality of status at Inzigkofen to the predominance of one unit at Schambach. When all three houses are compared with simpler hall-house plans like Serville (Belg.), Quinton (Northants.) or LudwigsburgPflugfelden (Bad.-Württ.), it is apparent that an altogether more complex and potentially more stratified society is emerging. Very few house plans, excluding the simplest, show true symmetry. Among the small number of villas in the Netherlands are several which come near it, and one, Buchten (Fig. 23), was built symmetrical and never altered. At each end of the large hall a passage between two square rooms leads to a sizeable hall-like room (7×4.5 m), the whole forming a three-room unit essentially like those at Kirchberg but entered differently. Just what that difference signifies is not clear but for the present 87

— Chapter Six — the likeness, bringing together elements of plan not hitherto thought to be connected, is sufficient.

DISPROPORTION BETWEEN THE ENDS OF THE HALL In many houses inequality between the ends is apparent even in the first building phase. Nuth-Vaasrade (Fig. 23) is a Dutch example; Manderscheid (Fig. 39) is perhaps an extreme case, having at one end three small rooms, two with hypocausts, and at the other the undivided room mentioned earlier. Otherwise it was symmetrical, with porticus-and-pavilions on both display front and entrance front, and two doorways on the north side. This unusual degree of symmetry makes doubtful the assumption that the simpler end was no more than a service room or servants’ accommodation. The two doorways might, at a pinch, be thought to provide for family and servants, but if that were so, two external doorways would be even more necessary. In fact, here as much as at Inzigkofen, everything speaks of equality except the long narrow room, which was perhaps living-space, as was conjectured at Wahlen or Grenchen, though here markedly inferior. Stuttgart-Stammheim (Fig. 27) shows less discrepancy between the two blocks of small rooms. At the north end of the east row (the upper end) 3 has, surprisingly for so small a room, a doorway in the middle. To south the largest room (5), linked to the hall only (on the archaeological evidence) by a narrow walkway, surely had a better approach from the hall by a bridge over the cellar staircase to a doorway in the corner. Annexe 4 may form a unit with 5, leaving 3 as a room with some quasipublic function. These arrangements are superior to those of the west end, where only two square rooms – 10 with painted plaster – were domestic. Access to 9 is restricted, as with its counterpart to the east, by a cellar staircase. A room at the north end, next to the entrance and open to the rear aisle, served some work purpose and confirms the inferiority of this range to that opposite. Although a long narrow room is probably inferior to several smaller rooms at the other end of a hall, it is not always easy to be sure. At Mundelsheim (Fig. 22) an undivided room (5), already argued to be domestic, faces a row of three, of which 2 and 3 form part of a bath suite, respectively hot room and warm bath. Room 1, said to be a cold bath, is entered from the pavilion,25 which can thus be inferred to be part of the bath suite (dressing room?) or a passage-room leading to it. It thus appears that either room 5 was the only apartment or the pavilion (7) and room 1 in reality form a superior one. Some villas with principal houses having the same general disposition of rooms as Mundelsheim show other clear signs of two households. RheinbachFlerzheim (Nordrh.-Westf.) (Fig. 23) has at one end of an elongated hall what look like two apartments with adjoining lobbies, and at the other end two rooms perhaps forming a third apartment. More important than their details is the small ornamental basin blocking the conventional point of entry in the middle of the porticus. Its placing implies that there were two entrances, and the only possible point of that is to provide separate access for the two households (at least) which comprised the houseful.26 88

— Developed Forms of Hall House —

LARGER APARTMENTS: BOCHOLTZ-VLENGENDAAL Those houses so far mentioned are all recognisably of hall type, in the sense that the hall dominates the rest. That is less true of a few German and Dutch houses which have sufficient other rooms to reduce the comparative importance of a large square hall. Not all the additions are living-rooms; a bath suite is usually included, sometimes also a workhall, and the presence of the latter raises difficult questions concerning the division of function between it and the hall. First, though, the living-rooms will be analysed to see how their grouping relates to the dispositions observed in the preceding sections. The Dutch villa of Bocholtz-Vlengendaal [I] (Fig. 24) accommodated two households, as the two entrances into the porticus and the large pool blocking direct access to the hall demonstrate. To west is a block of three rooms and a fourth with one intruded corner which is the key to the circulation pattern. It is, in fact, a large lobby, almost an entrance-hall, of the same shape and serving the same purpose as room 22/25 in the five- or six-room south block at Blankenheim IA (Fig. 70). And if the Rhenish villa provides an analogy, room 16 opened off the hall, like 30 at Blankenheim, and is connected with the representational function of the hall; and removing 16 from the west block removes the need for a second passage-room. In the hall (19) two projections look like bases for pilasters beside a grand entrance.27 They would not be at all surprising in that position if the hall were entered from the rear; although if that were so, external pilasters might be expected on that side too, as at Friedberg-Pfingstweide 1 (Bad.-Württ.) (Fig. 41), and as occurs in more elaborate form at Köln-Braunsfeld (Nordrh.-Westf.) and Hartlip K (Kent). They presumably formed an impressive frame to the doorway into the porticus, but the purpose of that by itself is obscure. At the other end of the hall is a simpler, more public group of rooms reached from the east end of the porticus, not by a doorway but a wide opening which, no doubt, was carefully differentiated by timber and plaster ornament from the more private and probably smaller doorway at the west end. It led into a large, hall-like livingroom (23) off which opened two good-sized rooms (20 and 21), larger than the end rooms of Inzigkofen and its like; they were, perhaps, independent living-rooms comparable to the rooms opening off transverse lobbies at Hérouville-Lébisey (Calvados) and other simple row-houses. The social basis of Bocholtz changed very considerably by [F]. The two doorways into the porticus were blocked; the pool was filled in; a new and impressive approach was created and was presumably lined with columns of timber or brick; and on the north side a porticus (18) and the pavilion (22) were built. A bath suite was added to the west rooms, including the large room 11 which may have had something of the character of a recreational room where the social aspect of bathing was enjoyed; it may be compared to a room with hypocaust added to the villa of Leiwen (Fig. 72). Under the west end of the new porticus is a cellar entered by a staircase in room 16. This may well have necessitated closing the doorway into the middle room and been accompanied by some change of use.28 Interpreted in this way, Bocholtz reveals the same kind of change as some other villas discussed earlier, away from an original division of occupancy into two distinct 89

— Chapter Six — and more or less equal parts to one in which one part was clearly superior. An aesthetic consequence of the original dual character of the house is a symmetrical disposition of doorways facing the porticus, something which recurs in a number of the grander houses such as Blankenheim, Oberweis (Rhld-Pf.) and Bierbach (Saarld).

OTHER LARGE END APARTMENTS: KINHEIM The addition of rooms at the ends and rear of a house, whether due to an increase in the number of people living there or to improved living standards, diminishes the importance of the hall. Kinheim (Rhld-Pf.) (Fig. 24) illustrates this process in some detail. The hall, no more than about one-third of the total living space and smaller than the block of five rooms to east, shows how far that room had declined in importance. The lesser rooms present the problem encountered earlier, of how they were entered without making two of them passage-rooms. Bocholtz and Blankenheim point to the solution: 9 was a large lobby from which three rooms, 6, 7 and 10, were reached; the remaining room (8) has a doorway into the hall and there is good reason to think it did not intercommunicate with 7 or 10. The hall itself is no longer reached from the porticus (and from outside) by a doorway in the middle, and if there was a central hearth it did not relate to an inner room; the social significance of its placing has gone. Instead, the lower end is occupied entirely by the cellar stairway and two furnaces for hypocausts and neither the south nor the west wall is suitable for a dais. Although the hall has evidently lost much of its importance as a focus of daily life, it provides the only access to 8, a room central in every sense to the life of the villa. Here, in the middle of the room, was found a large stone sculpture of the god Sucellus; here, not the hall, was where religious observances were conducted. This explains the function of the corresponding rooms in other villas, such as 30 at Blankenheim I/II (Fig. 70)29 and 16 at Bocholtz; all stand in the same relation to the hall and the entrance to it. Given that 9 is a lobby and 8 did not form part of the principal suite of rooms, what of the remainder? The end room (6) was presumably a domestic hall or workhall serving some of the functions once carried on in the main hall. Room 7 presents problems as to how it was lit, being entirely surrounded by other rooms. The explanation, which has a much wider bearing than Kinheim, must be that in the Roman Empire, as in seventeenth-century England, the general need for light was far less than it is now, and the ability to move easily about fairly dark spaces general, for otherwise lobbies would not have been as common as they were at both periods. In all houses of complex plan much use was made of borrowed light, and a window or two, not necessarily glazed, in the workhall (6), set opposite grilles in 7, would have been sufficient to meet most needs. This applies to rooms at Bocholtz, Raversbeuren (Rhld-Pf.) and other big hall-derivative houses. As for the function of 7, it was probably a living-room serving, as such rooms no doubt usually did, as a bedroom; 10 was probably the same. What of the west end? Mostly it is taken up by a bath suite, leaving only 14, over the cellar, for any other purpose. That may have been where the equivalent of a butler or 90

— Developed Forms of Hall House — major-domo, who need not be envisaged as an employee, lived. The bath attendants and other indoor servants are likely to have slept in the hall. So whereas Kinheim [I] was a large hall with few smaller rooms, [F] saw a concentration of private rooms at one end of the house, with rooms 6, 7, 9 and 10 forming one suite and the two betterappointed rooms with hypocausts another. Its religious aspect was the only important function associated with the hall; all services were grouped at the far end, where lived the inferior members of the houseful who performed them. Schuld, a house with a complicated architectural history not elucidated during excavation and virtually impossible to sort out now, has, as the east end of the hall, a complex block of rooms including two with hypocausts, and two more are at the opposite, severely eroded lower end. The process observable at Bocholtz and Kinheim is not particularly evident here; the hall has lost some of its importance and nothing like a shrine room is identifiable, but the two elements of a kin-group which are presumed to have been the origin of the double-ended hall are still discernible.30 Social development was uneven.

SOCIAL SIGNIFICANCE OF ROOM GROUPINGS In villas like Kinheim, Bocholtz and Blankenheim I the blocks of end rooms are articulated by different kinds of lobby from those employed in row-houses. Two of these houses have the stubby L-shaped lobby which needed a complementary smallish oblong room, 4 at Bocholtz, 18 at Blankenheim. There is no obvious need to intrude the pavilion or the workhall into the body of the house; why this complication? That it was felt necessary arises, on the one hand, from careful assessment of room requirements in the various units comprising the end blocks and, on the other, from the constraints imposed by the canons of classical architecture. Blankenheim I, where some of the relevant doorways were uncovered, illustrates the clash between these conflicting demands. Division of the functions carried on in the putative pre-I hall between a workhall and a representational hall, coupled with the wish to introduce a religious function at the heart of the house, required, if ceremonial dignity were to be achieved, that some distance be put between the two rooms. Putting the workhall in a wing ensured that everyday business and the callers connected with it were away from the main residential rooms and the hall; the convenience of this arrangement is demonstrated by its adoption in the very different kind of plan at Farningham-Manor House I (Fig. 15). That was easy. How to reconcile the presence of so large a room with a symmetrical classical front? The solution was to put into the north wing the largest and most important domestic room (60/62) of the house. It is oblong, comprising the square living-space usual in such a room with circulation and service space opposite the doorway. That customary square form was one limitation of the possible length of the wing; another was the implied need for an impressive doorway facing the porticus. Together they dictated the length of the major wing, and the point at which the room was entered made a return of the porticus unnecessary – it may even have been thought to preclude it.31 So, if the workhall had to be a particular size to perform its functions, identical wing length could only be had by pushing the workhall back into the body of the 91

— Chapter Six — house. That in itself did not require an L-shaped room but, given the width of the workhall and the need for some smallish rooms of customary rather than oblong proportions, the shape of the lobby/entrance-hall became inevitable. The lobby (22/25) was circulation space and nothing more. Whether the division between it and 21 amounted to their being separate rooms or was only a wide opening is not known because the evidence had disappeared, but since 21 provided the only way into 19, some wider opening than a doorway may have existed. I am inclined to think that two interlinked spaces formed one hall-like room, off which three others opened; the largest was superior to two equal smaller ones and, appropriately, was reached almost directly from the porticus. Had it been more of an appendage to 22/25, a large inner room, the doorway would have been well inside. If room 28/29 were of inferior status to the apartment adjoining the hall to north the demands of symmetry could be met by providing the latter with its own doorway from the porticus, the former not. If this conjectural design process is correct, it explains why neither symmetry nor a wholly convenient plan could be achieved within the constraints imposed by the social structure of the houseful. Bocholtz is another house with an L-shaped room. Here the intrusion of a pavilion produced this effect and the necessity for such a design solution is harder to understand. It must be bound up with the need for sizeable pavilions which are larger than any other rooms but the hall, the desire for the entrance to be in the middle of the room, and the avoidance of strong projection and a wing-like appearance. The result may have been a quite convenient pattern of circulation, with only the small square space next to the porticus effectively acting as a lobby, leaving most of the room free for other use. Perhaps lobby-room and 12 stood in a similar functional relation to one another as 22/25 to an inner room at Blankenheim. Kinheim avoided this kind of complication completely, perhaps because its social structure was less complicated, perhaps because it was more sharply differentiated. The principal difference between the end rooms in these villas and those at Newel (Fig. 24) or Schupfart III (Fig. 71) is that the latter are more sharply divided by lobbies into separate rooms, a pair and two singles. Substitution of a passage and lobbies for a room that is part passage, part in common use implies a looser association between the occupants. In the light of that observation, the west end of Bocholtz and the south end of Blankenheim I look like an equivalent version of the small hall with rooms off it, such as Frocester Court (Fig. 69), whereas the equivalent parts of Newel and Schupfart II are coming closer to the row-house concept. What kind of social organisation does the second alternative imply? One possibility at, for example, Newel is that each of the end rooms 5, 6, 7, 9 is for a separate family, in the sense of a unit of consumption within the varied size range of the conjugal family’s life-cycle. Another is that the rooms are paired. Living-room and bedroom, the usual reaction to this idea, is anachronistic and hard to reconcile with the placing of the doorways, and the same applies to any family activities; why should they be separated so awkwardly? A third is that a single owner, a farmer or proprietor and his family, occupied all the rooms; in fact, in the currently favoured model, they and servants (number unspecified) had the whole house – and hey presto! we have the nineteenth-century bourgeois household in togas.


— Developed Forms of Hall House — The last of the three possibilities is the most unsatisfactory, the first the least, but it leaves a problem of comparative interpretation in similarly planned villas; for if doorway positions were unavailable, room 8 would be interpreted as a lobby. In the report, that room is said to be for a staircase on the grounds of its position and narrowness and a reconstruction defies the evidence of the doorway at the south end and has the stair approached from the hall.32 It is hard, though, to suggest an alternative for 8 or a use for the west end of the passage (4), beyond the doorways. Other important differences appear in the halls of these buildings. The square hall at Blankenheim I/II is an imposing room with a wide entrance from the porticus; menial tasks were conducted elsewhere. So too at Raversbeuren; the long flight of steps providing a stately approach for formal occasions is sufficient indication of that, and Schuld [I] may not have been so different even though neither the south porticus nor steps were found. Newel and Kinheim have halls of different shape, slightly oblong and entered, not in the middle but at or near one end. At Newel careful excavation showed that this was a utilitarian hall, representational on occasion, probably, but also needed for workaday functions – quite unlike the Bocholtz hall with its broad pilasters proclaiming a stately entrance. At Kinheim the two opposite doorways at one end of the hall and the two hypocaust furnaces at the other make it difficult to envisage the kind of grand use that can so easily be imagined at Bocholtz or Ravensbeuren; nor can upper and lower end have meant much. The hall is in manifest decline, lacking the social importance retained to the end by halls in smaller villas such as Mayen.

DEVELOPMENT OF WIDE-NAVE HOUSES Only two houses are worth mention under this head. Hölstein I (Switz.) (Fig.7) was open throughout its length but divided functionally. The east end was always the principal living area, entered from a porticus which terminated short of the west end in order to clear a wide entrance to the part devoted to work. In II two rooms, 12 square, 13 a double square, were cut out of the upper end. Entry was perhaps past what may have been a porter’s lodge into the larger room, and on into the inner room. Presumably they formed an apartment. In III a block of eight rooms was added at the opposite end of the hall; some formed a bath suite but it is impossible to fit them all into the conventional sequence, so Fellmann suggested that two were winter living-rooms.33 An interesting aspect of the changes within the nave at Hölstein is that the only longitudinal partition ignores the structure of the building, both aisles and axial post, in contrast to British houses, where new walls invariably utilise the rows of posts. That is true of another wide-nave house, Kaisersteinbruch (Aus.) (Fig. 67), where the three rooms at the north end, which may belong to [II], respect the aisles.





any villas clearly belong to the types or their derivatives discussed in the foregoing chapters. Among the remainder are two kinds of plan whose relation to them is arguable. The first comprises some left over from the hall type; they are like a larger Inzigkofen, either yards flanked by one or two small houses, each equivalent to a row-type unit, or inordinately wide halls with a few rooms at the ends. Structure and function alike require discussion. The second comprises simple houses, mostly of three-room plan, which do not easily fit the row type and are of similar size to the houses or units of the first group. Next, coming to light in increasing numbers, are one-room buildings which sometimes stood alone and were sometimes the first phase of a sizeable house. Commonly they reveal little or no indication of use so that the status of some of them as houses may be uncertain. The same is true of a considerable number of house plans, large and small, whose diversity masks whatever they may have in common. It is hoped to make these unusual buildings sufficiently comprehensible to fit into a general discussion of villas.

BONDORF: SMALL HOUSE OR LARGE YARD? The matter of yard or hall was discussed earlier with the emphasis on halls and the purposes they served. Now the widest of such villas will be looked at from the opposite standpoint by assuming that they were yards and considering how the rooms around them might have functioned. This is more than a debate about a handful of rather exceptional buildings; some of the yard claims implicitly cast doubt on Oelmann’s demonstration that Stahl was a hall. The reaction against this almost universally accepted view appears to have begun with Reim’s opinion that Inzigkofen (Bad.–Württ.) had a central yard1 and was heightened by unease about the problematic villa of Starzach-Bierlingen, which Baatz thought had an atrium and impluvium. So many archaeologists now adopt the yard interpretation, even for quite small buildings,2 that it is now the prevailing orthodoxy in southern Germany. More important, though, was Baatz’s awareness, in discussing Starzach-Bierlingen (Bad.– Württ.), of the need for complete excavation of the central spaces, about which little was known.3 94

— Problematic House Types — Baatz practised what he preached, with the result that his work at Bondorf (Bad.– Württ.) in 1975, written up by Gaubatz-Sattler, provided fuller detail on these problems than ever before. It has to be said, though, that despite its high quality, the report does not match those of Mayen and Rijswijk for depth of treatment of such fundamental issues as reconstruction, ownership and social structure, nor Blanken-heim for analysis of room function. In consequence, the hall/yard debate is not advanced beyond the position in the early 1920s, when some accepted the ‘Stahl type’ and others favoured yards. The arguments in favour of the yard interpretation at Bondorf (Fig. 25) are these: (1) The clear span of 22 m precludes a roof without supporting posts and the walls are in any case inadequate for the weight; (2) one of the cellar drains begins at the foot of the stairs leading down from the internal space, from which it is inferred that the cellar steps were intended to carry away water from an open yard; (3) the light shaft in the east wall of the roofed-over cellar stair gave light from an open yard, the unstated implication being that it would be useless in a hall; (4) a stone footing (feature 58) argues a roof over the south-west part of the hall and the cellar staircase, thereby providing a windshield for the furnace and storage for the fuel supply. All that is said about the hearth is that its position shows that the internal yard belonged to the household part of the building; significantly, neither hearth nor cooking equipment was found elsewhere. Of these arguments, the third appears the most convincing but, if the internal yard were really a hall, entered by the customary wide doorway in the middle of the south wall, the light from the open doors would be adequate. Point 2, drainage of a courtyard through a cellar, is inherently unlikely and would have more force if there were drains to direct water to the cellar steps. Point 4 satisfies the obvious need to protect the cellar from rain but a roof is totally superfluous if the steps act as a drain. And it is odd to provide a roof over the fuel supply while leaving the hearth it was destined for unprotected and unusable in inclement weather, the more so if the furnace opening was sheltered from the wind. Given the lack of positive evidence for a kitchen, the big open hearth is the obvious place for cooking, as Oelmann argued very cogently at Mayen; but if not for that, and bearing in mind that it belonged to the household part, what? Moreover, a yard brings other problems. The four rooms at the west end can hardly have included two passage-rooms if they were all one residence; and if they faced a yard, why no trace of posts for a veranda, as in a row-house? And what was the eastern half of the internal space, parts of which were paved, used for if not domestic purposes? What conceivable purpose can a small yard beside a part-roofed, part-open domestic area, and behind a front which was the emblem of Roman civilisation, have served? The arguments that the span is excessively wide for a roof and the walls incapable of bearing its weight are the only ones which can neither be refuted directly nor dismissed as improbable. Considering first the width, 22 m is a not impossible clear span with the hanging king- and queen-posts familiar to Roman carpenters,4 perhaps using a built (jointed) tie-beam, and if it were of the low pitches common in Swabia such a roof would not exert much outward thrust or be endangered by wind pressure or the weight of snow – that is proved by the survival of old roofs.5 Whether the walls were capable of supporting such a roof depends quite largely on their construction. 95

— Chapter Seven —

Figure 25 Halls or yards?


— Problematic House Types — Were they wholly of stone? Did posts set in the masonry carry the roof trusses? And if so, were the posts braced in any way to the tie-beams? These matters are virtually impossible to decide on the published evidence. A judgement has to be made between two situations. On the one hand is the lack of post-holes and the dubious strength of the walls; on the other, the lack of drainage in an enclosed yard, the improbability of a hearth (probably used for cooking) in the open air, and the difficulty of finding any conceivable purpose for half of an open yard. In any case, what is such a yard doing as part of the principal building of a farmyard? Any reconstruction incorporating one has a very extraordinary appearance.6 Thus far the problem has been discussed solely on the basis of published details and plans. A further possibility to be considered is that evidence for internal posts like those long known at Bilsdorf (Lux.) (Fig. 26) or discovered more recently at Ludwigsburg-Hoheneck (Fig. 26) has been overlooked or has long been removed. It is very hard to believe that this could be so at Bondorf, whatever may apply elsewhere, and consequently the case of that particular villa has to be decided in the light of the points made above.

OTHER VILLAS OF BONDORF TYPE All villas of this kind having a porticus-with-pavilions raise the question of why such a front is needed for a yard. All, too, present the problem of roofing a wide span. Thus Reim, in challenging orthodoxy,7 but without attempting to counter every point made in connection with Stahl, argued, from the absence of roof supports in an overall width of 16.6 m and the insufficiency of the walls to carry a tiled roof, that Inzigkofen was a yard, not a hall. Yet the span need not have been so wide, for the post bases on the west side are of a size more suited to heavy trusses than a lean-to roof; they leave a manageable clear span of 13.5 m, which is not excessive in the light of Blankenheim and Bruckneudorf (Aus.), 12 m and 14.5 m respectively, let alone Laufenburg (Bad.-Württ.) (Fig. 38) at 17 m. The importance to be attached to the slightness of the inner porticus wall and

Figure 26 Halls with freestanding corner posts


— Chapter Seven —

Figure 27 Halls with rear columns

the alleged inadequacy of its footings depends on the form of construction: a wall partly or wholly of timber would provide adequate support with only a slight foundation. Sigmaringen-Laiz A (Bad.-Württ.) (Fig. 25) is one of the few villas of this type (or


— Problematic House Types —

Figure 28 Halls subdivided by timber partitions

sub-type) to have been excavated and published to modern standards. An admirably objective analysis left the hall/yard question open8 but failed to recognise that the house had, above the two cellars (4, 5), a porticus between quasi-pavilions (2, 6) or terminal rooms9 which overlooked a valley in the way characteristic of so many continental villas. That of itself makes a hall more likely than a yard, even with a span of 17 m; moreover, a second porticus, defined internally by a timber structure, could reduce the span to 15 m.10 Other houses in Baden-Württemberg are as broad as Bondorf, two such being Messkirch (Fig. 25) and Tengen-Büsslingen, but the broadest is Remmingsheim (Fig. 25) which, from its size, has every appearance of a rectangular yard some 33×28 m. In front is the usual porticus-with-pavilions; one pavilion has a small room adjoining, the other two or three. Within, two larger structures (2, 6) face each other. The front rooms at least were domestic accommodation, but divided so clearly into two blocks as to have been in effect two apartments forming one house; and each of the internal structures adjoins one of these blocks. Here the yard is believable insofar as the two groups of rooms do not depend upon it but can be reached from the porticus; yet, since they are not linked in any coherent way and neither resembles any other kind of provincial Roman house, acceptance of a yard rather than a hall brings its own problems. In discussing Remmingsheim, Paret did not shrink from the implications of Stahl and Mayen; the formal front, he said, had been added to the dwelling house,11 thus implying a hall within which rooms were subsequently built; he did not suggest how so wide a space was roofed. Without special reference to Remmingsheim he recognised that posts not discovered by excavation would have been needed to support the wider roofs, and did not consider the alternative of three or four parallel rows, as in some big French barns. Nevertheless, he meets the point about the porticus-with-pavilions.

THE PROBLEM OF FUNCTION The matter of wide halls or small yards can be approached in two ways not dependent on detail. First is the point that several German villas where the whole walled farmyard has been examined have, more or less in the middle, a house of the kind under discussion. Messkirch, Bondorf and Tengen-Büsslingen are claimed to have a courtyard 99

— Chapter Seven — complete with cellar staircase and a hearth, like Mayen. But is it really credible that the most important building should be, essentially, a small courtyard partly surrounded by lean-to structures? As for its purpose, no suggestion has been advanced. Until those questions are answered the probability must remain that we are dealing with large open halls. A second approach relates to those houses where the central space is surrounded by rooms. Messkirch was like this,12 with rooms on two sides and two porticuses; there is no way of entering a small farmyard. Bad Rappenau (Bad.-Württ.) (Fig. 25) also has a completely enclosed middle space with no direct access from outside; so do Harburg-Gross Sorheim and GrossSachsenheim. Inzigkofen, Bondorf and others where access from outside is possible are to that extent credible as yards; where not, a use for the space is hard to envisage, and no trace of a garden has yet been found. The problem of function arises in connection with the domestic rooms in and around the yards as well as the yards themselves. Alpnach (Switz.) (Fig. 25) which has been variously regarded as a villa or a posting station,13 is wide for a roof of any kind – 19.5 m – but no less troubling is the idea of four rooms, not interconnecting and without porticus or veranda, opening on to a yard.14 The principal and largest room stands alone in a corner and at the other end are three rooms of varying size. Hechingen-Stein (Bad.-Württ.) (Fig. 27) is comparable in this respect. On balance Alpnach was a villa like Hechingen-Stein and each had a large open hall.

RANGES OF END ROOMS IN BROAD HALLS In southern Germany a number of the villas wide enough to prompt the question whether they are halls or yards have, at one end (and occasionally at both ends), a row of four rooms in which one is narrower than the others. Thus a range of four small rooms only about 4.5 m wide at Eckartsbrunn (Bay.) (Fig. 25) includes what looks like a lobby between two equal-sized rooms (7, 8) and between them and the porticus is a larger room (9); the whole resembles a row-house on a small scale.15 What must surely be a very large hall at Olfermont north building (Haut-Rhin) (Fig. 41) has a row of four rooms of the same width as those at Eckartsbrunn. No orthodox row-house is as narrow as these but an approximate parallel can be found in what is perhaps the smallest independent house so far found in Germany, Kirkel-Forst (Saarld), where the rooms are of the same average width. Better set out but with fewer rooms is Kempten-Loja 1 II (Bay.) (Fig. 29), and with the same number of even smaller rooms, Kempten-Loja 2 (Fig. 29). Bondorf is generally similar to Eckartsbrunn16 and confirms the impression that one end of the house was much more important than the other; the west or left-hand end, where the row of rooms is, also contained the hearth, the cellar entrance and the only two hypocausts; and adjoining the rear corner was the bath block. A Bavarian villa, Treuchtlingen-Weinbergshof, though double-ended, shows a clear emphasis on the east end compared with the simpler arrangements at the west end. Some of the passage-like rooms were definitely not lobbies. Stuttgart-Stammheim has just such a one (4) on the east side which illustrates the complexities of interpretation. Access to the hall is much restricted by the ramp to the cellar and a 100

— Problematic House Types — narrow ledge-like walk17 that can only have been a service approach to room 5 and perhaps the porticus, hence this small room is likely to be an annexe to, and entered from, one or both of the flanking rooms. It cannot possibly be a lobby, yet if not, how was the largest room of this block entered? Not, probably, from the front of the building because that would make the pavilion (6) a passage-room. The suggestion that the decayed wood found in the ramp may have been a floor or a roofing over of it18 allows for a wooden floor over the south part of the ramp to give access from either the hall or (by a small porch or lobby) the porticus. It appears in the reconstructed ground-floor plan. Not all lobby-like spaces are properly explained as such, as Wollersheim (Nordrh.Westf.) shows, for there the space in question is a staircase leading to the cellar under a corner room. Yet the existence of the cellar and of a hypocaust in the room at the opposite end make a different point about these end ranges, that despite their narrowness, they could be as well appointed as a small row-house. As for the overall plan of the range, it too bears a considerable resemblance to Kempten-Lojakapelle 2 I.

YARD RATHER THAN HALL? Finally, halls or yards without a porticus are still more problematic; of all the south German buildings so far discussed, these are unquestionably the ones most susceptible of a yard interpretation. A Bavarian building, Niedereschach-Fischbach 3 (Fig. 25), has, at the corner of a space 17.5 m wide, an L-shaped block of eight small rooms, the biggest only 4.25×3 m, and in the corner are three bath rooms, a furnace room and an oven. That leaves at each end two rooms equivalent to a unit at Brixworth; and it is hard to see how the whole functioned as a house without some considerable building adjacent or a veranda extended to cover the oven. The baths do not absolutely require an adjoining room of considerable size but, like those in the similarly divided row of small rooms in the aisled house IIIB at Collingham-Dalton Parlours (Fig. 68), make better sense with one. Wagner described Niedereschach 3 as a farmyard, while noting the large number of (roof) nails inside and outside; and he observed that ‘the roof must have sloped outwards’.19 A simpler example is Westheim-Hüssingen (Bay.), a structure about 19×17.5 m with an open hearth and two rooms inside it and three more adjoining them outside. The hearth is evidence that the walls enclosed a hall; otherwise, its placing towards the middle is hard to understand.20 It is rare in disputable buildings of this kind to find rooms built outside the hall or yard, but one villa at least is much harder to explain as a hall than as a yard. Reimlingen (Bay.) comprises three rooms of different sizes, two of which have hypocausts fired from a trapezial walled enclosure (to use a neutral term). Rooms could not be identified inside this space but wall plaster and a stretch of lime mortar floor were found. Reimlingen may therefore have resembled Hüssingen and the argument for that building’s being a yard would apply equally, were it not for the difficulty of explaining why such an eccentric plan was adopted and how it was roofed. Here, too, no veranda existed to link the rooms if they flanked a yard; indeed, it was precluded by the two furnaces which cut through the surrounding wall. 101

— Chapter Seven — All the villas analysed in this section have living-rooms no bigger than those of a British row-house, hence smaller than continental buildings generally, and some are remarkably small. Two factors may operate here: first, some of the activities carried on in a row-house were performed in the hall; and second, the households were smaller, with some separation of the modestly privileged from those who lived and worked in the yards and fields. In a historical context the wide yard/hall houses are interesting for their distribution, which on current showing is restricted to southern Germany and the adjoining Germanspeaking parts of Switzerland. Although that may under-estimate their connection with halls not so much smaller which have a wider distribution, the limited spread of this sub-type suggests some distinctive regional form of either agriculture or social organisation or a combination of both.

THE SMALLEST ROW-HOUSES? One of the most informative collections of villa plans ever published is Agache’s page of rural houses in northern France21 which illustrates, among other things, the many variations found in three-room houses. A striking aspect is the lack of true symmetry in all but two of the thirteen there depicted;22 six more convey an impression of symmetry;23 the rest are conspicuously asymmetrical. Taking the group of six, we find that all resemble one of the units at Downton (Fig. 10) in having a middle room or cell narrower than the others. The largest, Wancourt (Fig. 29), has an informative parallel at Plouneventer-Kerilien II (Fin.) (Fig. 29), which differs only in the detail of the porticus and shows how the flanking rooms relate to the transverse entrance lobby; their respective doorways were at opposite ends to protect privacy, the better room being entered further from the porticus. This kind of arrangement is intended to separate two sets of occupants, not unite them; they were two distinct households. It is reasonable to assume, in the absence of evidence to the contrary, that all six houses in the group were entered in the same way as, probably, was Hucclecote (Glos.), and all are thereby differentiated from houses where the middle room is as large as the other two or the largest. Bouchoir (Fig. 29), with veranda and pavilions, must have been entered in the middle of the front elevation so that the middle room which, by its proportions, was the minor element of the plan, is likely to have been held in common. In that respect it was like Lamargelle-Versingues without the transverse lobbies. It follows from what has just been said that houses formally of the same type might be very different in social composition. The two large rooms of Wancourt are each somewhat larger in area than the one-room house Sontheim, a.d. Brenz 2 III (Bad.-Württ.) (Fig. 30); those of Dury (Somme) (Fig. 29), than Sontheim II (Fig. 74). And a room at Wancourt, which is much the same size as one of the core rooms of Niedereschach-Fischbach 2 (Fig. 29), is almost four times as big as its counterparts at Dury. The conventional British view regards a house such as Dury as the residence of a conjugal family having a total living-space not much more that a quarter of that at Wancourt; even allowing for possible different densities of occupation, the latter must qualify as a row-house with two families, and probably extended families. 102

— Problematic House Types — H.O. Wagner drew another conclusion from Niedereschach-Fischbach 2, that it was ‘a representational dwelling, without an area of intimate family life’.24 No argument is offered in support of the idea, which may have resulted in part from the position of the building as the dominant middle one in a line of three. However it was reached, the conclusion depends, in the absence of evidence of non-domestic use, on assumptions about household size and composition, and where the doorways were; Plouneventer shows this and Vouneuil-sous-Biard (Fig. 20), for example, provides for a different relation between the parts of a kin-group. Asymmetrical houses with no obvious division into two parts are harder to interpret and make any idea of two families difficult to sustain. L’Etoile and

Figure 29 The smallest row houses


— Chapter Seven —

Figure 30 One-room houses


— Problematic House Types — Harbonnières (Somme) (Fig. 29), each with three progressively larger rooms, illustrate the problem: was the largest room used for some common purpose such as a workhall? And if so, can we, bearing in mind that complete separation of domestic life and work is unlikely, envisage its being occupied by part of an extended family? If so, Ovillers (Fig. 40), for example, might divide into two parts, one comprising an endlobby and living-room, the other a workhall/living-room.25 The questions raised cannot be answered until one or two such plans have been elucidated by excavation. Three-room row-houses are hardly to be found in Germany. One may be Rheinfelden-Herten/Warmbach 2 I (Bad.-Württ.) (Fig. 29) for, despite being described as an agricultural building, it is of a size and has the porticus and terminal room appropriate to a domestic structure. It is true that the north room has a wide doorway suitable for bringing in produce but the two other rooms have, respectively, a hearth and the foundation of an oven. Perhaps other three-room buildings also served a dual purpose, part domestic, part agricultural.

ONE-ROOM BUILDINGS: HOUSES OR WHAT? Little attention has been paid to the simplest villa buildings. Not surprisingly, most houses which are simply a hall with either no other rooms or only a porticus were enlarged subsequently. Noyers-sur-Serein I (Yonne) was no more than hall and porticus and was in a poor state of preservation following the building of a much larger establishment. A rare survival and in some ways a more certain one is Pin-Izel (Belg.) (Fig. 30), its title to be a villa, however modest, being established by its porticus and cellar. For the more numerous villas where [I] can be inferred to be a simple square hall, a suitable starting-point is provided by what may be the only instance established by excavation: Sontheim a.d. Brenz-Beim kleinen See (Bad.Württ.) is one of several unspectacular sites reported on by Eduard Neuffer with exemplary clarity. Two of the three buildings comprising the villa were halls, 1 apparently undivided, 3 with porticus-and-pavilions and of more than one building phase. The third, 2, had a complicated history (Fig. 30). 2/I was a plain structure only 4.6×4.2 m internally, producing no associated finds. It was replaced by 2/II, about one-third larger, in one corner of which was a hearth. That gave way to 2/III, an undivided hall nearly four times as large. Its successor, 2/IIIB, acquired two hypocausts, one of them no less than 14.5×3.8m26 and one of many such elongated rooms whose use is difficult to understand. By V 2 was the best appointed, though the smallest, of the three buildings comprising the farm. So, leaving aside the problematic 2/I, what was 2/II? The presence of a hearth rules out a temple; it must have been a dwelling, but for whom? Understandably, no explanation was offered. A one-room family house can certainly be envisaged in a provincial Roman context. Plouneventer and its analogues are pairs of one-room houses, and at LamargelleVersingues two apartments each comprised only one room and a lobby; and the two largest rooms at Alpnach, although each may have been linked by usage to the nearer pavilion, appear to have been essentially one-room dwellings. Even in more sophisticated houses like Newport an apartment might comprise no more than one large and one small room. The problem is not so much whether so small a house is 105

— Chapter Seven — credible but why it coexisted with two halls several times as large. This problem is bound up with another equally intractable one: why did the smallest of three buildings become the most comfortable? A few other sites have produced squarish buildings with stone footings or walls. Vierherrenborn-Irsch [I] (Rhld-Pf.), so far as it was discovered, was only a hall 10×9 m with hearths.27 A similar building appears to form the core of Koerich-Goeblingen 2/[II] (Lux.) (Fig. 30); although not mentioned in the report, its existence is established by straight joints and changes in wall thickness. The smallest building of this kind, a minor one in the villa of Biberist-Spitalhof (Switz.), is B IA (Fig. 30). It is described as a workshop for washing either freshly woven cloth or – in the country-house model – the laundry for the whole villa. Only in the nature of the work performed there can it have differed much from, say, Sontheim a.d. Brenz 2II. A comparable building, Gargrave-Kirk Sink E (Fig. 12),28 was twice replaced, and, though larger than Sontheim 2 II (c. 9×7.5 m internally), lacked any partitions. At one stage it is said to have been linked by a covered walk to the house D. As to its purpose, ‘the best interpretation is that [such buildings] served as the farm offices, where bailiff or owner would issue orders, receive tenants and their rents, and keep the estate records’ – just like a late nineteenth-century farm except for the disproportionately large size of the office and its unsuitable position at the innermost part of the enclosure. ‘Alternatively, extra accommodation for the owner’s sons and their families or the equivalent of dower houses may sometimes have been in question with these additions.’29 How suitable a single large room might be for either office or dower house is not considered, nor how appropriate its location for the latter; significantly, early modern English dower houses are at a considerable distance from the family seat. Some domestic use is hinted at by the covered walk. Such undivided buildings are the most rudimentary kind of hall and can usually be distinguished from shrines or temples either by their greater size relative to the villa or by a different kind of location. The type persists with the addition of a porticus or a narrow inner room but may be disguised by rebuilding. Thus Monreal [I] (Rhld-Pf.) (Fig. 30) is a nearly square hall entered by an unusually wide doorway (3.4 m) almost in the middle of the east wall; it had an inner room.30 Larger, and of one building phase, Neuhausen auf den Fildern-Horb (Bad.-Württ.) is a rudimentary hall on the way to becoming, with its porticus and cellar, a villa, and may perhaps be compared with Noyers-sur-Serein [I] or Ober-Ramstadt-Pfingstweide [I] (Hessen). The latter comprised a hall of much the same size as Neuhausen, 12×7m, with a hearth and a porticus.31

DOUBLE-DEPTH PLANS The history of villa houses is, for the most part, one of increasing complexity of plan. Usually this is expressed by extension in line or by wings and courtyards, or around an open hall; as a general rule, rooms are in single file. A small minority of houses has two parallel ranges of rooms side by side or separated by a corridor. They have never been subjected to serious analysis.


— Problematic House Types — Basse-Wavre (Brabant) (Fig. 31), mentioned frequently and confidently in the literature as a ‘corridor’ villa on a lavish scale comparable to Haccourt III or the younger Pliny’s Tuscan villa,32 is hard to understand for those who read a plan. Except for the great room in the middle the house is characterised by two rows of rooms – a block of square ones to the east, oblong ones interspersed with what are alleged to be courtyards elsewhere. The smallness of these rooms, only about 4 m wide – narrower than a British row-house, or the rooms at the ends of StuttgartStammheim – is as remarkable as the number of them. To what extent the two rows intercommunicated is unknown; perhaps hardly or not at all because, as the legend of the published plan says, most of the doorways are assumed. Most extraordinary are the nine or ten oblong rooms, every one of which was entered either at one end and thus necessarily a passage to the other end, or in the middle and so comprising two distinct parts. Yet Basse-Wavre shares this curious characteristic with just one other villa, that of Villers-Bretonneux (Somme) (Fig. 31), even to having a rear porticus considerably wider than the front one. Possibly the former served utilitarian purposes, as is suggested at Basse-Wavre by its being interrupted by inserted rooms and incorporating a hypocaust furnace. Some rooms of the Belgian house were certainly altered, as the excavator noticed, and a large wing room was added to east and baths to west; but the original form of the house will be discussed later in the light of other double-depth houses. Some are quite small. Frilford (Berks.) (Fig. 31) has the two rows of rooms, squarish and oblong mixed, but all extraordinarily narrow (2.75 m); and there is a larger room with hypocaust. The most inexplicable feature is the tapering of the front rooms; it looks like a clumsy way of emphasising the dominant room 14, to the left of the point of entrance. Smaller still, and uncertain in detail, is GünzenheimStaatsforst-Sulz (Fig. 31).33 Nevertheless, it has parallel narrow rooms, perhaps two rows of three,34 and the marked tapering found at Frilford, although here to opposite effect; the larger pavilion adjoins the narrow end. These two houses give some plausibility to an improbable-looking tapering squarish building at Chiddingfold (Surrey) which has nine small rooms. A few quite small double-depth houses are known. They include KernenRommelshausen (Bad.-Württ.) (Fig. 31), which had four rooms, a workhall and a cellar; and it is of some interest that this very minor villa has at the rear a room-andworkhall unit not dissimilar to those at Basse-Wavre. A Hungarian villa, SmarjeGrobelce 1, has, essentially, six square rooms, one with a mosaic and another with good wall-paintings; and even the tiny Somerton-Catsgore 3.14 belongs in this category. The first two especially present some of the same problems as the much larger houses. Before I discuss these very curious plans it will be worth considering one or two which are formally indistinguishable from them but are more susceptible of explanation.


— Chapter Seven —

Figure 31


— Problematic House Types —

Figure 31 Adjoining parallel ranges and back-to-back halls

BACK-TO-BACK HOUSES One of the most instructive double-depth houses, Munzenberg-Gambach (Hessen) (Fig. 31), appears, on analysis, to be a highly developed row-house. In I it had two ranges of rooms of equal width, each having in the west half a lobby flanked by two rooms; they look like the two units of Downton (Fig. 10) placed back to back. One room (9) had an open hearth. In the north range a larger room with a wide doorway at the east end is a workhall forming, with a lobby and a square room to the west, a separate unit. On the south side the remaining space is occupied by two unequal rooms, one of which may also have had some work function. That was the superior side, with the largest pavilion, which had a hypocaust of some kind, and a cellar. In effect each of the two ranges is intelligible as a row-house of two units and would be so regarded if they had been built separately, like two of the houses at Gargrave. A comparable plan was excavated in the early nineteenth century at Lendin (SeineMar.) (Fig. 31). There the double-depth block comprises principally two four-cell ranges, the one facing the courtyard looking much like Northchurch IIA, the other like Welwyn-Lockleys I (Fig. 11). Some rooms had painted plaster and at least one had a hypocaust. At one end are three small rooms of unknown purpose; a bath suite, possibly. Where, though, are the larger rooms corresponding to the one or two workhalls of Gambach? They may be on the opposite side of the courtyard, in a building which comprises a square room at the inner end and a long one divided by an oblique wall into two not quite equal parts. It is a strange way to divide a building yet it occurs in early modern England. At Ashwell (Herts.), Ducklake Farmhouse is divided into two parts, each of which is a house, by a skewed timber-framed partition standing at about 80 degrees to


— Chapter Seven — the long walls: a manifestation of the unit system. Why the partition was set out obliquely is unclear; it must have had a significance now lost.35 By analogy the oblique wall at Lendin is taken as a sign of joint proprietorship, however obscure the reasons for making the division that way. Another such oblique division appears in the villa of Anthée (Fig. 75), and in so prominent a position in so grand a house cannot be the result of careless setting-out; it may be secondary, but I cannot properly explain it. A more recent example, and the closest to the Ashwell house in its domestic purpose, is the oblique wall dividing the hall at Saint-Aubin-sur-Mer (Fig. 9).

THE INTERPRETATION OF DOUBLE-DEPTH PLANS Since this kind of plan brings more problems of lighting and ventilation than single ranges it must have had some important advantages over the commoner row-type plans. Gambach is the easiest to understand; it is two row-houses, each comprising two unequal units, set back to back rather than end to end like Romegoux or Aylesford-Eccles I (Figs. 13, 66), and the two face in opposite directions like Marshfield (Fig. 66) or the slightly separate houses at Collingham-Dalton Parlours (Fig. 68). Approaches from different directions were important in this kind of grouping, as can be seen where there are two architecturally unrelated houses, for example KoerichGoeblingen (Fig. 74). Perhaps the principal advantage of a unified block of building was its more imposing appearance at a time when villas generally were growing in size, and this is very obvious at Gambach, especially after the addition of the pavilions. A single-span roof over the whole would have produced a massive bulk comparable to that of the biggest halls but without the technical problems of a wide clear span, so providing an impressive appearance relatively inexpensively. Another benefit must have been a reduction in building costs; what a seventeenth-century English architectural writer called ‘spare of walling’36 mattered, and use of a spine wall saved the cost of another of equal length. This analysis may help to explain Villers-Bretonneux and Basse-Wavre. At VillersBretonneux all the rooms are either small and square, appropriate to a living-room, or long and narrow, like a workroom. Each row can be divided into three livingroom and workhall units, the best, with the largest living-room and workroom, being to the south, and two others of equal size; but the latter have an extra livingroom between them. And those three are doubled, making six units in all. Selfevidently the narrowness of the rooms was fundamental to the design; given that, a one-room-deep house would either have been preposterously elongated or necessarily grouped around a courtyard, and the latter would perhaps have produced problems of relative status among so many equal units. A double-depth house resolved these difficulties with great economy of walling. This helps to explain Basse-Wavre. Starting at the east end, the distinctive block of six square rooms flanked by pairs of workrooms matches Villers-Bretonneux exactly: four units and two uncertain living-rooms.37 How the remainder was divided up in [I] is matter of argument but an attempt has been made (Fig. 31) on the assumption that its walls are likely to have been incorporated in [II] wherever possible.38 Even if these guesses are wrong in detail, the shape of the rooms demonstrates a general 110

— Problematic House Types — likeness to Villers-Bretonneux and so makes a unit-system interpretation with several apartments likely. Frilford finds a place here. The long room 13 stands out in the same way as the biggest room at Munzenberg-Gambach and may have had the same purpose, that of a workhall. The smaller rooms resemble the many such at Basse-Wavre and VillersBretonneux, though differently proportioned; 10+11, 4+5, 7+8, 3+2 all combine a small room with a slightly larger one; 6 and 12 are like those extra rooms not easily allocated in the continental villas. Whatever argument there may be over details, Frilford’s points of resemblance to Gambach may denote a division on the same general lines. The most puzzling aspect of Basse-Wavre and its analogues is why the rooms should be so narrow compared with those of Gambach or nearly all row-houses, yet the consistency in this respect shows it must have been fundamental to this form of plan. The problem is analogous to that presented by the end rooms in German halls, which can resemble closely a three- or four-room row-house but are invariably narrower, and by the pairs of two-room units in some row-houses, for example those at the south end of Towcester-Mileoak (Northants.) which could come from Villers-Bretonneux. Differences of this nature may arise from differences in household size or the purposes the rooms served; did each household at Basse-Wavre, for example, conduct in the oblong room some work operations which elsewhere were carried on communally or severally in a workhall or hall? Excavation is the only hope of solving this problem. Something must be said about three houses with partial or quasi-double-depth plans. Frankfurt-Bornheim (formerly Gunthersburg Park) (Fig. 32) and Ashtead (Surrey) (Fig. 32) are linked tenuously to Villers-Bretonneux by the unusual characteristic of having the narrower rooms at the front; also unusual in houses of this size is the absence of pavilions or terminal rooms. In other respects they depart entirely from the pattern in having rooms or groups of rooms which ignore the spine wall, Frankfurt at the north, Ashtead, perhaps by alteration, at the north-east end (rooms 1, 2, 2a). Ashtead also has what is interpreted as a reception room (9) serving as an anteroom to a dining-room (8);39 their unusual proportions relative to one another are not explained. A related Belgian villa, Ambresin (Fig. 31, scale uncertain), with two rows of rooms of markedly different width, is like Frankfurt and Ashtead in having one larger room 12 which cuts across the spine wall; the recurrence of this

Figure 32 Anomolous double-depth houses


— Chapter Seven — feature may represent a stage in the development of the type towards the dominant middle room found at Basse-Wavre. Frankfurt-Bornheim becomes intelligible if entered by the rear porticus (10) and if the middle room (6) was a passage-room to 8 and 7; and it certainly led to the bath suite. That leaves a block of four rooms (7, 9, 12, 13) as a domestic suite, and the largest room (3) as a place for leisure and recreation closely connected with bathing. The accommodation provided by Frankfurt is equivalent to two blocks of rooms each comparable to the one at the end of a hall such as Kinheim (Fig. 24), but without the hall itself, and it leaves the porticus as a place of recreation. This means that any resemblance to Ashtead is purely formal; consequently the Surrey villa remains, for the present, unique.

THE AXIAL CORRIDOR Allied to the double-depth type is one characterised by an axial corridor with rooms on both sides. Two such have been excavated, neither completely. Hohenfels (RhldPf.) (Fig. 33) has a porticus, probably with the usual two pavilions, behind which lies the essential house. Next to the porticus and heated by a hypocaust was an elongated room (6) comparable in its proportions to a room at Sontheim a.d. Brenz 2 IV/V. The building seems to have been entered, like so many villas, on the opposite side to the display front; a wide hall (14) led to the internal corridor and also a short corridor straight ahead to the only entrance to the principal room. This was a wide doorway facing an equally wide one on the other side: an opposition implying that the two shared some important common function, rather like the series of rooms linked by aligned doorways in eighteenth-century country houses. Yet such an approach to a grand room is itself strange; wide doorways and a narrow corridor are a contradiction in architectural meaning. A curious feature of the opposed doorways is that they are set slightly off axis to the rooms, as if the important part, the place of privilege, were on the porticus side, and this, too, is unusual. Neuburg a.d. Donau (Bay.) (Fig. 33) has a few points in common with Hohenfels. Though lacking a porticus, it looks eastwards down a slope and was approached at the rear through an entrance-hall (6). At the far end was an extraordinarily long and narrow room with an apse at one end; it looks like a porticus but is in quite the wrong place, and better analogies are the long rooms at Sontheim and Hohenfels. The next largest room (3) may have been the principal focus of daily life; how the rest related to it is uncertain, but the narrow room 5 presumably had the same function as its counterpart at Hohenfels. The only other certain example of an axial corridor appears to be Mézières-en-Santerre/la Croix Saint-Jacques (Somme), which is known from an aerial photograph and is bigger than the others of this type. It looks like two row-houses, each with two or three transverse lobbies, and was apparently entered at the front.40 Related to these axial corridor plans are villas which are really two houses sharing a common corridor. A large instance of this rare kind, Bad Dürkheim-Ungstein (Bad.Württ.) (Fig. 33), has as its principal building a highly developed hall house with two wings.41 Behind it and separated from it by a corridor is a tripartite complex explained 112

— Problematic House Types — as a courtyard with a wing at the east end and a large room at the other. The wing is said to be a cart shed on the grounds that it has a wide double doorway, the room to be an industrial or craft area. So what was the purpose of the yard between them? In view of the uses to which the ends were put, a hall might be a better interpretation, and its size, 21×11 m, is entirely appropriate. The problem this creates of lighting a corridor 35 m long and only 2.5 m wide may appear greater in the twentieth century

Figure 33 Houses with internal corridor or yard?


— Chapter Seven — than it did in the fourth because the amount of light needed was not very great; borrowed light from the hall and the two largest rooms of the house, and light coming in through the doorway at the west end – normally open in daytime – would be enough. The only other such villa is Geislingen-Heidegger Hof (Bad.-Württ.) (Fig. 33), dug in 1795, written up in manuscript in 1816, published in 1911,42 and for which not much more than a plan is available. It is really two separate buildings rather than ranges of rooms, flanking a space too wide (9 m) to be called a corridor but which, nevertheless, performs more of that function than the term ‘narrow courtyard’ might imply.43 Detail which might decide whether it was roofed or not is lacking. The west-facing front (?) range of the villa comprised in [I] a narrow hall, end rooms and porticus and was about the same size as Collingham-Dalton Parlours J [I] (Fig. 68);44 pavilions were either altered or added later and the house was enlarged subsequently. Behind it is the yard or corridor and behind that a double-depth range of rooms, the best of which faces east, although it is not at all clear that it was given a formal architectural treatment on that side. For all the deficiencies of our information, the middle yard does appear to have functioned principally as a corridor; perhaps it was made that wide, and left open, to avoid problems of lighting. It is a kind of plan to be differentiated from that of a villa at Bregenz (Aus.), where both ranges face inwards to a courtyard of about the same width as Geislingen.

THE SOCIAL BASIS OF AXIAL-CORRIDOR PLANS Axial corridors occur in two very different social contexts. Geislingen and Bad Dürkheim are the easier to explain; they represent the bringing together of the separate buildings usual in villa planning, as courtyard, peristyle or big elongated houses like Maillen-Ronchinne or Jemelle-Neufchâteau do in other ways. It is a development apparent to varying degrees in most of the European provinces. Yet, however unlike Bad Dürkheim and, for example, Sudeley-Spoonley Wood (Fig. 70) look, they embody comparable elements: a principal domestic and representational range; another domestic range, east at Bad Dürkheim, south at Spoonley Wood; and baths, at the west end of the main range and the south range of the German and British villas respectively. Lastly, there is a large tripartite open hall, separated by a corridor from the house at Bad Dürkheim, by an interruption in the porticus between house and north wing at Spoonley Wood: different ways of expressing social inequality between groups of people who lived in closer proximity than would be appropriate to a wage-labour relationship. Beyond the residential court of Bad Dürkheim the part-domestic outbuilding 4 corresponds to the aisled building, probably part domestic, at Spoonley Wood (Fig. 44). Such comparisons could be made between many villas. Not all the components just considered appear at Geislingen, where excavation is known to have been incomplete. Although there is a resemblance to Bad Dürkheim insofar as the main house is separated by the access court from an inferior block comprising principally a hall and end rooms, the presence of two large hypocausts in the former and a small one in the latter shows that the social gap between them 114

— Problematic House Types — was less than the physical one might suggest – another point linking these unusual kinds of villa to commoner types. There were many ways of achieving rather similar social ends. Hohenfels and Neuburg a.d. Donau are so much smaller as to make any explanation of this kind unlikely. So what was the point of putting a corridor between two parts of not very large houses? Neither is divided in a way suggestive of social difference. Hohenfels is the superior house; it has the porticus-withpavilions lacking at Neuburg. Both have a small bath suite, three squarish rooms and a narrow room by the entrance; the difference is narrowed down to the large hypocausted room 6 and the associated room 8, which give the appearance of a large, comfortable and rather grand suite. Perhaps the intention of the axial corridor was simply to create a more centralised plan than could be achieved with rowtype derivatives, and at Hohenfels the best rooms were next to a porticus which was clearly intended for show and recreation, not an entrance. If this analysis of an idiosyncratic plan is accepted, the only parallels are likely to be equivalents in number and kind of rooms – socially rather than architecturally analogous. Neuburg is a humbler attempt to the same end.

BACK-TO-BACK HALLS Row-type houses back to back are invariably the principal residence of a villa. Backto-back halls are rarer, seem always to be subsidiary buildings so sited as to demonstrate socially inferior status, and are not always of a single building phase. Graux is one of two instructive Belgian examples, with the halls (which must be workhalls) standing to the south of the house. First came the south hall, with a south-facing porticus, an undivided room at one end and three rooms at the other; it was the superior of the two. Adjacent to it on the north, and only just touching it, was added a hall with north-facing porticus and later a long end room, so that in two stages the first hall was virtually duplicated. Just how deliberately the new one was sited appears from its awkward relation to the older building; it looks as though the builders first set out the front wall and porticus to face in a very precise direction, not simply parallel to the existing hall, and then found difficulty in fitting in the east gable wall. Walsbetz (Fig. 31)45 presents the same idea. Two slightly different workhalls are sited parallel and a short distance apart and to demonstrate their independence they face in opposite directions; they are the Romanised version of two similarly sited halls at Rijswijk (Fig. 65). A similar explanation may underlie the siting of two buildings (3, 4) at the Hungarian villa of Szentkiralyszabadja-Romkut; also the bipartite hall and the building parallel to it lying south-west of the Dwelling House at Langton (Yorks.). These halls reflect stages in the growth of social inequality, the main house being occupied by a superior group under whom were two subordinate groups inhabiting virtually identical houses, whose not quite equal standing was expressed by their facing in opposite directions and thus in differing relation to the main house. These two buildings were not occupied by people who were at their superiors’ beck and call – mere labourers – otherwise the relations between them would not have 115

— Chapter Seven — influenced the positioning of their houses. It is a quite different situation from that at Sudeley-Spoonley Wood, where only one sizeable hall stands within the courtyard and the lowest members of the villa community lived in the buildings where they worked, or that at Köngen, where two halls each stand in a slightly different physical and social relation to a third principal one. Graux and other villas were built according to a code expressive of social relations, a code that was taken for granted, not put in writing, and was constantly changing in its local applications to meet specific situations. A few basic principles could have been stated had anyone thought fit to do so, but the finer points evidently depended on subtly different power relations that were capable of infinite variation. To take one example from those mentioned above, a scale of development can be observed running all the way from the tiny Somerton-Catsgore building through VillersBretonneux to Basse-Wavre, and some factor made it desirable to build the basic units side by side, even in the smallest house of all. That factor, whatever it was, did not find the same rigid expression in any other kind of house, not even in MunzenbergGambach. And somehow, right to the end, the large Belgian villa preserved elements of social relations, however modified, that had existed in the little building in Somerset; total rebuilding never became necessary. Archaeologists are driven by the nature of their work to concentrate on change, and the timing and causes of change are very much the historian’s business too, but it is important to recognise the underlying stability expressed by the continuity of plan elements in houses and villas.





n the preceding chapters two principal types of house have been isolated, each corresponding to the fundamental social organisation of the family group that built it. To very many houses was added the kind of front defined by Swoboda as the Portikus-mit-Eckrisaliten, the porticus-with-pavilions, that was the badge of Romanisation in most of the European provinces. It must also, by providing two rooms of considerable size, have modified the conduct of daily life in the core rooms, but unless the changes were sufficient to require structural alteration they will be imperceptible from architectural remains. As Koepp remarked in a critique of Swoboda’s book, the fact cannot be underestimated that, in architecture, purpose is the first consideration and form the second,1 so it is imperative to discover the uses to which these substantial additions were put. Although it is impossible to do justice to the complexities of this kind of change on the evidence at present available, a beginning can be made by studying the relation between pavilions and the main body of the house – how they are approached, their size relative to other rooms, and the differences they presented in appearance. This will permit a distinction to be drawn between pavilions and wings in the true architectural sense, for it is as desirable to wrest the word ‘wing’ from the British archaeologists who have so long misused it as it is to prevent Germans from applying Eckrisalit indiscriminately to every flanking structure.

PAVILIONS: THE CLASSIC FORM A threefold intention underlay the porticus-with-pavilions: aesthetic, to create a balanced facade in which the pavilions provide a strong visual termination to the lighter and more open appearance of the colonnaded porticus; cultural, to proclaim to all who came near the adoption of a Romanised way of life; and social, to provide additional high-status rooms. To rank these intentions in order of importance would be pointless because they are inseparable, but it is important to bear in mind that the extra rooms thus provided were not, at least in the initial stages of Romanisation, so closely integrated with the core building as to require intercommunication. This is self-evident for some villas and can be 117

— Chapter Eight — demonstrated for more, and although in the majority of houses it is uncertain, there are not many clear instances of pavilions being linked with core rooms. The classic form of pavilion appears most clearly in the simplest hall houses, those where the central mass of the hall touches the pavilions corner to corner and rises above them, while the porticus joining them provides a more or less open screen; how far open is matter of debate. Great Staughton (Cambs.) (Fig. 40), Mayen IV (Fig. 1) and Serville (Fig. 1) are like this. In a few instances the pavilions stand a little way beyond the dominant block and the whole is unified by the colonnaded porticus; Stahl comes into this category and Voerendaal-Ubachsberg (Neth.) just scrapes in. There are besides a few houses where an intention to give hall and pavilions this elongated symmetry is evident but is clouded by coeval structures which made a minor visual impact and presumably were thought not to detract from the overall effect; Bondorf (Bad.-Württ.) (Fig. 25), where a range of four rooms adjoins the hall to west, is one such.2 Sideways projection of the pavilions is not very common and implies a situation in which separation from the main body of the house did not detract from the room’s usefulness. Serville illustrates the problems. It has a hall with no adjoining rooms, a colonnaded porticus and two pavilions; that to south, the only room in the villa with painted plaster,3 stood over a cellar; that to north, which de Maeyer thought was perhaps a bedroom,4 was slightly the larger. For the south pavilion historical analogy may provide a clue. Not a few early modern farmhouses in England and Wales have a parlour, the most private and usually the best-appointed room, under which is a cellar for the storage of beer or cider, which in a Roman context translate into wine or family treasures. The point of this conjunction is to give the head of the house control of the cellar, and the same motive may reasonably be assumed at Serville and villas similar in this respect. So how did the house function, bearing in mind the limitation that no subsidiary buildings were discovered? The hall is uncomfortably large for a nuclear family; at the very least either relatives or servants, whether indoor or outdoor, have to be allowed for; and if servants, it presupposes a more equal relation between master and man than seems likely in the light of other evidence about society in the Roman Empire. The south pavilion, though the smaller of the two, was the more important and was presumably given over to the head of the household; but in what capacity? He could hardly exercise any representational function there. An office is sometimes suggested in villas, although that implies a level of organisation and a separation of functions hardly to be expected in so simple an establishment. A reception room or dining-room are possibilities, yet the better pavilion lacks the prominent position close to the entrance which is usual where such purposes are assumed. This is not a fatal objection because a house with a hall as dominant as that at Serville has no other obvious place for a new function. As for the north pavilion, de Maeyer’s suggestion of a bedroom is anachronistic if a special-purpose room is meant. If the foregoing possibilities are set aside, the only probable function left for the pavilions is living and sleeping accommodation. For whom? If we assume ownership by a conjugal family, they could be for master and mistress (each with a personal servant), south and north respectively, perhaps, although the space provided seems anachronistically large in a society just emerging from a common life in a large hall, 118

— The Porticus-with-Pavilions: Pavilions — and there seems to be little evidence of this kind of separation. Alternatively, it would be consistent with the dual occupation suggested for some halls if each pavilion were occupied by a senior household, the south for the head of the kingroup for the time being, the north perhaps for the next most senior, leaving the hall to kinsfolk who were dependants of various degrees. These suggestions, though totally speculative and quite inconclusive, show that the functional possibilities of pavilions in any particular house are not unlimited and can be narrowed down.

WHERE WAS THE PAVILION ENTERED? In trying to understand the various uses of pavilions and the reasons why enlargement of the house takes that form the first problem is to ascertain their relation to the core rooms. Pavilions which are somewhat detached make a useful starting-point. Most pavilions overlap the main body of the house by about half their width, some by less, others by more, a few hardly at all. Nor do all project to their full depth; some are partly or even largely recessed into the hall behind. The extent of the overlap determines whether it is possible to fit in a doorway from the main block to the pavilion, a point that has to be taken into account even though surviving doorways and other relevant evidence suggest that entrance to a pavilion is usually from the porticus. Mundelsheim (Bad.-Württ.), Farmington (Glos.) (Figs 22, 3) and Aiseau S Bldg are hall houses where the pavilions overlap for about half their width. At Farmington, where there is overlap enough for an entrance to the west pavilion from the room at the upper end of the hall, a doorway in that position is virtually precluded by a fireplace, like the north-west pavilion at Neumagen-Dhron-Papiermühle 1 (Rhld-Pf.) (Fig. 68); and access to the east pavilion at Farmington was restricted, though perhaps not entirely prohibited, by a fireplace in the hall near it. Neckarzimmern-Stockbronner Hof (Bad.-Württ.) (Fig. 27) further emphasises the point. There the overlap is insufficient to permit a doorway of normal width from the hall to the south pavilion; intercommunication at that point would have required a bridge over the staircase,5 and how awkward one would have been is shown by the cramped entrance from the hall to the cellar beneath the pavilion. That this is not an isolated instance of clumsy planning is shown by the rather similar cellar entrance at Furschweiler (Saarld) above which, certainly, no entrance to the pavilion existed. Many villas have the sophisticated form of porticus which returns forward to provide a small square space just in front of the pavilion; Gayton Thorpe N (Norfolk) (Fig. 39) is an example. It is a comparatively late development with the intention, probably, of shielding the entrance to the pavilion from the gaze of those excluded from it; and Weitersbach I (Fig. 69) had a simpler way of achieving the same end. All villas of this kind reinforce the idea that the customary point of entry to such pavilions was in the porticus. Not all pavilions were so entered. Lauffen am Neckar I (Bad.-Württ.) (Fig. 41) is a hall with porticus-and-pavilions where the evidence is unusually clear: the south pavilion has a doorway from the hall and enough remains of the wall towards the porticus to show that a doorway cannot have existed there in that phase. In II a hypocaust was built in the pavilion, its stokehole blocking the entrance from the 119

— Chapter Eight — hall; at that stage a doorway must have been made into the porticus, there being no other place for it.6 No trace of a doorway was discovered in the north pavilion, but there, too, a hypocaust introduced in II blocked any doorway from the hall. The three-sided apse added to create a dining-room7 demands an entrance from the porticus, so the same change is possible in this pavilion. Although we cannot be sure of that, it is quite likely that structural change in the pavilions brought about another change, from intercommunicating with the hall to a situation of greater social distance.

ASYMMETRICAL PAVILIONS The classic form of pavilions-with-porticus is symmetrical, yet in fact the symmetry is rarely perfect. Even at Serville the pavilions are not of quite equal size, the betterfinished of the two being marginally the smaller, and although the disparity of size is rarely as great as at Ludwigsburg-Pflugfelden (Fig. 1), pavilions can be asymmetrical in less obvious ways. A common disposition is to have one flush with the end of the house, or nearly so, and the other either partly or sometimes almost wholly projecting beyond the opposite end. The purpose of any arrangement of this kind is clear enough in general terms: the pavilion that is the more clearly distinguished from the main body of the house has, implicitly, some function that ties it less closely to the core building than its fellow. Even so, the differences between them can be slight and confusing, as at another hall house with porticus-and-pavilions, Biberach [I] (Bad.-Württ.): the smaller south pavilion stands over a cellar, is flush with the end of the hall, and has painted walls, whereas the larger one to the south projects insignificantly beyond the gable-end and has a hypocaust. The one to the north is the better finished, the more prominent and comfortable and for that reason, perhaps, the more important. A possibility that springs to mind in a villa as simple as this is that the heated pavilion was a dining-room; yet against that there is not even a single service room such as might be expected, and would surely be needed, in an establishment sophisticated enough to have a specialised dining-room. We simply do not know the code governing the size, functions and placing of pavilions and it will take much work to recover it.

PRACTICAL ASYMMETRY? THE CASE OF ROTHSELBERG Rothselberg (Rhld-Pf.) (Fig. 6) suggests how one of the pavilions, which projects too far to be reached from the hall, might have been used and why it was so sited. Since the east end of the hall was given over to agricultural or other work purposes, the pavilion could have been used to oversee the wide entrance, where traces of waggons were observed, and to supervise the yard. To suggest that this pavilion was in part an office may not entirely preclude some domestic function, for its plastered walls were painted red, like those of the porticus, these being the only places where such decoration was found. The layout of the hall throws light on the pavilions. The circulation space between two opposite doorways divides the hall into a farm-related working part and a 120

— The Porticus-with-Pavilions: Pavilions — smaller domestic part, and to the latter are attached two rooms, a large one with a wide doorway and cement floor and a smaller one above a cellar. Although the precise function of the west pavilion is uncertain, its position at the superior end of the house underlines the physical isolation of the east porticus from the purely domestic rooms.8

ASYMMETRY AS AN EXPRESSION OF STATUS While it cannot be doubted that pavilions express both the status of a house and of those occupying them, it is rarely obvious which was the superior one, even where differentiation was more pronounced than at Serville. Osterfingen I (Switz.) (Fig. 68) was a hall with pavilions of equal projection. When it was enlarged in II by half its length a new pavilion, larger in floor area but the same width as the one it replaced, was built abutting the gable wall. Entered only from the porticus, it was distinguished by three pilasters as being in some sense the most important room in the house; nevertheless, its fellow was provided with a hypocaust and a small inner room with painted wall plaster and was certainly a more comfortable apartment.9 It looks as if the pilastered pavilion was important for those approaching the house, perhaps as a place to meet for the negotiation of business or other matters, whereas the inner room to the south was important for the household as the seat of internal authority. Other houses resemble Osterfingen in having one of two pavilions embellished with pilasters whose significance is complicated by their being regarded usually as buttresses, an interpretation the more plausible because so many continental villas stand at the edge of a river valley. Thus at Grenchen (Switz.) (Fig. 22) the most prominent pavilion (of three) is treated with pilasters on two sides, as though the villa was approached from the south-east and the most conspicuous feature of the house was an imposing pavilion for, if the projecting features really were buttresses, it is hard to see why more were not needed. It is interesting to note, in the light of the conflicting views about a bath suite at Osterfingen, that the Grenchen pavilion may have been a bath at some stage.10 Moreover, a blocked doorway discovered ‘in the south-east front’, at a place not precisely specified, makes the function of this pavilion yet more dubious. These uncertainties matter less than the evident intention to dispense with true symmetry in order to proclaim the greater importance and, in this case perhaps, the greater height, of one pavilion. Symmetry was not valued for its own sake; some imperfection was important to establish to all and sundry the social relations between the various parts of the house. One of the few British hall houses with a strong resemblance to those in Germany deserves mention here. Byfield (Northants.) (Fig. 3) appears to have been a simple undivided hall. At first glance the pavilions look equal, yet they are not identical; that to the north is fractionally larger than that to the south, and projects very slightly more beyond the gable-end. The house model from the Titelberg (Fig. 34) gives some idea of what it must have looked like. So little difference is there between the two pavilions that it could be dismissed as a minor discrepancy of setting-out, yet just such small variations occur in many villas,11 and when combined with the customary accuracy of provincial Roman builders in setting out right angles are 121

— Chapter Eight — likely to be deliberate and significant of some social differentiation. In this respect, too, the Titelberg model is helpful; it shows how the size of the pavilions and the height of their gables would have stood out against the length and especially the dominant roof of the hall in a way which emphasised their respective sizes.

Figure 34 Pavilion details


— The Porticus-with-Pavilions: Pavilions —

SMALL DIFFERENCE, SIGNIFICANT SOCIAL IMPLICATION Two comparisons from early modern England strengthen the argument. Aston Bury, a mid-seventeenth-century manor house in Hertfordshire of basically rectangular plan and one room in depth, is unusual in having two virtually identical staircase wings at the rear.12 The staircases themselves have characteristic ornament of the period and are virtually identical. Only on close inspection does a slight difference in ornament and painting become apparent, and only measurement shows that the marginally simpler staircase stands in a marginally smaller wing. For present purposes the measurements are more important; both the width and the projection of one staircase are slightly smaller than those of the other; they are 4.42 m and 4.27 m wide, and project 3.20 and 2.90 m, respectively. These differences are of the same order of magnitude as are observable between the two pavilions of many a Roman villa, and it is likely that in an age when people necessarily had to depend on close observation, because they had no written aids, two such structures would be instantly recognisable as conveying a social message, or rather, reinforcing it, for their size would not be the only signal; ornament and decoration played a part too. Walnut Tree Farm in Pirton parish13 teaches the same lesson. It has two porches equidistantly spaced which exhibit the same order of difference as the Aston Bury staircases. They were intended for two households, one superior to the other; one comprised only living-rooms on the ground floor, the other had a kitchen and common hall/dining-room. If the house had been razed and subsequently excavated, the plan, except for the chimney stacks, would have points of resemblance to a Roman villa. Anyone approaching the house and confronted with two porches would have known instantly which one was appropriate to his status and purpose, quite apart from incidental signs of occupation. Those who doubt that small differences between pavilions and matching rooms are significant have to argue away the relevance of these analogies.

OBLONG PAVILIONS Most pavilions are squarish. Some, as at Stahl (Fig. 1), are oblong and project boldly in front of the house. A few villas have an oblong room apparently of equivalent function which stands parallel to the house and is roofed parallel to it: Raversbeuren (Rhld-Pf.) (the north-east pavilion) and Badgeworth (Glos.) are examples. The two kinds of pavilion conveyed different social messages; the latter group is anomalous and its significance obscure, the former is the more instructive. As a general rule a squarish shape proved convenient for most kinds of livingroom and consequently elongated pavilions demand explanation. Stahl (Fig. 34) points to one. The left or west pavilion has splayed corners intended to provide a setting for someone or something of importance, probably the former; it looks like a variant of those rooms reduced in width at one end which are found at Bierbach (Saarld) (Fig. 19) and some other villas. Functionally the room divides into a place of honour emphasised by the splays, where, perhaps, the head of the houseful sat; above him a window lit the body of 123

— Chapter Eight — the room, leaving him in comparative obscurity. He faced an area more or less square where the main activity was carried on, and a circulation and service space at the doorway end. Thus divided, a dining-room would provide a suitable use, or it might have had a representational purpose: in the first case the service space might have been reserved for a sideboard or its equivalent as a waiting place for servants; in the second for a stand displaying precious objects such as a portrait bust or figurine of a deity. Broadly similar considerations apply to the use of the floor space in the east pavilion. From their size, shape and position these are clearly the most important rooms in the house and, just as happened with the parlour in late medieval hall houses, they must have taken over some of the principal functions of the hall, including representational ones; for the upper end, in the sense of a seat of honour where the head of the house habitually sat, seems to have disappeared by the time the pavilions were added. A British hall house, Kingsweston [F] (Fig. 8), throws some light on how this not very common kind of pavilion might be used. The west pavilion (7) (Fig. 34) is an inner room in relation to the adjoining room (6) at the end of the hall, the two forming a functional whole. Presumably persons coming past the hall hearth entered 6 at the north end, where the plain border of the mosaic is widest, and, if privileged to do so, continued on to 7 where, at the far end on the wider part of the border, was the place of honour reserved for the head of the household. Uses for the entire space have to be envisaged on the lines of those suggested for Stahl; but the east pavilion (11), entered from the porticus, had, insofar as it was more accessible, a less private and thus inferior status. Oblong pavilions are uncommon and generally smaller than these. Examples are to be found at Ersigen (Switz.), where they are as prominent as those at Stahl; fronting a divided hall at Saint-Aubin-sur-Mer (Pas-de-Calais) (Fig. 9); and at the row-house of Laufen-Müschag (Switz.); none appears to be susceptible of the kind of conjecture ventured for Stahl and Kingsweston. No general explanation has been offered for the quite common addition to pavilions of a slightly smaller flanking room, of which Bondorf (Fig. 25), Biberach (Bad.-Württ.) and the north side of Manderscheid (Rhld-Pf.) (Fig. 39) are examples. They may be small inner rooms equivalent to what would have been called a closet in seventeenthcentury England – a small room used for one of several purposes by one person.

PAVILIONS NOT RECTANGULAR It might be expected that the Roman predilection for polygonal and apsidal buildings, forms particularly suited to display, would manifest itself in pavilions which, as much as any part of a house, were intended as a mark of status. Rather surprisingly, these more interesting shapes are rare. One is the otherwise unremarkable BrewoodEngleton I (Staffs.), embellished by the addition in II of two bow-fronted pavilions.14 But at Brewood, as in so many houses, one pavilion is flush with the gable-end wall and the other, the left one, is given greater prominence by projecting beyond the end of the house, so that shape is not the only mark of distinction. Among buildings of its size and type Brewood proclaims its owners’ advanced taste. 124

— The Porticus-with-Pavilions: Pavilions — Lauffen am Neckar has, as already mentioned, a polygonal room which is a fairly sophisticated kind of enlargement intended, perhaps, to convert the principal pavilion into a dining-room. At Hüfingen (Bad.-Württ.) (Fig. 34) the apsidal front of the left-hand pavilion indicates a use rather like that proposed at Stahl, with the place of honour facing the house and the doorway into the room close to the hall, leading directly to the supposed circulation/service area. Polygonal pavilions in the late phases of Fliessem and Stadtbergen are irrelevant because these villas are so large and imposing as not to be comparable, yet apart from them such a feature is hardly to be found among the huge number of villas in Germany and France. A remarkable outlier is the Hungarian villa of Hosszuhetény, which has two markedly unequal hexagonal pavilions – a shape so awkward by conventional standards of room use as to suggest they were designed for architectural effect and leisure. The same can be said of the two round pavilions at Orlandovtsi (Bulg.) (Fig. 60). Regular geometrical shapes do not exhaust the possibilities. A few widely scattered villas have pavilions of irregular shape in otherwise rectilinear buildings. At Whittington III (Glos.) (Fig. 34) the south pavilion forms a five-sided figure, has a hypocaust and mosaic floor and was the most important room in the house, facing a pleasant prospect down to a stream. Where it was entered is not known definitely but, if surviving walls are a guide, it was from the larger adjoining room. That also has a hypocaust and the pavilion may have been an inner room to it, with the place of honour in the three-sided apse. Interpreted thus, it is a small version of the pavilions at Stahl and Kingsweston. Yet another is found at Regensburg-Burgweinting 3 (Fig. 41), where a canted pavilion forms an inner room to an irregularly shaped room intruded into the hall. The point of these irregular shapes may be to reproduce on a small scale the imposing setting created by the canted corners of the Stahl pavilion, and making them inner rooms conveyed the impressive sense of distance otherwise resulting from size. Some houses have one of two pavilions built with the front canted towards the path by which the villa was reached, as if to enable persons approaching to distinguish which was the superior side of the house. This may have been intended at Burgweinting, and Friedrichsdorf-Seulberg (Hessen) is another instance among a considerable number like this.

MINIMAL PAVILIONS Pavilions first came to serious notice as ‘corner-projections’ and the notion of projection, though fundamental to their definition, was applied sufficiently flexibly to embrace villas in which the visual effect was slight. Swoboda extended his definition with reference to a building, itself looking like a small villa, at the entrance to the courtyard of the great establishment at Fliessem (Fig. 43). Taking the then current view that the space in the middle of the building was an open yard, he described it as an ‘extended villa with small internal court and outer porticus on one side between pavilions or [their] rudimentary forms’. 15 This was said of a house which has a pavilion of minimal projection and, at the other end, a room flush with the 125

— Chapter Eight —

Figure 35 Minimal and quasi-pavilions

porticus. On the other hand he made no mention of villas with corner projections which are not really pavilions in the usual architectural sense, those called here minimal pavilions. An example is the principal house of the small villa of Langenau Osterstetten (Bad.-Württ.) (Fig. 35). Its two pavilions, about 7.25 m in depth, project for little more than a metre at the front and less – and unequally – beyond the ends of the hall,16 so that, as in some previous instances, the slightly larger and better appointed pavilion is the less prominent. No porticus was found, and some kind of construction comparable to that at Stratford-upon-Avon-Tiddington (War.) (Fig. 35) is likely.17 It follows that the front wall was solid and the pavilions projected hardly more at the front than they do at the ends of the Titelberg model.18 What was the point of such minimal pavilions? If the hall had low eaves they may have had more of the customary visual effect of the standard house front than the plan suggests, yet some of the contrast between dark hall and light pavilion, and some of the privacy must have been lost. On the last two counts they cannot have differed very much from rooms at the end of open halls like Quinton (Northants.) and Rothselberg. Their primary purpose was to make a symbolic statement, to claim a tenuous affinity with the Roman world at the least possible cost.

PAVILIONS IN ROW-TYPE HOUSES However hard it may be to assign specific functions to the pavilions of a simple hall like Serville, they clearly provide a new kind of accommodation. That is not selfevidently true of even the simplest row-house, and only the indiscriminate use of blanket terms like ‘wings’ or ‘Eckrisaliten’ has concealed the problem. Because row-houses are so varied in the number and size of rooms, it is difficult to select one form of plan and argue a range of functional possibilities, as could be done with the simplest halls. A beginning may be made with the three-room houses of the Somme basin, which are simpler and less altered than the many well-excavated rowhouses in Britain. Bouchoir (Fig. 29) and Marchelepot (Fig. 43) have pavilions in the classic relation to the core building but Rainecourt (Fig. 68) shows the commonest application in row-houses, in which the notion of projection is abandoned and the pavilions are reduced to small appendages rather than major features of the front elevation. In its classic form the pavilion looks like a private room or parlour, as in


— The Porticus-with-Pavilions: Pavilions —

Figure 36 Terminal rooms

hall houses like Serville. Not so, though, at Rainecourt (Fig. 68), Sparsholt (Hants.) and other row-houses where the rooms at the end of the porticus are too small to perform any such function. Some may have been internal porches or lobbies; ShiphamStar 2 I (Fig. 37) suggests the possibility that others were service rooms to the rooms they adjoined; a porter’s lodge is another alternative.19 Small rooms like these do not conform to the basic idea of an Eckrisalit because they do not project beyond the perimeter of the house-plan, only, perhaps upwards above the porticus. Swoboda, discussing the little villa of Bachenau (Fig. 36) and its analogues, did not call such rooms pavilions,20 nor did he give them a name, so I propose to call them, with implicit reference to the porticus, terminal rooms; and that, whatever its deficiencies, is an improvement on ‘wing’, as in wingedcorridor villa. Sparsholt has two terminal rooms which have earlier been suggested as lobbies (chapter 4). Two are likely at Downton (Fig. 10) because it is so symmetrical in other respects, and there they are unlikely to have been lobbies because each apartment already has its own. The one terminal room excavated was an inner chamber opening off a large one, and again the term ‘closet’, meaning a small subsidiary room adaptable to many purposes, may be useful. Terminal rooms are very rare in hall houses. One example is Beckingen (Saarld) (Fig. 36), where a slight set-back in the front wall shows that the terminal room 3 at the north end antedates the bath suite and is original. Another is Broichweiden (Fig. 36), where a wide porticus allows a larger than usual terminal room as large as some pavilions.


— Chapter Eight —

DETACHED WINGS OR QUASI-PAVILIONS A fairly rare kind of structure which has caused some confusion is the detached room or wing standing in front of a porticus and separated by it from the core building. It is very difficult either to devise a reconstruction for villas which possess this feature or to find a reason for separating comparatively small parts of a not very large building. The prime example is Houdeng-Goegnies (Neth.), which misled several people into believing it analogous to Ditchley (Oxon.).21 Another is Newton St Loe S (Wilts.), where, at the east end, a quasi-pavilion separate from the core rooms provides balance to the terminal room at the west end. A third villa having this feature is Brewood-Engleton III (and probably also II), but the evidence is less clear than for the others. This strange departure from general villa practice may be related to the way the units of a row-house were entered by lobbies. Brighstone-Rock I (Fig. 36) has at each end a transverse lobby which is reached from the porticus. Had the pavilion (6) been added to the core structure in the usual way it would, by blocking the porticus, have made the east lobby redundant and necessitated much rebuilding at that end to create a new one. Adding it outside the porticus preserved existing social arrangements to a large extent, although they must have been modified in ways now hard to see. Houdeng-Goegnies may be explicable in this way and probably Brewood too.22

ONE STOREY OR TWO? In recent years British archaeologists have argued increasingly that the pavilions of their villas had two storeys. Insofar as any evidence is adduced it is the greater thickness of the pavilion walls than those elsewhere on the site, and if one pavilion has thicker walls than the other, symmetry demands that they be the same height, come what may. External projections are another favourite, to be unhesitatingly interpreted as buttresses without regard to probability, perceptible necessity or structural considerations. Why a pavilion should need several buttresses at the front – and it is usually at the front – and only the odd one at the side is never explained, yet the thrust of a square structure is fairly uniform. In this as in all other matters the need for a reconstruction to conform to structural logic23 is totally ignored. Nor, as far as I am aware, has anyone ventured to mark clearly on a plan where the staircase stood – in some cases the only possible place is over a hypocaust – and still less to indicate what form it might have taken. Very rarely is anything said about the materials of which the staircase was built: timber, presumably, but a structure intended to carry people at all mindful of their own safety, let alone one destined to have heavy loads carried up or down it, needs a foundation of some solid material, whether stone, brick or concrete. Among the large number of villa plans examined for this book only three instances have come to light where plausible evidence has been adduced for a staircase. Two are in the Iberian peninsula: Guadaira-Alcala/ Casa de Pelay Correa (Seville) and Vila de Frades (Portugal) both retain stone steps – in the latter villa, more than a dozen.24 The third is Stanwick-Redlands Farm (Northants.), where a stone foundation was interpreted as being that of a staircase; the problems of first-floor circulation it raises have not so far been addressed.25 128

— The Porticus-with-Pavilions: Pavilions — The space needed to do the necessary demolition job on the many alleged upper floors is better used more positively. Besides, such an exercise can provide for anyone of a sceptical turn of mind hours of innocent amusement for the long winter evenings in exposing the absurdities that abound in every aspect of villa reconstructions, not just pavilions; it would be a pity to spoil the fun.

CONCLUSION Pavilions need closer analysis than has been attempted here. Some have hearths or ovens, some were built as part of a bath suite or converted to that purpose, some contain mosaics. A small proportion intercommunicate with an adjoining room, whether by an original doorway or one broken through later. On the whole, though, these instances appear to be the product of particular circumstances. Most pavilions are explicable in their initial phase in terms of those examined above; some that underwent enlargement will be considered below (pp. 261–2). The distinction between pavilion and terminal room is important for the bearing it has on the transmission of ideas; the relation between the two, and why the more localised form developed, are questions to be discussed in the wider context of villa development.





he problem bedevilling the understanding of many elements of villas, that they serve more than one purpose, makes it important to analyse which of the possibilities is primary in any given context. This might be thought not to apply to the porticus, yet this apparently simple feature has several variant forms which relate to the houses behind them in quite different ways. A porticus is nearly always a mode of architectural display. Functionally it may be a corridor giving access to the rooms behind it, a gallery overlooking the countryside, or an enclosed room. Entrance may be at the front, either by an imposing doorway in the middle or by two not quite equal doorways; at the rear, by a middle doorway; or into one or both ends of the porticus. Its length and breadth can be varied to serve particular needs or to give a desired emphasis. In any one villa two or three of these purposes can be combined or they can change from one phase to the next, and it can be hard to decide the intention, part functional, part iconic, underlying a particular example. These problems will be dealt with under the three heads of social structure, recreation and living-space.

OPEN-ENDED PORTICUSES: AN EXPRESSION OF SOCIAL STRUCTURE The simplest kind of porticus is open at both ends and extends from end to end of the house, as at Aylesford-Eccles I (Kent) (Fig. 66). It perpetuates the purely utilitarian function of the timber veranda open at one end and not extending the full length of the house but, merely by its greater symmetry, creates greater possibilities of architectural treatment. To what extent this opportunity was seized is not clear and, if turned wooden columns were used on a waist-high stone wall, may never be known. But Eccles I, like every provincial-Roman house of any quality, has a superior and an inferior end, here north and south respectively. So what was the point of having both ends of the porticus open, and how did those approaching it know which was the appropriate point of entrance for them? In a villa sophisticated 130

— The Porticus-with-Pavilions: Porticuses — enough to incorporate a long ornamental pool the immediate surroundings will have been laid out to govern the decision. One path led to the entrance at the south or lower end where an open-ended room was probably a domestic shrine, as a similarly placed but shallower room at Farningham-Manor House I (Kent) (Fig. 15) appears to be. Nearly everyone entered here. The end of the Eccles house may have resembled in a general way the shrine from Mainz-Kastel (Fig. 37),1 but with a square-headed opening under a pediment, a king-post roof and simple ornament or painted walls (Fig. 37). Blankenheim I (Rhld-Pf.) (Fig. 70) has, at the entrance to the porticus, a wide, shallow version of this.2 Architectural detail would have made clear the importance of the shrine and those entering the house would expect to make an offering. Inside the porticus some architectural feature or ornament or merely painted decoration distinguished the entrance to each apartment; additionally, emblematic barriers such as a wooden arch or a carved tie-beam may have marked stages in the progress towards the furthest and most important apartment. They were the predecessors of the arch spanning the porticus at Ormalingen (Switz.) (Fig. 15), and formed customary barriers warning the caller not to venture uninvited, though in fact an attendant waiting near the shrine is likely to have guided any visitor’s progress. Only the privileged would enter at the north or upper end of the house or perhaps go out to a garden. The pool was placed nearer the south or inferior end without overlapping the putative shrine, thus emphasising both it and the best apartment at the north end as viewed from the west; the latter may well have had some external treatment to distinguish it further.3 No other villa as long as Aylesford-Eccles has an open-ended porticus unless, perhaps, Mareuil-Caubert (Somme).4 Only Ault in Picardy has a porticus on both long sides; why should so small a house need two? They may connote two equally important elevations, front and garden being the conventional possibilities. More likely, in a house as small as this, they indicate the social importance attached to separate entrances to two apartments, and they extended the full length of the building because the two were not fully independent; here the largest room – a workhall? – is at one end. Whatever idea underlies Ault appears in modified form at Monchy-Humières (Somme), where the wider porticus of two is open at both ends and the other is closed at one end by a wing larger than any of the range of five rooms; its size suggests a workhall in a position comparable to those beside the entrance at Farningham I and Blankenheim I. Open-ended porticuses giving access along the front or rear were appropriate to a series of apartments with some degree of hierarchy but no dominant one, and the choice of this direction of approach rather than one at right angles is fundamental to porticus development. Aylesford-Eccles I is the most conspicuous example and an early one, but the same idea was expressed much later in the advanced planning of Blankenheim IIIA. This kind of porticus needs a series of openings, which might take the form of a colonnade, to give light, but otherwise, lacking a focal point, does not lend itself to symmetrical treatment or architectural enrichment except at the entrances; Halstock (Dorset) in its later phases exemplifies this point in an otherwise rather exceptional porticus.5


— Chapter Nine —

HIERARCHY IN PORTICUSES OPEN AT ONE END Porticuses open at only one end are easier to understand as a way of approaching units of not quite equal status than those open at both ends. They have a precursor in the timber veranda, of which the best-known example is Welwyn-Lockleys I. An earthfast veranda always terminates short of the end of the building, and the reason for this is plain: because entrance to the house is gained from only one end the far end room is entered at the nearest point, and to extend the veranda would have had no conceivable advantage; and where the main elevations do not lend themselves to display, nothing was lost by curtailing the simple timber structure. This utilitarian attitude persisted in some porticuses built or replaced in masonry, such as Bramdean6 or Köln-Mungersdorf II (the east porticus). As masonry footings and, increasingly, masonry superstructure superseded timber, the architectural possibilities of the new material, coupled with a growing exploitation of symmetry, caused the porticus replacing the veranda to be extended the full length of a house. This left a space at the far end, beyond the last doorway, for which a function had to be found. Chilgrove 2 II (Sussex) exemplifies this kind of development, as, probably, does Manfield-Holme House (Durham) (Fig. 68) where the evidence is less well preserved. Shipham-Star IIA (Som.) (Fig. 37) is a particularly interesting example. A hall house in I was replaced in IIA by a row-house entered at one end of a full-length porticus, at the far end of which, and flush with it, is a minor room, probably a service room. The two doorways on the west side of room 3 point to there being a timber partition between 4 and the porticus and suggest a kind of use for the end room that may have occurred elsewhere. Northchuch IIB, a two-unit house, had a porticus open at one end and closed at the other by a small terminal room of uncertain purpose.7 The open-ended porticus could also lead to a pavilion, as at Démuin (Fig. 37) and Plachy-Buyon. Some social nuance must be implied by one instead of the usual two pavilions, and in the context of a kin-group it may imply the emergence of a dominant household – one of the many ways in which that kind of social change is expressed. A single pavilion, being larger than a terminal room, has a more important function, and no doubt carried something of the symbolic significance of the porticus-withpavilions front. Porticuses open at one end or both are commoner in Britain and France than elsewhere. They are found, as the preceding examples demonstrate, in row-houses lacking a large middle room. It appears that once such a room, large and in the most prominent position, is introduced into a row-house of two units, the porticus is entered in the middle. Some major new function has been brought in with resulting changes in social life. A hint of what it may have been emerges from Newport (I.o.W.) and Downton, both with traces of what can plausibly be interpreted as a religious function, in addition to the commonly ascribed one of a dining-room.

THE ULTIMATE OPEN-ENDED PORTICUS: CSÚCSHEGY More important and more complex than these is Budapest III-V Csúcshegy (Hung.) (Fig. 37). At the entrance to the porticus a doorway on the right leads to a workhall 132

— The Porticus-with-Pavilions: Porticuses — which is, as is commonly the case, the largest room in the house; typically, it was also the kitchen and accommodated the stoke-hole for a hypocaust in the adjoining room. The porticus led on past a pair of interconnected rooms with hypocausts, terrazzo floors and wall-paintings; they were entered from the next room in line, which had a boarded floor.8 The good-sized room in the north-east corner also had a boarded floor and possibly the adjoining one to the south too, so they were certainly a pair of living-rooms. So far the proportions of the rooms fall within the usual limits of variation; the two at the rear of the house are outside them, one being twice and the other three times as long as wide. Opening off the former, the one in the south-east corner, is a bath, so it was a public room, not private; and the only definitely established entrance to the two rooms to the north, the ones with boarded floors, was from it. The larger of the two elongated rooms had a hypocaust and a remarkable plaster ceiling. How to interpret the plan? It appears to be centralised around the square room at the heart of the house. From there, certainly, the suite of two hypocausted rooms was entered; they form the best apartment in the house and cannot have been reached any other way. If it be assumed that the porticus leads solely to the centralroom, a

Figure 37 Open-ended porticuses


— Chapter Nine — doorway from the porticus at its east end matches that into the workhall by the entrance, so forming part of a symmetrical decorative scheme comparable to that found at Bocholtz and Bierbach. The common hall with its adjacent bath is reached from the central room; so, too, is the apartment comprising the two boarded-floor rooms to north.9 This interpretation gives a centralised plan with two apartments each comprising an inner and an outer room and each with easy access to the common hall from which the bath and the grand reception/?dining-room were approached. However speculative this is, it provides an explanation of the end entrance to the porticus and reveals a coherent, tightly organised plan, the most advanced of its kind. It is a highly developed form of the compact row-house, superior in its subtlety to those of central France. Csúcshegy, indeed, has one of the most remarkable villa plans anywhere and illustrates how adaptable the end-entrance type could be in the hands of an accomplished architect. In the preceding chapter the porticus-with-pavilions was described as the badge of Romanisation. That is true of most villas, yet those with an end entrance could, as Csúcshegy shows, be just as Romanised. Possibly the one indispensable element was a colonnaded porticus, as in Blankenheim II and III, which come very near to having an open-ended porticus (Fig. 70). These plans raise a difficult question, one that cannot be answered at present: what social factor determined this minority choice rather than the standard and universally recognised one of porticus-withpavilions?

TAPERING AND SPLAYED PORTICUSES A few large houses of obviously high status show a setting out of the porticus so far from rectilinear that it must be deliberate. Küttigen-Kirchberg I (Switz.) (Fig. 19) is a row-house comprising two three-room and lobby units and a middle representational room.10 Developments in II show that the right-hand (east) end was the more important, and it is towards that end that the porticus tapers. Wherever the house was entered, and that is not perfectly clear, it was not from the south into the main porticus but from the north or west side. The tapering must have been intended to demonstrate to anyone looking at the villa from below where authority resided, and to induce discretion in those entering the porticus. Similar considerations can be inferred elsewhere, for example at Laufenburg (Bad.-Württ.) (Fig. 38), a house of higher status to judge by its size and the finds. In I it had a stately hall 17 m wide.11 In II two pavilions of proportionate size, bigger than those at Stahl, were added; they were connected by a porticus tapering as strongly as the one at Kirchberg towards the bigger pavilion, from about 3 to 2 m, and at that end of the hall lay the private or inner rooms. Tapering could be used less obviously to the same effect. It is apparent to a modest degree in the Belgian hall house of Vesqueville, where the front porticus narrows from right to left, presumably towards the superior porticus, the one equipped with a hypocaust. Even slighter emphasis is found in the small house at Eaton-by-Tarporley (Cheshire) (Fig. 16) which is entered from the veranda by a longitudinal lobby widening towards the baths wing and narrowing towards the end with the small suite of private 134

— The Porticus-with-Pavilions: Porticuses — rooms. The divergence from parallel in the lobby walls is so slight that it would not be worth remarking if the rest of the building were not wholly rectilinear, and if the tapering did not correspond to the same order of importance in the rooms as it does in the grander examples mentioned above. Even though the visual effect was heightened in other ways, by paint or mouldings, for instance, the architectural code must have been universally recognised to be worth applying. Some similar message was conveyed by the few porticuses which splay from a middle entrance on one side only. There are two well-known examples, one Belgian, one British. At Chastres, otherwise Chastres-lez-Walcourt, the right (north) porticus is orthodox, the left is sharply angled in a way that makes it difficult to envisage in its original complete form. It must, nevertheless, have been something like Beadlam N (Fig. 17), where a similar juxtaposition occurs, and in both houses the widening part seems to have served the superior side. A third house, Broichweiden (Fig. 36), splays out from the entrance to the left, with nothing definite to confirm that side as the less important. It is clear that the message thus conveyed did not necessarily distinguish between two domestic units, only between superior and inferior ends.

Figure 38 Tapering and splayed porticuses


— Chapter Nine —

TWO HOUSEHOLDS: SYMMETRICAL ENTRANCES A few villas have two entrances into a porticus. Sinsheim-Wald (Fig. 8) and Maulévrier (Fig. 13), both published some 150 years ago, are like this and it seems that in all those years nobody has discussed what is surely a rather remarkable feature. Several more can now be added. At Sinsheim the right-hand entrance lies opposite a bay of the internal arcade; the other, nearer what may be the lower end, faces a stretch of blank wall extending between the two entrances; and the two were differentiated by their architectural treatment. Neither external nor internal symmetry is perfect, and the underlying intention can only be to differentiate between two households of not quite equal standing. Herschweiler-Pettersheim (Fig. 39) expresses the idea in a rather similar way. The main entrance faces the doorway into the hall, the lesser one a blank wall, so the visitor favoured with entry to the equivalent of an inner room – whose relative status is shown by its greater distance from the hall doorway – has to turn left to the slightly bigger pavilion. At Köln-Mungersdorf III–V (and no doubt I and II) (Fig. 71) the two doorways were at the north and south ends of the porticus, one leading to the workhall, the other to the apartments. Had pavilions not been provided from the first, the house would no doubt have been entered only at the north or lower end along an openended porticus. A generally symmetrical exterior was evidently a prime aim but internal arrangements reveal more clearly the social requirements met by the placing of doorways and passages. At Maulévrier the bigger of two entrances faced a slightly larger room than did the smaller, and the minor breach of symmetry caused by the different-sized doorways – no doubt matched by the treatment of the two internal doorways – was the minimum price of maintaining social distinction. Romegoux-La Vergnée (Fig. 13) shows a slight variation of this pattern in having two porches within the porticus, one of which, to the west, opens into a large hall-like room, the other into a smaller square room. Hemel Hempstead-Gadebridge Park (Herts.) (Fig. 13) achieved the same result differently. The main range appears to comprise two Downton-type units and a further room at the east end. Both porches led to transverse lobbies, that to the west facing its porch, that to the east not, so that anyone aware of architectural convention knew which unit, probably that to the east, was the more important. This kind of differentiation in the placing of doorways, locating them carefully to convey some social distinction within the houseful, is a constant factor facilitating the interpretation of a villa plan. It was possible to have axial symmetry externally and confine the expression of social differentiation within the house. Chedworth W (Fig. 74) does that, with a single entrance in the porticus by which the range was entered, whence the person entering could go left or right to either of the two units. It is a row-house equivalent of the way the hall at Manderscheid (below) was entered. This arrangement is a possibility in any hall-type house where no doorways were found and the principal hearth lies to one end of the hall, or in a row-house where the core rooms are clearly divided into two more or less equal parts. This group of villas at least can hardly be understood without assuming occupation by two or more households.


— The Porticus-with-Pavilions: Porticuses —

TWO HOUSEHOLDS: PORTICUS-WITH-PAVILIONS FRONT AND REAR A few hall houses have a porticus-with-pavilions on both long elevations. Manderscheid (Fig. 39) and the two Belgian villas of Doische-Vodelée (Fig. 39) and Habay-

Figure 39 Houses with two porticuses or two porches


— Chapter Nine — Rulles are the only absolutely certain ones; Dirlewang (Bay.) (Fig. 39), NeumagenDhron/Papiermühle (Rhld-Pf.) (Fig. 68), GrossSachsenheim (Bad.-Württ.) and Lemiers (Neth.) are all highly probable. They present problems: why, if the porticus-withpavilions is a proclamation of Romanised ways, should it be needed front and rear? And why are there so few? Swoboda’s explanation that it was a matter of taste12 is unhistorical. The display front (Schauseite) of Manderscheid, in the Eifel, overlooked the valley of the river Kyll to the south but was not the entrance. The only external doorway discovered, and at 3 m wide it was certainly the most important, was in the north porticus, from which, near the ends, two doorways opened into the hall. They must connote some spatial and social division within the hall which was matched by the provision of two hearths; one was probably in or near the south-west corner13 and another might be expected near the south-east corner, in which case the entrance/ hearth relation resembled that at Stahl. In the middle of the wall between them was another opening, as wide as the outer doorway but rather enigmatic in detail (Fig. 39, detail); it was usable, according to aus’m Weerth, either vertically barred as a window or unbarred as a doorway.14 Here, then, the functions of the porticuses are clear. One, commanding a view of the valley, gave access to the south pavilions. The other formed the house entrance and provided the only access to the north pavilions and the cellar; it was much like the porticus at Serville and many other villas. Manderscheid has a more obvious balance of room functions between front and rear than most comparable villas, something made possible by the placing of the small bath suite at the end of the hall, not at a corner of the house as at Habay-Rulles. Only Dirlewang (Fig. 39), where the plan was not completely recovered, may have had equal rooms on both sides of the hall, although Habay-Rulles III was much the same15 until rebuilding made the east or entrance side markedly more important, and the balance of importance was further reduced by incorporating one pavilion in a bath suite. Any use of pavilions for some special purpose common to the houseful marks a change in the near-equality on which the provision of two porticuses is founded. Doische-Vodelée and Neumagen-Dhron/Papiermühle both developed on those lines, so that the front porticus became linked to the social aspect of bathing. It appears that in most houses with two porticus-and-pavilion fronts one of them was built primarily to impress the outside world, since other purposes could be achieved in other ways. Usually the front porticus was both recreational space, as a gallery commanding a view, and utilitarian in affording access to the pavilions. The unusual opening from hall to gallery at Manderscheid deserves consideration in this connection: it was not an orthodox doorway, and the possibility that it could be closed completely and used as a window for longer or shorter periods implies that the gallery was not in continuous use and that the pavilions could be reached from the end rooms. That still leaves the historically more important question of why, setting aside special uses of pavilions, four virtually isolated rooms, sometimes of markedly different size and always necessitating two porticuses, should be required. If Dirlewang is viewed as a manor house or Herrensitz, the rooms, only three of which are certain, are difficult to envisage as private rooms used by members of a family; they are all too nearly alike to have been built for that, and some grouping together might be expected. In the light of the duality discernible at Manderscheid 138

— The Porticus-with-Pavilions: Porticuses — they may have been shared between two households, each having one on each side. That, coupled with the social necessity of a Romanised appearance and the desire for a view (and with it, perhaps, a sense of lordship over the countryside), could account for the architectural form the presumptive social structure took. But since an impressive entrance was not always felt necessary, as Horath and Great Witcombe show, a rear display-front was not essential; and the reason why much less prosperous and comfortable villas than those needed one must relate to the extent and completeness of the duality presumed to exist in the houseful. Manderscheid needed a second porticus-with-pavilions; Stahl did not. Few row-houses have a porticus-with-pavilions front and rear. MunzenbergGambach (Fig. 31) is the most interesting of them and the easiest to understand; two ranges, each of two households, require four pavilions. Gayton Thorpe N (Fig. 39), a house of gradual growth, is another example. Beginning in [I] as a core building of row-type, it was improved by the successive addition of porticuses and pavilions,16 in [II] at the front and in [III] at the rear. As in other villas, this duplication is taken to correspond to two households and, if Gambach is taken as a guide, each of those embraced two units of consumption; yet it is impossible to divide into four units without having one passage-room or assuming the north-east porticus was a livingroom. Possibly some alteration occurred in [IV] when the rear porticus ceased, following its partial blocking by the enlarged middle room, to be a vehicle for display, and the character of the front porticus was changed by creating a grand entrance. This architectural change is the consequence of social change, reflecting a new concentration of power within the kin-group; and the linking of the two houses may be part of this process. These considerations affect the interpretation offered for other houses with two porticuses of roughly equal status, such as Rockbourne [II] (Fig. 42) and Bray-surSomme. Left out of account are the many houses with porticus-and-pavilions at the front and a simple entrance porticus at the rear, such as Kleinsteinhausen (Hessen) and Merdingen (Bad.-Württ.). But the discussion also has a bearing on Radley-Barton Court Farm III, which has a modest display-front and was entered, as its small enclosure reveals, at the rear. Although many British houses are like Barton Court Farm, few have sufficient detail to show where they were entered and, perfectly naturally, a front entrance is assumed – perhaps wrongly in not a few cases.17

PORTICUS FUNCTIONS: RECREATION Pavilions or small rooms were not the only ways of terminating a porticus. A very simple alternative occurs at Marquivillers (Somme), where the porticus extends a short distance at both ends beyond the building, forming lug-like projections in order, perhaps, to set off the structure behind and emphasise its bulk. Such lugs occur at Vicques (Switz.), at the ends of the principal or front porticus and beyond the point where it returns forward to link detached pavilions a little distance away. Since the greater bulk of the pavilions would have destroyed any architectural emphasis of the kind supposed at Montivillers, these lugs had a different purpose. They are slightly narrower than the porticus and may have been intended as viewing places 139

— Chapter Nine — overlooking gardens, fields or a more distant prospect, as, too, were the comparably sited rounded projections at Hummetroth (Fig. 18); and the same intention may underlie the projecting north end of the porticus at Blankenheim IIA (Fig. 70), where the courtyard could be viewed from the superior end of the house. A variant form appears in the south building at St Lythans-Whitton VI (Fig. 65), where the ends of the porticus are a bulbous shape – a rather surprising touch of sophistication in a remote district and one for which the parallels are mostly distant, the nearest being Pulborough (Sussex) (Fig. 45) and Fishbourne Palace (Fig. 47). Others are the Bavarian villas of Westerhofen and Peiting, and of the latter, which is a sophisticated building, it was noted that apsidal ends are associated with great buildings such as Fliessem and Nennig.18

RECREATION: GHOST PAVILIONS A few houses have porticuses breaking forward at the ends as if to go around pavilions which do not in fact exist. Thus at Frankfurt a. M.-Bornheim (Hessen) (Fig. 32) each end of the porticus breaks forward quite widely, reducing the middle part to comparative insignificance. This is a particularly interesting example because the most unusual plan provides nothing to emphasise; behind the porticus the rooms are flush. Nor is this the only house in which the porticus breaks forward without visible cause, although such a break is usually combined with a balancing pavilion.19 Bergen -Auf-dem-Keller (Hessen) (Fig. 40) is like this; its frontispiece is not at all closely related to the rooms behind, a clear indication that two building phases are involved, with [I] having a plan much like Rothselberg. The empty projection forming part of [II] is wider than the balancing pavilion and projects beyond the end of the house; it may have been intended to emphasise the upper end. British porticuses breaking forward in this way appear in the hall houses at Wraxall (Som.) (Fig. 3) and Cherington (Wilts.), and also the row-house Great Weldon I (Northants.) (Fig. 70).20 A projection of this kind is found even in as highly developed a hall as Raversbeuren21 and its appearance in a house of this size and quality shows that the combination of pavilion and empty projection must have had a significance now hard to recover. The explanation may be that porticuses generally, and particularly those described earlier as galleries, were the provincial Roman counterparts of the Long Gallery in Elizabethan or Jacobean manor houses: places used in bad weather for exercise and at other times for talk and relaxation. On general grounds, provincials with Roman manners might be expected to behave more like the men of seventeenth-century England than people today, and from that most revealing source for social habits, Samuel Pepys’ Diary, emerges a preference for conducting business or private conversation between two people while walking – which is in strong contrast to the sedentary habit of the twentieth century. ‘Come, Mr Pepys, you and I will take a turn in the garden,’ said the Lord Chancellor Clarendon when he wished to discuss a very serious matter, and the diarist was well used to conducting naval business as he walked in the gallery at Whitehall with the Duke of York, or on Whitehall Bridge with a colleague; and once at the Exchange he ‘walked two hours or more’ with a 140

— The Porticus-with-Pavilions: Porticuses — merchant acquaintance. Even at home Pepys and his wife would walk up and down in their Great Room as they talked of their private affairs.22 The idea of a place for relaxation may be relevant to two Swiss villas which show the breaking-forward porticus in minimal form. Kirchberg [I]23 (Fig. 19) had at each end a squarish space too small to be a room; the one at the west end is only 2×2.8 m and that at the east end is scarcely larger. It is typical of the misapplication of Eckrisalit that the word is used of these little projections overlooking the valley of the River Aar;24 something different is needed, and since they must surely be places from which to admire the view the term ‘viewing-places’ may be appropriate. It can properly be applied to a second example, Ferpicloz, a villa built on a peninsula in the River Sarine, where two such projections overlooked the yard and approach in front and the two arms of water on either side. They appear to be the equivalent of lugs or rounded ends. Viewing-places are appropriate to rather spectacular situations but can occur elsewhere, as in Picardy at Villers-sous-Ailly.25

THE PORTICUS AS LIVING-SPACE Many porticuses were fronted by a half-height wall carrying stone columns of classical appearance above, but not all can have been like this. At Sinsheim and Kings Weston (Glos.) the wall between hall and porticus was pierced by arched or square-headed openings in which no evidence of doors was recorded. Consequently it has to be supposed that the front of the porticus was enclosed by glazed windows or shutters, the latter hinged or otherwise movable, and at suitable times light would be gained by opening doors. A further reason for thinking that some porticuses were enclosed is the presence of quite elaborate mosaics. A little hall-type house at Great Staughton (Cambs.) (Fig. 40) is like this; structurally it is perfectly symmetrical but decoratively there is

Figure 40 Ghost pavilions and porticuses as living-space


— Chapter Nine — a difference, one of the two mosaics in the porticus being better than the other. One purpose was probably to show in which direction the superior pavilion lay, yet, in a building of this small size, so well-appointed a space may have been more than just a corridor. Although each half of the porticus has unusual proportions for what is likely to have been more of a room than a corridor, when compared to the elongated rooms found at Hohenfels, Sontheim a.d. Brenz and Wahlen it is within the normal provincial-Roman range. The part of the capital of a stone column that was found may have been to do with the main doorway rather than a colonnade. The plain piece of pavement between the mosaics is, in effect, like the internal porch found in some porticuses, of which Romegoux is an example (Fig. 13); a porch of that kind would be pointless unless the porticus were enclosed. Evidence to the same effect is the presence of secondary hearths. These are usually regarded by British archaeologists as a sign of ‘squatter occupation’ in the final stages of a villa’s existence. Newel IV (Rhld-Pf.) (Fig. 24) demonstrates the contrary. Built in the north half of the porticus are an oven and a hearth which would, in Britain, have been attributed to people living in comparative squalor during the final stage of the villa’s decline. Here, that explanation is ruled out by the fact that a stone wall was built to separate the newly created room from the rest of the porticus; it is an improvement in the amenities of the house at the cost of its architectural formality. Where squatters are invoked it is rarely clear that the principal heated rooms were abandoned before the use of a porticus as living-space began, and it is never explained why a porticus with its presumptive open colonnade should have been preferred to the makeshift adaptation of rooms: a house falling into disrepair is likely to have damaged a porticus through falling roof tiles before the core rooms became unusable. Since so many houses suggest in various ways that porticuses were commonly used as rooms, squatters should be consigned with bailiffs to the dustbin.

PORTICUS AS WORKPLACE: WIDE OR CONTINUOUS PORTICUSES A minority of porticuses serve a utilitarian purpose. Some are merely service corridors as, for example, the one at the rear of Köln-Mungersdorf II (Fig. 71) which links the cellar and the dining-room; they are easily identifiable. A more enigmatic group comprises those that are conspicuously wider than the norm for a given kind of house and this is something to be decided, at present, by eye rather than measurement, because what looks right for a row-house in Britain is too small for one in Switzerland. A second problematic group overlapping with the first includes porticuses running continuously (or nearly so) around a comparatively small row-house, so that the amount of what is commonly called ‘corridor’ is altogether disproportionate to the amount of living-space. A few are so wide as to make it certain that their function cannot have been simply that of a corridor. Thus the three-room house at Flocques (Somme) has a continuous porticus on three sides which is half the width of the core rooms, and if it had a timber veranda at the front they amounted to nearly half the total floor area. 142

— The Porticus-with-Pavilions: Porticuses — The extreme case is Ovillers (Somme) (Fig. 40), where the three rooms occupy a third of the whole house area. Ribemont is the more remarkable for having a porticus only at the front, and so wide (5 m; rooms 6.5 m) that the applicability of the term may be questioned.26 Differences of this kind appear in British row-type villas too, for example Marshfield-Ironmongers Piece, where the north house had a front porticus of normal width and the south a wide one. Continuous porticuses are rare; those on three sides, one long and two short, slightly less so. Wancourt (Somme) (Fig. 29) is like this; the porticus is structurally in two parts, looking like two L-shaped porticuses open at one end, so that each of two one-room units was entered at opposite ends of the house or at least at the ends of the front.27 These porticuses do not have the iconic associations with pavilions identified by Swoboda, and the wide ones cannot have been simply corridors or galleries. On this ground alone such porticuses had some utilitarian function not at present known, although their general nature is apparent at Marshfield-Ironmongers Piece28 (Fig. 66) where one of the two end-to-end row-houses has a considerably wider porticus than the other; this was the inferior end of the villa, where much of the work was carried on. In southern Germany a yet smaller house, Rheinfelden-Salzbrünnele (Bad.-Württ.), being surrounded on three sides by a porticus forming half the total area, conveys the same point as Marshfield. Sizeable parts of the porticus must have served work purposes and been separated from the rest by customary use, not partitions. Storage of winter fuel is possible, as happens under the boldly projecting eaves of old farmhouses in Bavaria even today, and so is the kind of ancillary farm task that needed light, shelter from rain and space for equipment, such as fashioning wooden tools. The three distinct parts of the porticus were differentiated by width, no doubt corresponding to function; the narrowest part on the south side was the point of entrance, with the successively wider parts to east and west serving different, or additional, purposes. Recently Gaubatz-Sattler described the porticus of a villa, with reference to Bondorf, as a multi-purpose area which might include cooking, corn-drying or a dwellingplace.29 This is a timely reminder for British archaeologists of how inappropriate their use of ‘corridor’ is, for Gaubatz-Sattler’s remark is certainly true of late villas and probably always of wide porticuses, even though hearths may be lacking in the earlier periods. A curious feature of some rear porticuses, whether continuous or not, is that the perimeter wall is too close to permit entrance on that side. Villas at opposite ends of Europe are like this: to west Mézières-en-Santerre/Le Ziep (Somme) and to east Chatalka-Delimonyova Niva (Bulg.) and Bihac-Zaloz?je (Yugosl.).30 In Chatalka villa the encircling porticus may have been intended to enhance the imposing quality of a large building which dominated the courtyard and must have been clearly visible outside. That is less likely to apply to Mézières, where a rear corridor may have been desirable from the first in so long a row-house. It may have been utilitarian if, like that at Köln-Mungersdorf II-IV, it was principally for the use of servants.





he only aspect of villas so far discussed is the principal house. It did not exist in isolation, being only one of a group of buildings which usually stood within an enclosure. Virtually no attention has been paid to the varied shapes of enclosures and courtyards, or why the buildings within them are set out as they are. Discussion of these matters here is limited by two considerations. First, although the need to explore the farm buildings as well as the house was recognised long ago, late nineteenth-century methods made robbed foundations unintelligible and missed timber buildings completely, so that it is hard to be certain when completeness was really attained for any one site or phase. Second, the purely agricultural functions of buildings will not, because of my own ignorance of farming, be discussed; excavators’ opinions, where relevant, will be followed. Courtyard and farmyard forms appear in great variety, ranging from the irregular shapes common in Germany to perfectly rectilinear ones in Switzerland, and from the small cluster of buildings of some British farms to the many found in some French and Belgian villas. The profusion of types will be analysed by working from loose groupings of buildings through to the most geometrical and formal of courtyards. This progression, like that of houses, is purely typological, because to include chronology would only create the kind of confusion that has hitherto made a house typology impossible.

IRREGULAR YARDS Regensburg-Burgweinting One of the most interesting German villas is Regensburg-Burgweinting (Bay.) (Fig. 41). The yard, which is approximately trapezoidal, looks like the piecemeal replacement of an earlier enclosure in less permanent materials – a hedge or a wooden fence – and the apparently incoherent disposition of the buildings within it contrasts strongly with the geometrical layout of those in the regular yards of Picardy. Entrances provide a key to its interpretation. Two, on the south-east and south-west sides, were established with certainty; those 144

— The Elements and Forms of Villa Complexes —

Figure 41


— Chapter Ten —

Figure 41


— The Elements and Forms of Villa Complexes —

Figure 41 Irregular and divided yards


— Chapter Ten — on the other two sides were regarded as probable1 and here are taken as certain because then the plan becomes intelligible. They imply a division of the yard into four quarters, two of them having clusters of mostly domestic buildings, two emptier and with only agricultural buildings. The most important sector, the north-west, contains a good-sized tripartite house of incoherent plan which must be of several periods of construction. The core appears to be a squarish hall with a porticus facing the northwest entrance. To it, and facing away from it to west, was added what is virtually a separate house, comprising two hypocausted rooms, two smaller rooms – one with a cellar – a corridor and a porticus; and on the other side was added a second hall-like building. Perhaps then a room was added on the outside (to north-west) of the old porticus, which may have become redundant when the new front was built and was now reduced to no more than a corridor. A number of villas, for example MontacuteHam Hill (Som.), have a comparable combination of elements in very varied sizes and shapes, much as units in a row-house can be arranged in several ways. In the south-east quarter three separate buildings provide in total rather similar accommodation and amenities. The largest is the double-ended hall abutting the yard wall, whose two trapezoidal rooms have already been mentioned (chapter 6). One of them terminates a short porticus, at the other end of which is a terminal room with a cellar. This was probably the side on which the house was entered, and the unusual point of entry2 was probably chosen to face away from the one in the northwest quarter and away from the squarish hall to west; and if the latter point is correct, it implies that the two buildings are contemporaneous or, more likely, that the one to west is the earlier.3 In that building a hypocausted room is evidence that it was a hall house,4 and in the nearby corner of the yard a second hypocausted room formed part of what was presumably another small house.5 Five more buildings, none with signs of habitation, were found. Two in the west half of the yard are more than empty rectangles and require explanation in terms of specific agricultural functions. Probably all the stone structures were found but others less permanent, of timber and clay, probably existed. The important points are that Burgweinting sheltered at least three households, of whatever size, one of which was superior to the others, and that the apparent incoherence of the yard masks a disposition of buildings carefully ordered to reflect social arrangements. One detail of Burgweinting is of wider application. Placing the entrances to the two principal houses (1 and 2) at 90 degrees to each other is found elsewhere, as for example, at Koerich-Goeblingen (Lux.) (Fig. 74). It was evidently a standard way of relating two houses of not quite equal importance.

Ludwigsburg-Hoheneck A second trapezoidal yard at Ludwigsburg-Hoheneck I (Bad.-Württ.) (Fig. 41) shows every sign of being of a single building phase;6 its walls are straight and the two principal buildings relate to the entrance and to one another. There can be no doubt which of the two was the more important; the broad hall on the north side dominates the yard and has an air of authority over the rest. Was there a second house? Paret in 1911 and Klein in 1992 thought not, regarding


— The Elements and Forms of Villa Complexes — the building abutting the south wall as no more than a bath block;7 like most archaeologists, they force the rooms of any not very large building which has baths into a Procrustean bed of rooms derived from Vitruvius. Here, as in so many instances, this procedure fails to explain why the baths took a particular form. As Klein explains them, a short range of rooms in I was enlarged in II into an L-plan around a small space, a yard some 9 m square, in which were two furnaces. Why so unusual a plan was adopted is not discussed.8 I think it more likely that 2 began in a timber phase ‘which cannot in principle be ruled out’,9 as a small hall which was first improved by the addition of a bath suite and then rebuilt in stone. Although the decline in importance of the secondary house compared with its counterpart at Burgweinting is clear, it no longer turns away from the main house but faces it with a porticus, perhaps because, architecturally, it no longer looked anything like an equal. The process whereby one habitation of two or three became dominant takes a variety of forms and reveals so many nuances of relations within kin-groups that is hard to put them in a scale of increasing dominance culminating in the overlordship of one household.

DIVIDED YARDS One expression of the process was to separate the yard into two more or less equal parts. Ewhurst-Rapsley (Surrey) and Bruchsal/Ober-Grombach (Bad.-Württ.) show ways of doing this. From the first Rapsley (Fig. 41) seems to have been divided but only in III can two yards be seen to be separated by a wall.10 East of it was a small house (6) akin to the row-type but not of any orthodox form and to west lay what was definitely another house (1), which may have been aisled but was of equally unorthodox structure. In the superior yard, east of the wall and close to it, was a shrine, so placed, probably, to be accessible from both sides; Marshfield (Glos.) may provide a parallel. Period IV saw the building of another aisled house (2) in the west yard and the conversion of the older one to monospan construction, yet what appears to be the downgrading of that side by the concentration in it of buildings with a partagricultural purpose is denied in V by the embellishment of 2 with pilasters.11 Ober Grombach (Fig. 41) is likewise divided into two principal parts, north and south, and has unimportant minor enclosures. The principal house (1) (above, 31) was a hall with porticus and pavilions; it faced, apparently, not a shrine but a limekiln, the view being interrupted by a small rectangular wooden building on unmortared walls whose purpose is unknown. No certain domestic building was found in the north part, where survival of footings was not so good, unless, perhaps, 3?, divided into small rooms, was something of the kind. More importantly, what may be a second house (2) lay across the line of the dividing wall, whose full length was not discovered. It may be a house; the purpose of four small rooms with independent hypocausts, if not domestic, is hard to imagine, and there was an entrance into the middle of what might be regarded as the hall. Enigmatic though the building is, its position in relation to the dividing wall may be compared to that of some houses in larger villas, for example Apethorpe (Northants.), Gentelles and Grivesnes (both Somme).12 Division was not always by a wall as at Bad Homburg (above, 44); a ditch would do, as at Welwyn-Dicket Mead (Herts.), where it is called a canal. 149

— Chapter Ten — Physical division of the yard is not the only line of descent from Ober-Grombach and Hoheneck. Lauffen am Neckar (Bad.-Württ.) has an almost rectilinear yard (Fig. 41) in which the principal house (1), a hall with porticus and pavilions, faces two habitable buildings. Of these, 2 began as a small building (8×6 m) in which was a hypocausted room about 3.5 m square; it is comparable to the hypocausted room and something else in the corner of Ober-Grombach.13 The other building (3) is described as a ‘granary with work- and living-space’ and space for crop storage and tools; yet, as is well said in the report, ‘fireplaces are certainly unusual in a granary, because the fire danger is great’.14 For good measure, the building has no sign of a chimney or of a staircase, yet it is alleged to be two-storeyed. These contradictions are resolved if 3 is thought of as a modest workhall; then the whole looks like a smaller version of the two blocks of buildings at Ober-Grombach, the main house being the most important element from the start but the small building 2 having similar comfort. It is easier to see how the fundamentally bipartite character of Burgweinting persisted in one form or another than to understand the kind of houseful the various buildings provided for. The point is borne out by the villa of FriedbergPfingstweide (Hessen) (Fig. 41), which exemplifies another mode of development having the principal buildings set parallel to but not opposite each other. Friedberg has, unusually for a German villa, a nearly square courtyard whose geometrical regularity is not matched by the positioning of the buildings within. Dominating the courtyard, but not placed quite centrally, is a most extraordinary principal house, made even more extraordinary by the conventional explanation of its elongated middle space (28×7 m) as a yard. It must surely be a hall, roofed continuously over the rear rooms and flanked by a porticus-and-pavilions in the usual way. At the front, two square garden(?) enclosures separated by a path are matched by two sets of three rooms, separated by a passage, at the rear. It is as though the rear rooms correspond to the two units of Downton and they share the adjoining hall. In the west garden is a square building which may be a temple. It is likely to have been common to the houseful, as was the bath building adjoining the east garden; hence, perhaps, the little square structure in the middle of the entrance path which, it may be, was a roof on four columns marking where it was permissible to cross from one household’s private space to another’s. This important house, the centrepiece of the yard, is canted slightly towards the entrance and lies parallel to the road, not the yard wall. It follows that the line of the road is significant. Dividing the yard in the proportion of 70:30, it corresponds approximately to the relative importance of three households, the two more important to the south, a minor one in the hall near the east entrance. Ober-Grombach yard is divided into two more nearly equal parts. Despite the clear inferiority of the north part of Friedberg, those living there were not all labourers. The hall (2) near the east entrance was a quite imposing building, finished externally on three sides with pilasters and internally by two flanking the opening from the porticus. In that respect it is like the hall at BocholtzVlengendaal, although its altogether different proportions appear to provide for two households, not a representational purpose. One other building (3) in this north part has some claims to be part domestic, the one abutting the yard wall, 150

— The Elements and Forms of Villa Complexes — with a porch and porticus-like front. The two narrow front rooms have not been explained, here or anywhere else, in terms of an agricultural purpose; their proportions and the narrowness (c. 2 m) of the porch prompt comparison with similar living spaces for craftsmen and labourers shown in the Plan of St Gall.15 Compared with Ober-Grombach, Friedberg shows a marked shift towards inequality among the various households comprising a kin-group. Another shift in that direction, though one less easy to evaluate precisely, appears at Olfermont (Haut-Rhin) (Fig. 41). The bipartite row-house [I] survives with added wings of row-type appearance; and beside it has been added in stages a yard of irregular plan, a small square enclosure, and a large workhall.

COURTYARDS OF GEOMETRICAL SHAPES The squarish shape of Friedberg is not common;16 an elongated yard is usual and takes many forms. It can be a trapezium, tapering from the entrance side towards the house, and a variant of this has successive flanking buildings stepped back to the same effect; or fan-shaped, broadening out from the entrance towards two or three houses; or rhomboid; or rectangular, entered in the middle of either a short side (the commoner) or a long side. These forms refer to the principal courtyard, which alone carries a social meaning; where there are flanking or outer yards they vary in position and shape. All geometrically shaped courtyards raise questions about social structure, especially about the power relations between the privileged who lived in the principal house and those who lived in the buildings around the yard. To understand them requires analysis of the principal house which, even at the head of a long courtyard lined with buildings, can be no more than a few simple rooms, or a sizeable residence enriched with mosaics and hypocausts, or a group of twenty or more rooms with no obvious coherence. Moreover, houses of any considerable size may be grouped around smaller courtyards of varied shape and consequently raise further problems of social development. Geometrical yards present other problems of a different kind. At what stage in the evolution of a farm was it decided to undertake the costly work of laying out a courtyard? What determined the size and shape of courtyard adopted to replace what was, obviously, a smaller establishment? To embark on such a change implies a confidence that the family or kin-group is stable for the foreseeable future, without any marked expansion or contraction, and that agricultural or other prosperity was assured. Then there are questions about how the large amount of building work was carried out and paid for, and although these two questions will only be touched on briefly, they present problems which have to be faced by anyone seeking to account for courtyard villas. The largest concentration of courtyard plans at present known is in Picardy and it will provide many of the examples in what follows.


— Chapter Ten —

TAPERING YARDS Le Mesge (Fig. 42) is a simple and comparatively uncomplicated example of a tapering courtyard which shows the essentials of the form. The focal point is the house closing the end of the courtyard, and this type of plan is not to be confused with a tapering yard like Mayen, which expresses a quite different relation between the three houses. Le Mesge has a row-house of quite modest size, its core much the same as Downton, with one two-room and lobby unit to south of the large middle room and an equalsized one of indeterminate form to north. Inner and outer courtyard walls have what looks like a gatehouse in the middle, showing that the house was approachable by a path 250 m long – a prospect designed to impress; but whether a less ceremonious approach was possible, to the rear open-ended porticus, for example, is uncertain. The sides taper fairly equally to give a symmetrical yard, as if neither range were of preponderant importance. Ecoust-Saint-Mien (Fig. 42) is another large villa like this but with one interesting difference; the gatehouses and approach path are aligned on a cross-wall. This implies that the house was divided like Maulévrier but was entered by a single doorway at the front of the porticus, from which the visitor turned either right or left, as at Chedworth (Fig. 74) or Manderscheid (Fig. 39). Sparsholt [I] (Fig. 68) may be a comparable British example of a tapering plan if the slight evidence of an outer yard is a guide. It has three buildings around the inner courtyard, which tapered symmetrically and implicitly focused on a house, the existence of which must be presumed in order to give the yard shape any point. The only building excavated outside suggests a tapering alignment on that side, as if the angle changed at the inner courtyard as it does at Clairy-Saulchoix (Somme). Yards tapering symmetrically are exceptional. Most taper more sharply on one side than the other, as do Fontaine-le-Sec (Somme) and Citernes (Somme), while Cromhall (Glos.) has a plan developing on these lines. Some taper on one side only so that the plan is like a truncated right-angled triangle, as at Mézières-enSanterre/La Croix Saint-Jacques (Somme). The intention must have been to emphasise the importance of one set of buildings rather than the other. A scale of importance is implied by the successive stepping-out of buildings on the east side of Cachy (Somme) (Fig. 42), whereas the west side shows no such differentiation. It is unlikely that such careful discrimination was needed for buildings which had a solely agricultural purpose. Tapering yards are rare in Germany. A very small one is Wiesbaden-Höfchen (Hessen) (Fig. 44), where two halls stand close together in much the same relation to one another as several row-houses and aisled houses do in British farmyards; Mansfield Woodhouse is an instance. One example, of slightly irregular outline, is NiederzierHambach 69 (Fig. 42); too few buildings were found to permit discussion. Mayen (Fig. 42) has an interesting variation in which the principal house is not at the head of the yard but to one side and canted towards it, so that it would be more prominent to anyone approaching past the little temple or shrine. British villas usually reveal an embryonic plan lacking a yard wall of masonry. Either one never existed or was not found or the yard was enclosed by a bank and ditch or a hedge. Mansfield Woodhouse (Notts.) (Fig. 42) shows the beginnings of a tapering plan, Cromhall a later stage of development, and there are several more.17 A minor variant in France 152

— The Elements and Forms of Villa Complexes —

Figure 42


— Chapter Ten —

Figure 42


— The Elements and Forms of Villa Complexes —

Figure 42 Yards not rectangular

has all the yard buildings aligned parallel to one another but stepping inwards progressively; Davenescourt (Somme) (Fig. 44) has four buildings set out like this on the east side of its 200 m long yard.


— Chapter Ten —

FAN-SHAPED YARDS A tapering yard drew the eye to the house at the far end to emphasise its commanding position and demonstrate where power lay. Fan-shaped yards created a different effect, one which is not always obvious. Darenth [F] (Kent) (Fig. 42) shows this well. Those approaching the villa saw from the entrance, which was some 180–200 m away, the prominent and impressive front elevation of the large principal house and, tucked away at its ends and largely masked by outbuildings, two smaller ones. To east was the gable-end of the best-appointed house on the site, whose external appearance may have proclaimed the fact by ornament or paint, and to west a similar building more like a workhall may have proclaimed its relative status in similar ways. Whatever the fan shape was meant to convey, it did not do so by specifically architectural means. Nevertheless, Darenth had a different social basis from Le Mesge, as would have become obvious to someone advancing down the yard; he was confronted by, first, a small square temple and then a nymphaeum which blocked a direct approach to the largest building. An appropriate offering made, the person proceeded to left or right and entered either of the flanking houses in the usual way, by a doorway in the middle of the long side. So what idea underlay the contradiction between the dominance of the main house and the denial of direct access to it? Clearly the conventional manor-house interpretation will not do. Perhaps the intention was to proclaim architecturally the importance of the kin-group through the main block, while preserving something of its basis in [I] through the two separate houses. Explanation is complicated because we see only the final phase of the plan, which embodies elements appropriate earlier but hard to change in altered circumstances: an inconvenience frequently arising from the long life of structures which were expensive to build or replace. In that light, Darenth [I] comprised a principal row-house which was one of the larger of its kind, comparable to Pulborough [I] (Sussex) (Fig. 45); a hall house (uncertain in detail) to west; and what may have been another hall – because its subdivisions look like insertions – to east. Allowing for the different main houses, it had a passing resemblance to Katzenbach [I] (Bad.-Württ.) (Fig. 74), where the alignments imply a fan-shaped courtyard; both villas provided for similar social relations. Subsequently Darenth developed in the same general way as some German hall houses, concentrating the service functions at the west end and improving the east end by subdivision and the improvement of mosaics. Most important of all, the balance of the house was shifted by the addition of a grand room (14.8×4.9 m) linking the east and main blocks, which had a middle opening about 1.45 m wide; it looks like the equivalent of the middle room of a row-house such as Newport (I.o.W.), and it may account for the lack of evidence of any such room at the middle of the main house. It is surprising to find a much humbler villa, Hambleden-Yewden Manor (Bucks.) (Fig. 42), presenting something of the same contradictions as Darenth and perhaps for the same reason, that alteration has confused an originally clear social statement. Phase [I] comprised three principal buildings: a small row-house (1) of stone18 and two halls (2 and 3) – to regard them as purely agricultural makes the yard plan pointless – which were possibly of timber; and the yard was enclosed by some impermanent material, a hedge or bank. Enlargement of the row-house took place 156

— The Elements and Forms of Villa Complexes — in [II], when, probably, hall 3 was rebuilt in good-quality masonry;19 and in [III] house 2 and the wall were rebuilt. But 2 [F] still had a good-sized hypocausted room, part of a domestic suite of three, and in earlier phases, before 1 had received its final additions, was a relatively more important element of the villa. Hambleden and Sparsholt resemble one another in having the same three components which, by their setting out in yards of contrasting shape, emphasise the principal house to different degrees – slightly in the former, strongly in the latter. Sparsholt, which began with a tapering yard, was realigned with a slight fan shape in order to give greater prominence than formerly to a secondary building. At Hambleden, where the original splaying out of the buildings made the principal house less of a focal point, the two secondary buildings declined to become very obviously subordinate to the principal house. Fan-shaped yards are rare on the continent. Only one has come to notice, Bargen im Hegau (Bad.-Württ.), where the yard widens from east to west.

SUB-TRIANGULAR COURTYARDS Sub-triangular courtyard is a shorthand form to describe a variety of courtyard distinguished by a right-angled triangle at the base and cut off at the top, and is adopted in preference to the more accurate but clumsy ‘truncated right-angled triangle courtyard’, there being no short, precise name to describe such a figure. The term is necessary to isolate a distinctive clustering of domestic rooms around the rightangled corner which distinguishes it from other tapering courtyards. The only fully analysed example, Rockbourne (Hants.)20 (Fig. 42), reveals the idea behind the adoption of such a plan. It began in II, the first Roman phase, with a hall and a row-house set well apart at right-angles. This potential triangular yard was abandoned in III in order to build one on a larger scale: the row-house was demolished; its successor, a larger one of two units and important middle room, was aligned with the hall; and traces of buildings, including a bath house, mark the beginnings of a new north-east wing. These changes are instructive. II comprises a house and workhall. In III the growth of a much larger villa was foreseen and the new plan demanded, not simply the addition of a house adjoining the workhall, but the preservation of their distance apart. There may have been a utilitarian reason, that here was the way in from the fields, but it is interesting that the workhall was never incorporated in an inferior range; like a German workhall, it always retained the dignity of its intermediate status. The new wing also incorporated a small bath building, and it stood clear of the houses because it served both wings. In IV the south-east wing is established with an aisled building containing some livingrooms; this was the beginning of a range of work buildings. Not until VI was the northeast wing rebuilt in intelligible form as a row-house of five rooms, and its greater width compared with the adjoining range is matched by the corresponding range at Chedworth (Glos.) and elsewhere.21 A difference of this kind must be another point in the code governing social differentiation because it is applied in villas of various types. Few other villas are definitely known to have been laid out with a sub-triangular yard. Bignor IIA (Sussex) (Fig. 42) is one and Liestal-Munzach (Switz.) (Fig. 76), 157

— Chapter Ten — though greatly altered, was perhaps another originally. One or two incompletely excavated villas, for example Atworth (Wilts.), look as though they may belong to this category of plan.

RHOMBOIDAL COURTYARDS AND THE PROBLEM OF HAMBACH 512 Why one courtyard plan should have been preferred to another is often hard to understand. This is especially true of types which appear to convey a similar message about the social relations through the layout of buildings in the yard. Thus the concentration of the purely domestic buildings of a villa, with the most important workhall nearby, which is the point of the triangular plan, reappears in two rhomboidal courtyards in Germany. Hambach 512 [I] had a small and, for Germany, quite unusually narrow hall (5 m) which is almost a row-house; with rooms at both ends and porticus-and-pavilions it represents much the same stage of development as Frocester Court I or BrewoodEngleton I. It faces down a long courtyard from one corner. Nearby and a little in front a workhall stands at right-angles to it. The grouping of buildings in the corner resembles that at Rockbourne, but whereas the latter is an L-shaped block open to the yard, the Hambach workhall turns its back on it to create, in effect, a small inner courtyard.22 This beginning of withdrawal from the general activity of the farmyard signals that new social relations were emerging, bringing a sharper distinction than hitherto between members of the kin-group. A number of German villas were enlarged in a rather surprising way by removal of part of the enclosure wall and building out over it. Hambach 512 [I] is an instance of this kind of change; close to the small hall house and aligned with it was added in [II] what looks like a second, larger hall, also with a porticus-and-pavilions. It was, in this phase, an independent house and its construction may have required demolition of a stretch of wall long enough to destroy whatever purpose and meaning the yard originally had, yet no new enclosing wall was found, or, perhaps, built. Physical limitations alone do not account for the chosen method of enlargement; they would not have precluded an L-plan or a change of the kind found between Rockbourne II and III, involving removal of a building. For some reason it was essential to align the new hall with the old, as was done at Leutersdorf II (Fig. 45), for example, but here the two were only joined together in [III]. The new hall seems to have been part of a large extension, to judge by a building parallel to it and outside the courtyard, and it may be that so extensive an enlargement was only possible by creating, in effect, a second yard.23 There is an alternative interpretation, one suggested by the curious placing of two wells, one on either side of the yard wall only about 10 m in front of the new hall. This implies that the wall still stood at least as far as that point, and the plan gives no reason to suppose it stopped there. It may have continued and abutted the porticus, like the wall of the divided yard at Bad Homburg (Fig. 9); and in that case the siting of the new hall astride the wall served the same symbolic purpose of uniting two landholdings as did the similarly placed houses at Marshfield and Condé-Folie (Somme) (Fig. 66). 158

— The Elements and Forms of Villa Complexes — To get back to courtyards, another Hambach site, 69, also has a house in one corner (Fig. 42). Unfortunately, no workhall is identifiable in its vicinity, that being the part of the enclosed area where least evidence remained. It reveals what 512 did not, the entrance, which is markedly off the yard axis, as is also true for Hambach 516. This displacement from what appears the obvious place does not put it opposite the principal house, only more nearly so, hence its siting may respond to an informal division of the yard akin to that at Bruchsal/Ober Grombach or Friedberg-Pfingstweide, an idea reinforced by the distribution of buildings.24 A third site, Hambach 403, was too incomplete to be certain it was rhomboidal, but the little hall house and a large workhall were placed in one corner at right-angles, facing the yard in the manner of Norton Disney (Lines.) or Mansfield Woodhouse (Notts.). The comparisons, as with so many aspects of villas, are international.

LONG RECTANGULAR COURTYARDS Fan-shaped, triangular and rhomboidal courtyards are not recorded in France. There courtyards are usually rectilinear, and in the Somme basin are mostly elongated. Few show any sign of being the result of gradual growth except, sometimes, for enlargement of the principal house; indeed, the form itself argues that buildings were expected to line the yard on both sides. Among the few villas of this kind where more than one phase is obvious in the yard are Meckel (RhldPf.) and Liestal-Münzach (Switz.). The surprising number of large villas in Picardy with long rectangular courtyards includes two which are remarkable for being in the same commune of Warfusée-Abancourt. At first sight they look much the same. Warfusée-Sud has a house some 40 m long which is quite large but not in relation to the 300 m long courtyard in front. There is no sign of a second house. A quarter of the way down the yard is a square building looking like a gatehouse, governing access to an inner court. Warfusée-Nord (Fig. 43) is shorter, perhaps only half the size of -Sud, with walls or a terrace separating the main yard from the smaller court in front of the house. The wall is interrupted by two gatehouses, not one, and the house, very large in relation to the courtyard, 25 has in the middle a cross-wall dividing it into two equal parts: Maulévrier again on a larger scale. A quite large houseful can be inferred, comprising several households. Warfusée-Sud is also large enough for two or more households but their separateness was much less. The agricultural workings of the two villas must have been similar; the groups of people occupying the principal houses were organised rather differently. Picardy has eight of these very long rectangular courtyards.26 Of the five where the detail of the house is adequate, only Mézières-en-Santerre/Le Ziep is certainly and Athies possibly divided like Warfusée-Nord; Estrées-sur-Noye definitely not; and Ribemont-sur-Ancre has a unit-system house like Lamargelle-Versingue 2 (Côte-d’Or). Anything like a big country house is absent from these vast establishments, and that is true of the classic villa of this kind and the largest of them all, Anthée (Belg.) 159

— Chapter Ten —

Figure 43


— The Elements and Forms of Villa Complexes —

Figure 43 Rectangular courtyards

(Figs 75, 76). Not only are there no country houses but, as Agache’s plans reveal, the houses at the head of these impressive ranges of buildings vary greatly in size. The Ribemont yard, not dissimilar in overall size from Mézières, has a house only half the size, one which, if excavated in Britain without the other buildings, would be regarded as a quite modest farming establishment; and for good measure someone would contrive an estate to match. That is equally true of Marchelepot (Fig. 43), where two three-cell row-houses – a major and a minor one – dominate what seems to be a quite large yard. These villas form a striking contrast to the large courtyards around Trier, where Blankenheim has a sophisticated house and Fliessem (Fig. 43) is dominated by a subtle and complex architectural composition, while the different but very impressive villa of Oberweis might be a third if its yard were better known.27


— Chapter Ten — Britain has nothing comparable to the big continental courtyards, the nearest approach being Pitney (Som.), which may have had some agriculturally related buildings as well as domestic ones. It is 84 m long; Picardy has nothing as small as this. Only Chedworth (Fig. 74) has the characteristic inner court (incompletely explored); it also has part of what may have been a long courtyard which, unlike those in most French villas, has a continuous line of rooms of equal width, all domestic. North Wraxall (Wilts.) (Fig. 44) is a small villa where some order seems to have been imposed on what began as dispersed buildings. Few elongated courtyards show signs of gradual growth. Meckel is one; the long south wall of the courtyard breaks alignment and so, twice, does the wall separating inner and outer courts; and the buildings flanking the inner court relate poorly to the main house and the perimeter wall. This means that a long rectangular yard was added to and in part replaced a looser arrangement of buildings on a different alignment, and the house was rebuilt parallel with the new inner courtyard wall. Levroux-Trégonce (Fig. 76) underwent complete rebuilding of the outer yards, which left the wings of the house misaligned with the two ranges of buildings, and a partial, somewhat less drastic reordering of the yard can be inferred at Liestal-Münzach. A commoner indication that the outer or working part of the courtyard has been rebuilt is provided by the misalignment of the house and the wall of the inner court. Oberentfelden (Switz.) (Fig. 76) and Belleuse (Somme) (Fig. 44) have this characteristic and so, too, very faintly, does Estrées-sur-Noye. Davenescort (Fig. 44) illustrates the placing of the house to one side of the courtyard it faces. This not uncommon characteristic might be taken as another sign of alteration if there were any confirmatory detail in the several instances, such as a hint that some structure has been removed from the blank space; but nothing of the kind has been noticed.

DOMESTIC COURTYARDS AND COURTYARD HOUSES Some British villas have courtyards corresponding to the inner courts of the longcourtyard villas just described, without any sign of the much larger working yard. A well-known example is Sudeley-Spoonley Wood (Figs. 44, 70), which comprises three apartments, service rooms, a workhall and baths arranged on three sides of a court. It shows signs of gradual growth: the addition of the smaller part of the double room in the middle, because the opening is not quite symmetrical to the bigger part; the addition of the porticus-and-pavilions to the north range, because one wall of the west pavilion abuts what looks like an addition to a double-ended hall;28 and alteration around the baths. Other complications must certainly remain undetected. Spoonley Wood [F] is a unit-system house comprising three apartments, one in the south wing (the best appointed), two in the main or east range, and an imposing workhall to north. Hence it differs in three respects from continental inner courts: by comprising in a close-knit group several buildings which are usually separate in other countries; by its not being planned from the first in the form it takes; and by not having a regularly planned outer yard, anything like the Picardy courtyards being precluded by topography. 162

— The Elements and Forms of Villa Complexes — The term ‘courtyard villa’ has long been used in Britain to denote one having domestic rooms on three sides – something which, while it finds some continental analogies, is quite unlike the generality of establishments in Picardy. Baths are included as a domestic amenity, as are workhalls of the Spoonley Wood kind, providing living as well as working space; also a room serving the purposes of private justice and administration – a basilica – with its appendages.29 Courtyard shapes vary, one or two such as Box being rectilinear, others like North Leigh (Fig. 42) and Bignor [F] (Fig. 42) trapezoidal. No British courtyard of a single building phase has come to light. Box I is a rowhouse with two short wings and continuous inner and outer porticuses which evolved into a three-sided courtyard house. It is unique in that the wings deserve that name; because they abut the main range the corner rooms are likely to be original and not, as in most of the analogues, an infilling between blocks which merely, or nearly, touched. Some comparable houses began as separate buildings on three sides of a court, Chedworth (Fig. 74) being the classic example;30 North Leigh I (Oxon.) had two buildings in line, with little to show more existed on the other sides of the yard. Box, Chedworth and the rest are all stages in formalising what begins as a group of detached buildings. Some villas did not become a unified whole: Badbury (Wilts.) was laid out on a large scale and may not have developed much; Folkestone [I] (Kent) was laid out with the north and west blocks too far apart to be linked later, and since no workhall has been found, the intended shape of the courtyard is unknown.31 These various ways of relating separate houses, and the extent to which they were unified, correspond to different degrees of subordination of one household or group of households to another. Continental parallels are not common but are thought-provoking. Marboué-Mienne (Eure-et-Loir) (Fig. 44), the biggest of them, has a courtyard 200 m long, as long as all but the largest of the Picardy yards and longer than Bignor overall, yet this is only the equivalent of an inner court; nearby is an agricultural yard some two and a half times as big. Little can be deduced from the schematic plan, although, since the site was excavated by a distinguished engineer of Ponts et Chaussées, it is probably accurate. It has several unexpected features for so large and splendidly appointed a villa: the south range is virtually detached from the west and north wings which form a continuous L-shaped range; the room with the three-sided apse is well off axis; and a line from the gatehouse parallel to the north range meets the cross-wall between rooms 8/9 and 10. That wall breaks the porticus and separates two threecell units which comprised, no doubt together with parts of the north and south ranges, phase [I]. Interpreted like that, it is analogous to Chedworth I (though rectilinear, not trapezial) and, in effect, a larger version of Maulévrier. Why the structurally detached principal house [I] should be some way off axis is not clear but in this respect Mienne is like Warfusée-Sud and several other Picardy villas.32 Social change caused the house to be lengthened to the north and provided with a grand middle room which effectively destroyed one of the original two apartments. To set it on the axis of the courtyard or opposite the gatehouse, either of which might be expected in a rebuilding, would have entailed recasting the house altogether, and objections to that are more likely to have been for social reasons than on grounds of cost. About the rest little need be said. Rognée (Belg.) has an unusual near-square 163

— Chapter Ten —

Figure 44


— The Elements and Forms of Villa Complexes —

Figure 44 Rectangular courtyards with hints of rebuilding

courtyard (72×66 m) and is hard to analyse. A villa with a more British look is Châtillon-sur-Seiche/La Guyomerais II, which in II possibly and III certainly is a small row-house at the head of a courtyard having, on the east (and only excavated) side, another row-house. That is enough to relate it to most other examples of the type. 165

— Chapter Ten —

PARALLEL RANGES AND DETACHED FAÇADES A few villas comprise two buildings linked by a porticus. Two were explored in Brittany in the 1830s, the better-reported being Plomelin-Perennou (Fin.) (Fig. 45). The two wings to east and west are little row-houses like Drax (Yorks.), two elongated rooms open off the porticus and there are two small rooms between them. There are differences between the east and west sides: the east rooms are bigger;33 in the former, the end room is closed off from the porticus, in the latter a doorway closes off the end room and lobby, and the larger (west) of the long rooms has the wider doorway. It is the familiar situation of two not quite equal apartments. The second example, Arradon-Lodo (Morbihan) (Fig. 45), is much larger and the differences between the east and west blocks are greater, as are the doubts about its dimensions.34 To west is a domestic block including four rooms with hypocausts, to east a hall-like room surrounded on three sides by smaller rooms, one with a hypocaust. This fairly sharp division of functions matches the very different architectural forms it took in hall houses and other types, and may find a pale reflection in the different-sized houses at Perennou. One British villa comparable to these is Great Witcombe (Glos.). Built on a fairly steep slope in a way more characteristic of villas in the Moselle, it comprises a porticus connecting two wings which seem to have been somewhat different from the beginning, although the way the rooms were used has never been worked out. Later the customary separation of functions took place, the west wing incorporating the baths and the east wing some service functions. Thus far it is much like Arradon-Lodo, but it also had, facing the view in the middle of the porticus, a large room which in I was rectangular and was later rebuilt as an octagon with an apsidal end. On the opposite side of the porticus from it was a quite elaborate shrine. None of the villas just described has any other buildings, so far as is known. Another British villa, Winterton (Lines.) (Fig. 45) combines a detached main (west) range having some resemblance to Perrenou with parallel aisled houses in which the ancillary occupations of farming were carried on. This suggests that the three related villas may also have work-related buildings waiting to be discovered. The west range G of Winterton (Fig. 45) is a curious blend of features. In I the porticus has pavilions of considerable size, each comprising few rooms but planned quite differently, so in that respect it resembles Arradon rather than Plomelin-Perennou. The north pavilion is slightly larger, corresponding to the larger and nearer of the two detached aisled houses. In IA the north pavilion was doubled in size and a small square room – a shrine? – was added in the middle of the porticus and in II it was replaced by an apsidal room facing out from the yard. These changes are the first sign that the porticus was turned to face outwards instead of inwards, without inhibiting the improvement of the two aisled houses. Change was complete in III when the apsidal room replaced the ?shrine, the pavilions were reduced to a single large room, and the whole acquired a purely recreational or representational function. The creation of a façade lacking any living-room represents a unification of the social structure; no longer do the two households share accommodation there, it is given over entirely to common purposes. Unification, though, was not on equal 166

— The Elements and Forms of Villa Complexes — terms: while the smaller house (B) underwent improvement to the last, the larger, D, is closer to the north pavilion which had a pilastered east front and always provided a stately entrance. Phase III also represents a change of attitude towards the villa as a farm and comes as near to a country-house mentality as was possible without building a completely new house – a villa in the modern sense, as Steiner put it.35 A near-parallel is the facade of Leutersdorf III (Fig. 45), which, however, incorporates some apparently domestic rooms, and another is La ChapelleVaupelteigne (Yonne). An unusual combination of parallel ranges and a tapering outer yard occurs at Orton Longueville-Orton Hall Farm (Fig. 45). Aisled buildings predominate, with four at least in the outer yard and another in the inner enclosed court, which has an imposing gateway. On the other side of the court is what looks like a single-ended hall. The innermost aisled building must, merely from its position, be a house, and if it accommodated two families, as suggested earlier, with a third in the house opposite, the social core will have differed in degree rather than kind from what can be inferred in many other villas: Ober-Grombach, Sparsholt and Rockbourne III–V, for example. The whole villa plan is without a close parallel, yet close inspection shows it is a combination of elements familiar in different contexts. Parallel ranges of more or less equal importance are rare. A French example, Verneuil-en-Halatte (Oise), whose beginnings are obscure, seems nevertheless to have taken this form from the first. In I, a phase approximating to the round houses and buildings A and C I at Winterton, the parallel ranges, north and south, are only revealed by the lines of enclosures and a few six-post buildings which were probably dwellings; and in II a ridge-post hall appears in the north range. By III an inner courtyard of rhomboidal shape existed, closed off from the working yard successively by a long range having a general resemblance to Winterton G I, a wall and a long ornamental pool like the one at Aylesford-Eccles (Kent). The pool made a direct approach to the middle of the range impossible, from which it may be inferred that the end rooms, though smaller than those at Winterton, were linked to the nearer of the two main ranges. By VIII the pool had disappeared and rooms, including baths, had been built on three sides of the inner courtyard, which was closed off, after the demolition of the phase III wall, by a porticus. Even without detailed information – for only a brief note is available at present – the process appears to be similar to what happened at Winterton but more complicated. The various ways in which detached facades, parallel ranges and ornamental pools were combined to express particular social relations show that a stock of architectural features and their iconic associations was common to the western provinces of the Empire. How that stock was drawn upon, and how the ideas were diffused, are questions not easily answered.

LINEAR VILLAS One way of setting out the several buildings that went to make up a villa was to put them, or the principal ones, in line. Houses like Halstock II (Fig. 46), North Leigh I (Fig. 42) or Saint-Germain-lès-Corbeil (Fig. 74), where the two principal houses are 167

— Chapter Ten —

Figure 45


— The Elements and Forms of Villa Complexes —

Figure 45 Parallel ranges and detached façades


— Chapter Ten — aligned, have the most rudimentary linear form. All probably had other buildings of timber and all developed courtyards of some form, unlike Bocholtz-Vlengendaal (Fig. 46), which has the house, a storeyed granary and a workhall in line.36 JemelleNeufchâteau (Fig. 46) shows the elements of this kind of plan unified. Two of these elements occur in Maillen-Al Sauvenière (Fig. 46), where a hall house developing row-house characteristics is joined by a porticus to an otherwise detached workhall lying to east of it which was certainly a building of quite high status; it has a pilastered entrance37 and the bath block, though connected only to the principal house, is positioned to serve the workhall too. In the same commune is a larger example of the same setting-out: Maillen-Ronchinne, which has a much more complicated plan of several phases, nevertheless falls into two distinct parts joined by a long porticus. One is a well-appointed domestic block with heated rooms and baths;

Figure 46


— The Elements and Forms of Villa Complexes — the other to the east is a workhall of which little was recovered except the porticus with the porch typical of buildings of that kind. Examples of this type are widely scattered. In the west of England, MontacuteHall Hill provides a further example of three buildings in line – a row-house and two halls all linked by a porticus; and at the other end of Europe Budapest III-Testvérhegy also has three buildings, all apparently domestic and just as irregularly set out.

RECTANGULAR FARMYARDS In those countries where small villas are common, principally Britain and Germany, rectangular yards are comparatively rare; Newel’s almost perfectly symmetrical plan is very exceptional and Hungen-Bellersheim (Hessen) (Fig. 43) tapers insignificantly. It has six buildings set out parallel to the yard walls except for a little square shrine or well-house (2). The largest building (4) is the principal house, 5 a secondary one comparable to Lauffen am Neckar 3; 6 is said to be a tower granary and 1, by its position facing the entrance, may be a little temple.38 It may be the only one of its kind known in Germany but in Brittany a somewhat similar farmyard was discovered at Malguenac-Guilly (Morbihan). What are described as entrenchments of earth and stone surrounded it on all sides.39 Its six buildings are smaller and more subdivided than those of Bellersheim, and it is characteristic of villa differences in France and Germany that Guilly should have as the principal building a row-house rather like Shipham-Star II and Bellersheim a hall or hall-derivative. What makes these yards different is the lack of a house of conventional plan and appearance, whether hall or row-house; none of the buildings has a porticus, not to mention pavilions.

CONCLUSION Villas were set out in an enormous variety of ways, those chosen for analysis here being only the more obvious ones. Perhaps the approach taken here will permit other villas which at first sight do not show any underlying principle behind the disposition of buildings to reveal a social and not just an economic significance. Questions are also raised by the scale of rebuilding of the larger courtyard villas, even though the answers lie outside the scope of this book: questions about where all the inhabitants came from, how the resources to rebuild were assembled, and how long they took to complete. Regional differences are remarkable too. What factors produced the remarkably uniform geometrical plans of the Somme basin and why did they apparently not operate in a prosperous district like Trier? These historical questions may be tackled in a different way if the component buildings of a villa are seen to be fairly uniform in function and variable in the number of each category, and the manner of their setting out is recognised to have social and economic implications.





he title may prompt a question: why lump together an administrative function, an architectural form and a standard of living? The justification for so doing is that they are closely linked insofar as all three are of essentially Roman rather than provincial origin and inspiration. A palace is a seat of government, whether of the emperor or the governor of a province, and the administrative functions it provided for were much the same in any part of the Empire; it was occupied by someone to whom luxury came naturally by virtue of position and wealth and to whose office Italian forms of planning were appropriate to cater for its bureaucratic and social complexity. Imperial palaces were few and did not provide for emperors who travelled widely throughout the Empire, of whom Hadrian, always on the move, is an extreme example. It seems not to be clear exactly where such an emperor and his retinue, which must have amounted to a considerable number of people, stayed when there was no town or military garrison to receive them; one appropriate residence, a provincial governor’s headquarters, was apparently never used,1 no doubt to avoid administrative disruption. Some emperors may always have travelled in military style like a small army on the march but, if not, it is difficult to believe that they could rely exclusively on official and military resources. Still less likely are they to have used the staging-posts (mansiones) provided for official travellers, which are completely lacking in a suitable sequence of rooms; furthermore, ‘There is no clear and positive evidence that chains of buildings distinct from mere mansiones, and designed essentially for imperial use, existed in the second or third centuries’.2 On general grounds it seems possible that they were accommodated in great houses, much as English monarchs in the early modern period from time to time descended on their richest subjects, and in the late Empire there were in several provinces a number of houses of noble size and splendour which might fitly have welcomed the imperial retinue. Private hospitality on the ruinous scale of Elizabethan England was common under the late Republic and early Empire3 and appears to have continued for long afterwards, and, although little or none of the evidence appears to relate to the northern provinces, practice was presumably similar there. The plan of the house at Trier rebuilt by Victorinus in the 260s, when he was an officer of the Imperial Guard under the Gallic emperor Postumus, is not known apart from its having a peristyle 172

— Palaces, Peristyle Houses and Luxury Villas — and a large mosaic-floored room, but a house in the insula to the south has rooms of a size and quality that can reasonably be called palatial despite uncertainty about their exact function;4 their arrangement around open courts is reminiscent of Nennig or Haccourt II. However problematic in detail Victorinus’ residence may be, it raises the question of how far palatial requirements influenced the planning of villas.

PALACES ‘Palace’ in its primary meaning denotes the residence of a sovereign or other ruler and, by extension, a palatial private house. A palace in the first, formal sense might be excluded from this book yet, as the second sense is sometimes used interchangeably with the term ‘luxury villa’, it will be useful to discuss three that correspond to the primary definition insofar as they assist the understanding of villas. The concept of a combined seat of government and residence suitable for official entertaining demands state rooms, rooms, that is, of a size suitable for the reception and entertainment of lesser rulers, ambassadors and legates, all accompanied by attendants and received by corresponding or greater numbers of persons in the imperial entourage. All of these would require space appropriate to their dignity, with a distance between them and the emperor or his provincial representative, and the rooms provided required stately access. Rooms of this kind were larger than any required for the kind of entertaining done by private citizens, however eminent, or by kin-groups, however large or powerful. Government also required that many persons be able to present petitions, conduct legal business at the highest level, and hear legislative acts proclaimed; hence, for the private citizens involved, suitable access separate from the corridors and entrances used by the governor and his officials had to be provided. These considerations will be used to interpret palaces, especially two early ones which make the essentials clear. Not surprisingly, Domitian’s residence on the Palatine Hill at Rome, the Domus Augustana or Flavian Palace (Fig. 47), the original Palatium, shows most clearly the official attributes of a palace. Dominating the north side of a peristyle is an audience hall or throne room, entered on the west from the basilica and on the east from both the lararium or chapel and its anteroom. The basilica – the seat of judgement – has independent access from the outside for litigants and lawyers, who no doubt reached their places via the north porticus or a doorway at the north end of the west porticus, while on the south side were the entrances used by the emperor and his officials. In one respect basilica and audience hall make an interesting contrast. The functional orientation of the former is clear: the literal seat of justice was in the apse, facing litigants, petitioners and lawyers. Not so the audience hall, which has seven doorways so placed as virtually to preclude a throne against any wall; the apparently obvious place in the niche on the south side stands between two doorways and provides a far less imposing setting than the apse in the basilica. From this it can be concluded that the throne was freestanding. Opposite the audience hall is the triclinium, a banqueting hall for state receptions. As for the rest, the west side was taken up by an entrance-hall having four doorways alternating with apsidal recesses; this was flanked on each side by two apsidal173

— Chapter Eleven — ended flanking rooms, the whole forming a brilliant display of curvilinear shapes. On the east side was a much wider range having, in the middle, a large hall almost entirely open on two sides to give a view of the two peristyle courts; at the ends, two bowended rooms intended to impress; and between, eight small or very small service rooms.5 Everything about this court except the service rooms was designed by its

Figure 47


— Palaces, Peristyle Houses and Luxury Villas —

FIGURE 47 Palaces


— Chapter Eleven — architectural subtlety and richness of ornament to over-awe those coming into the imperial presence. Such state would hardly have been appropriate to the palace of a client king or provincial governor, who, nevertheless, needed both the suite of rooms required by his official function and a residence in keeping with his standing. One of the few buildings in this category known with any completeness is the palace at Fishbourne (Sussex) (Fig. 47),6 the exact date of which does not matter for the present purpose, nor does the precise status of the owner, whether client king or some other ruler. Although not all the state rooms of the Flavian Palace are paralleled in what has so far been excavated, the number, size and interrelation of those already known are truly palatial and quite unlike anything to be found in the greatest villas. In the east wing was an aisled entrance-hall flanked on each side by five small rooms. The east end room on each side was perhaps reserved for porters who asked their business of those arriving, and for servants who conducted people to appropriate parts of the palace. In the hall officials waited to meet those on government business of every kind and in the small rooms or cubicles they or others of a superior grade discussed the appropriate course of action, as, no doubt, did their counterparts in the apsidal recesses of the Flavian Palace. Differences between Roman and provincial society may account for the different architectural treatments: sophisticated curvilinear shapes for grandees used to such things, simpler but more imposing rooms for those adjusting to Roman ways. All who entered saw in the fountain at the inner end of the hall a symbolic reminder not to trespass further. Directly opposite on the far side of the peristyle court was a smaller apsidal room that looks, from its shape and somewhat smaller size than the other state rooms, like the basilica but from its position was probably the audience hall or chamber. A statue in front of the entrance ensured that those approaching had to break their progress, however stately; it ensured that nobody could go boldly across the garden court and straight up the steps to confront the governor or ruler. On balance the basilica is more likely to have been the aisled hall at the north-east corner. Those seeking justice either went straight to the imposing portico fronting it or were directed from the entrance-hall along the east side of the palace, while the governor and his retinue entered from the opposite end. Between the entrance-hall and the basilica are two colonnaded courts in front of a range of the smallest and most inferior rooms in the palace; they must be what were referred to in early modern England as officers’ (officials’) lodgings, and their plan has been revised to conform with the principles of row-house planning (Fig. 47). In the London palace, about which little can be said, a similar block of rooms served the same purpose though differing in detail. It has at the rear a porticus which is a true corridor, designed, no doubt, to keep minor officials out of sight as far as possible and away from the more dignified parts of the palace. As for the rest of Fishbourne, the five or six apartments into which the north and west wings were divided were presumably occupied, at least in part, by counsellors, advisers and officials, with one or two reserved for visiting dignitaries. They need closer analysis than can be attempted here. A reason for placing the audience hall so far from the basilica may be sought in the vastly different societies over which the respective rulers presided: whereas only 176

— Palaces, Peristyle Houses and Luxury Villas — those of high rank are likely to have entered the Flavian Palace, some greater separation of a local ruler from ‘the dependants of the estate and . . . low-ranking clients’7 may have been thought desirable. As for the state dining-room, it must be supposed on this reasoning to have been in the south wing, flanked by the private apartments.8 A third palace, Budapest-Aquincum (Fig. 47), was built for a different kind of person, the governor of East Pannonia, and although the same governmental functions were performed here as at Fishbourne, the fact that it was the headquarters of a military command sited on an island in the Danube necessarily modified the principal official buildings and produced a close grouping in place of spacious formality. But first, where was it entered? Not from the east, by the grand porticus-with-pavilions front overlooking the river and proclaiming to the Sarmatians the power of Rome. To north the baths preclude entry, and to west the only possible point is the unlikely one provided by the long corridor 26. By elimination, the freestanding structure to the south, linked to the main building only by a vaulted passage, must be the entrance-hall where those on business were received by officials in the two rooms 19 and 79 flanking an octagonal room (20), perhaps containing the statue of an emperor. Some formal likeness between this building and the triclinium range of the Flavian Palace may be deliberate. From there some would proceed to the basilica, which was probably room 25,9 as near to the entrance as was possible where no external porticus existed as at Rome or Fishbourne. Others would advance further to the audience hall (5), approaching it from the south corridor (4). How the other rooms of this very grand front were used is problematic. Rooms 2 and 3 (and 1?) were perhaps for meetings about government business, while the intercommunicating rooms 6 and 7 appear to have formed, with 9, 8 and 16, the legate’s private quarters adjoining baths of a sumptuous quality appropriate to a household of exalted status. Finally, in the south range are two mirror-image blocks of rooms (Fig. 47) which may have been officials’ lodgings comparable to those at Fishbourne.10 On the opposite side of the porticus a large room (77) with hypocaust appears, from its position, to have been linked to them as a common hall or a dining-room for special occasions. Other governors’ palaces vary in the relative importance of the civil and military aspects discernible in their planning. The palace of the Dux Ripae at Dura-Europos separated the two functions sharply. As at Aquincum (and London) the principal rooms faced a river – here, the Euphrates – and the rooms around the peristyle behind them have, in plan, an austere, barrack-like appearance, with only the governor’s public dining-room as a reminder of his civil functions. These were conducted, otherwise, around a second peristyle, one side of which was occupied only by a plain, rectangular building where administrative and judicial business is plausibly supposed to have been transacted. The whole has an air of military austerity quite unlike the monumental scale of its counterpart at Köln. Here again is a porticuswith-pavilions front, facing the Rhine this time and nearby and to the rear is an aisled basilica rivalling in size the famous imperial audience hall called the ‘basilica’ at Trier. For the immediate purpose the interest of this building, which cannot be reconstructed as a functional palace, lies in the large apsidal pavilions – the apse being inscribed within a square end – which are open towards the river,11 and which resemble those at Aquincum. 177

— Chapter Eleven —

VILLAS AS SEATS OF LORDSHIP It is a commonplace that large and mostly late villas were the residences of magnates who controlled great estates and wielded power over numerous clients and tenants. To what extent can the exercise of such power be inferred from villa plans? One test is the extent to which the form of significant rooms in palaces, and particularly sequences of rooms, are observable below government level. Relevant to this question is another: how far did an emperor or provincial governor concern himself with the building activities of the richest and highest-ranking holders of civil office? In other societies government was much concerned about the kind of house a potential contender for state power might put up: in England, for example, the building of Thornbury Castle may have been a factor contributing to the execution of Edward Stafford, Duke of Buckingham, in 1521; and in the late 1660s the building of a great house reinforced suspicion of the Lord Chancellor Clarendon’s intentions and contributed to his dismissal and exile. It seems that the size and grandeur of private houses rarely caused political difficulties in Roman times12 but it would be interesting to consider, in those rare cases, exactly what the objections were. Recently it has been argued persuasively that the medium-sized villa at Box (Wilts.) (Fig. 48), through the addition of a large apsidal room (10/25/26) at a corner of the courtyard, became recognisably a seat of power.13 The room is compared with similar ones ranging from the Flavian Palace to Winterton (Lines.) (Fig. 45) and argued to have had ‘at least a partly “public” function’;14 and further, that ‘the distinction between “private” (dining) and “public” (audience) rooms’ is unlikely to have been sharp in such a ‘comfortable, but not exceptional, rural mansion’.15 But what is meant exactly by private and public in the context of villas, and in particular of villas owned by a kin-group? The word ‘public’ in this context means essentially private jurisdiction, that conducted below state level: the settlement of disputes between members of a kingroup, dependants and tenants, and the punishment of misdemeanours; and decisions about tenancies and inheritance, especially those relating to shares in kin-group property. Public in the sense of an audience room implies a well-defined social hierarchy in which one person has power over others in more than a judicial situation, power to hear petitions and grant requests; also to receive the homage of inferiors, in the sense of their public acceptance of jurisdiction and their acknowledgement of an overlord. Entertainment can be either public, to display solidarity with equals and on rare occasions even to receive superiors, or private, which includes the feasts given by the non-hereditary head for the other members of a kin-group, at which some representational acts might be performed. In the sense of receiving friends, some element of influencing them in a common cause, political or economic, might well be involved; the boundary between private and public is often blurred. To be recognisably a seat of lordship a villa must have a room whose suitability for public acts is clear. Where direct access to an apsidal room from the outside is easy, as at Box from the east porticus 21, suitability is established and the case for a public and particularly a judicial function is strong. Where it is not, private entertainment is the more likely purpose. On these grounds a separation of function 178

— Palaces, Peristyle Houses and Luxury Villas — is likely at Box, with the middle room (6) in the main range, which has a mosaic, being a dining-room: impressive when private entertaining was tinged with public concerns and luxurious on purely private occasions. Bignor (Sussex) (Fig. 48) has a similar conjunction of rooms in the north range; the apsidal room, the presumptive basilica/audience room,16 is entered independently on the west side, and the triclinium, serving both dining and (with its hexagonal piscina) representational functions, is nearby (though the latter might have been expected in the west range – a problem here). Easy access from the outside is one important point telling against the suggestion17 that the apsidal room 12 added to Winterton G II (Fig. 45) is a justice room; another is the lack of the small ancillary rooms needed for that purpose. Instances of the conjunction of two rooms capable of serving the distinct functions performed by their counterparts in buildings at the highest level make the lordship interpretation of the villa credible. Problems arise when there is only one important room. The claim, apparently very reasonable, that hospitality and justice are likely to have been dispensed in one room in all but the grandest villas, is difficult to accept when the room in question is, like the one at Winterton, in the middle of the house. If a villa was large enough to have a class of dependent persons sufficiently numerous for a justice room to be needed, that implies marked social stratification. In that kind of society those privileged to use the principal room on representational or social occasions are unlikely to have welcomed at other times the presence of their social inferiors on court business, and for that reason the exercise of private justice can only be argued when the appropriate room is largely separated from the rest of the house. Several Iberian peristyle villas show the conjunction of justice room and diningroom with more or less probability: Almenara de Adaja (Valladolid) (Fig. 48), for example, has in the west range an apsidal room interpreted as the dining-room, and in the north range a larger room with semi-octagonal apse which is associated with a ceremonial function – a provincial architectonic variant of the solemnity with which the throne room is invested in court architecture, as Castro puts it;18 they conform to the model proposed for Box. Aguilafuente (Segovia) has a room with a semi-octagonal apse and a large squarish room – basilica and dining-room respectively – standing in much the same relation to one another as at Almenara. There is no obvious independent access to the former and its function would be in question but for a wall and doorway closing off that walk of the peristyle serving the diningroom. Those seeking justice must have entered at the far end of the adjoining walk to reach the lobby and two rooms adjoining the courtroom where they could prepare their business. Separation of public and private activities may be a key to identifying such rooms.19 Rielves (Fig. 48) may be a highly unusual instance of such planning. The putative basilica is separated from the rest; the round room is clearly designed for some special purpose; a grand entrance provides for members of the kin-group, a minor one for suitors and petitioners. Although these identifications of rooms have some consistency, it has to be recognised, in going further, that all such interpretations except that of the Flavian Palace are insecurely based. Yet only by speculating in this way, trying to identify 179

— Chapter Eleven —

Figure 48


— Palaces, Peristyle Houses and Luxury Villas —

Figure 48 Seats of Lordship

broadly similar rooms or groups of rooms – for at this architectural level diversity is the rule – can we hope to establish the relative importance of the exercise of lordship in the several villas and the part they played in government at whatever level. Comparable plans are widespread. In Hungary, Nemesvámos-Balácpuszta 1 has a large apsidal room on the main axis of the peristyle – the dining/reception room – and a smaller one, reached by a separate entrance, which may be the justice room. Both had mosaic pavements. This last identification is called into question though, by the mosaic pavement, which, in the main body of the room, is composed of geometrical patterns but has a middle panel of two ‘pheasant-like’ birds on branches


— Chapter Eleven — facing each other; and in the apse, a wine-jar amid tendrils of laurel. None of these motifs seems to be iconographically appropriate and may invalidate the function deduced on more general grounds. In France the inadequately reported villa of Montcrabeau-Bapteste (Lot-et-Garonne) has a large room with semi-octagonal apse at one corner of the principal of two peristyles, but its relation to any other major room is far from clear because the plan lacks detail.20 Possibly the two rooms were situated quite differently from their counterparts elsewhere. British villas are not all as straightforward as Box, and what may be true of Montcrabeau is true of the better reported sites of Bignor (Fig. 48) and North Leigh (Fig. 48), both commonly bracketed together as typical English courtyard villas. The latter has, at the west corner, what is usually regarded as a dining-room.21 It is in two parts, a square vaulted inner chamber (1) and a narrower and longer room partly cut off from it (1 A), like an antechamber, and the whole is unlike the usual diningroom. The approach to this room is also odd, not reached directly by a wide opening from the porticus but through a lobby at right angles to it, and so into the ‘anteroom’. Now this is not the way presumptive dining-rooms are approached: rather, the lobby and the adjoining rooms suggest the minor rooms needed as offices and for interviewing clients and suitors next to a court room, while a passage from the farm courts provides an appropriate point of entrance. And, as at Aguilafuente, the south and west porticuses are separated by a wall and a doorway (with a servant to oversee it, no doubt) to ensure the supplicants for justice did not venture beyond their proper station. That leaves the representational/dining-room in something like its usual position near the middle of the north-west range. This interpretation reinforces the idea that the room betokening authority need not be apsidal, as has already been suggested for Fishbourne. Suitability in terms of size, approach and relation to other rooms, notably the ceremonial dining-room, are the criteria. Those considerations may assist in understanding Woodchester (Glos.) (Fig. 47), the British villa known best for its huge mosaic pavement, and one of the least analysed. It is obvious that the mosaic room, 15m square, dominates the inner courtyard, but what is less noticed is the remarkable size of the range opposite, which is as wide as the hall of Blankenheim I (12 m) and considerably wider than the biggest room at Basse-Wavre (Belg.);22 and in this range the outer room adjoining the entrance is 14 m and the inner 15.5 m long, both with mosaics. Quite likely they intercommunicated. Rooms this size must have served some public function; the outer one may have been in the nature of an audience hall or antechamber opening off the covered(?) entrance, the inner one a justice room. Only Fishbourne has three rooms comparable to the three principal rooms at Woodchester, and they are of course larger and grander. Nevertheless, this does suggest some special status, a rank above other large villas; in fact, Woodchester is one of the few villas which one could conceivably imagine an emperor or provincial governor taking over during his travels – a palatial villa properly so called. That point is reinforced by the recognition23 that the room containing the great pavement was entered on all four sides by doorways in the middle of each wall. It provides no architectural background, that is to say no obvious place for a seat of honour, which must have been freestanding, so that this large room was a distant cousin of the audience hall in the Flavian Palace and the corresponding room 5 at 182

— Palaces, Peristyle Houses and Luxury Villas — Aquincum. A ceremonial use is implied rather than a grand dining-room, for which purpose the size of the room and the position of its doorways are unsuitable. The distinctive placing of the doorways especially must be a conscious copying of the arrangements in very grand establishments; they imply the presence of many servants, some to watch the doors, some to provide a suitably large retinue for the presiding personage and yet others for the performance of ceremonial acts. Such stateliness is hard to parallel outside truly palatial establishments, even in the largest villas, and is more than ordinary lordship. The great room itself is so explicit – it gives the impression of a throne room – that a provincial government might not have looked kindly on it had it been built by a private citizen without authority. Woodchester is perhaps the only villa at present known, and certainly the only British villa, which has truly palatial features. The point has already been made that justice rooms or basilicas can take different forms. A distinctive room shape not so far mentioned is the rectangular block some two or three times as long as wide, with an apse on one of the long sides. The bestknown example is the west corner room 27 of Echternach I (Lux.), which in III is modified to resemble the ‘lop-sided’ apsidal rooms at Ljusina (Yugosl.) (Fig. 60) or Orlandovtsi (Bulg.) (Fig. 60); another is Lalonquette I (Pyr.-Apl.) (Fig. 50). All these are of suitable size and have the necessary ancillary rooms and suitable access to be justice rooms, as was in fact suggested for Lalonquette24 and the same use seems quite likely for an apsidal room at Nemesvámos-Balácpuszta I (Hung.);25 and in both villas the apsidal rooms stand in a comparable relation to the principal room. At Montrozier (Aveyron) (Fig. 52) there is the block formed by the rooms 38, 39, 41, 42. These and other configurations need comparative study.

PERISTYLE VILLAS The origins of the peristyle are of no concern here. It was in use at Pompeii and Herculaneum in the third quarter of the first century and spread in western Europe through Italian models, becoming commoner in the Iberian peninsula than in any other country or region. A useful key to the more important26 shows that peristyles are rectilinear except for three which are trapezial: Santa Colomba de Somoza (León) (Fig. 49), Albesa-El Romeral (Lérida) and Jumilla-Los Cipreses (Murcia) (Fig. 49).27 Such departures from the norm could be ascribed to construction in two or more phases did they not have one thing in common, that in each case the principal or dining-room (triclinium or oecus) is at or near the acute angle of the peristyle. This indicates that the distortion had a social purpose and, like the change of shape in the courtyard at Sparsholt (below, Fig. 68, p. 246) and other villas, was intended to emphasise a particular room or building. A similar concern is apparent at Eisenstadt (Aus.), where the off-axis addition on the north side of a big representational room housing statues of gods appears to be connected with the trapezial shape of the peristyle. Not all rectangular peristyles are of one building phase. The main range of Cuevas de Soria (Fig. 49) had at one stage a big apsidal room flanked by three smaller ones, the latter being separated by narrow apsidal lobbies (or closed corridors). The 183

— Chapter Eleven — absorption of the westernmost room 1 into the west range suggests that this involved a rebuilding of the latter and a realignment of the peristyle, for there is no reason to doubt that the peristyle belongs to the first phase of construction. That is not the case with certain Iberian villas in which greater irregularities imply gradual development towards a peristyle. Santa Colomba de Somoza (León) and Quarteiro (Faro) are likely instances and Valladolid-Prado is certainly of two periods, neither really intelligible. These and a number of other sites show the kind of unconformity present in the final phase of Lalonquette (Pyr.-Atl.), one of only two large peristyle villas as yet excavated to its lowest levels.

Lalonquette: Gaulish Unit System to Mediterranean Peristyle In its final state this villa (Fig. 50) looks much like other peristyle villas of southern France or Spain, yet its first two phases [I] and [II]28 have much in common with

Figure 49 Spanish peristyles


— Palaces, Peristyle Houses and Luxury Villas — the houses of north-west Europe. Lalonquette began with a core of two hall houses: the one to east has some likeness to Frocester Court I (Fig. 69); that to west (a workhall?) has a small corner ‘room’ that looks like a dais area partitioned off from the hall, for which Ober-Grombach A (Fig. 2) may be a parallel. These two houses were linked by a porticus but physically separated by a narrow passage open to the sky. The more developed east house is linked by the porticus (but not joined structurally) to two identical two-room wing-like ranges29 looking like units or apartments – the two parts of Brixworth I (Northants.) (Fig. 11) or the north-east wing of Weitersbach II (Rhld-Pf.) (Fig. 69). In the small courtyard is a well or shrine, like Ditchley (Oxon.), and to the west is a workhall with the usual porticus. Lalonquette thus begins with three halls at various stages of development, accommodating housefuls of varying composition and serving somewhat different purposes, so that at this stage it was, in essentials, like its northern counterparts at Köngen (Bad.Württ.) or Katzenbach (Bad.-Württ.). Only in [II], when the elaborate bath suite, the justice room and the larger temple were added, did elements then foreign to the northern scene appear. The most notable of them is the basilica which, if correctly identified, shows lordship making a much earlier appearance than in British or German villas. In [III] most interesting changes took place. The main range remained in two distinct parts but the principal hall house and its quasi-wings were replaced by a large row-type house which has, in the middle, a representational room projecting at the rear.30 To west is a two-room-and-passage unit of Downton type; to east, a very similar unit except, perhaps, for one very small room; that, though, may be an alteration, as, surely, are the rooms at the end and rear forming? [IIIA]. At the same time, the west building developed like a German hall, acquiring extra rooms at each end, and an imperfect peristyle, partly blocked by the baths, was created. The transformation from hall to row-house can be matched on other sites but the relation of the new house to the whole is extraordinary. Instead of facing the middle of the peristyle court the representational room faces down the east walk in a way reminiscent of the off-axis dining-rooms discussed earlier. Why so? It must be bound up with the location of other buildings and the desire to provide an impressive direct approach along the east walk. Two constraints operated: the baths were close to the house and the most difficult element of the complex to move; and the main lines of the peristyle were established and its enlargement would have required extensive demolition prior to rebuilding. Given that the new row-house comprised two units, they could be differentiated by making one face inwards to the peristyle court and the other outwards to a garden, which is why it has its own short porticus on that side. The basilica of [II] lost its apse and was turned to some other purpose, being replaced by another opposite the main house; it could be approached from both sides, either from the new grand entrance with its subtle changes of room shape, or directly from the outer service quarters. By now the social emphasis of the villa was heavily towards the east, including, apparently, an exercise yard (palaestra). Not until [IV] did Lalonquette acquire a complete peristyle.31 On three sides it utilised the earlier walks but on the west was brought inwards to produce a perfectly rectangular uninterrupted shape, and nearly everything else around it was rebuilt. The fifth phase, though it resulted in a balanced and more nearly symmetrical 185

— Chapter Eleven — disposition of buildings and is interesting in itself, does not affect the development of the peristyle. The important lesson of Lalonquette is that it reveals, beneath the very Mediterranean-looking final phase, beginnings linking it to houses in the rest of Europe, and that the slight breaks of alignment in other peristyle villas can be more

Figure 50


— Palaces, Peristyle Houses and Luxury Villas —

Figure 50 Lalonquette I–V


— Chapter Eleven — confidently regarded as implying something similar. Castro remarked on the fewness of Spanish villas with dispersed (or separate) buildings;32 that view may well be modified when deep excavation of courtyard villas takes place.

Montmaurin I The only peristyle villa built as such which has been excavated as completely as its preservation permitted is Montmaurin (Haute Garonne) (Fig. 51). Its story is simpler than Lalonquette, there being just two main phases. Montmaurin I was a fully developed peristyle villa.33 The first thing to notice is that the peristyle has large rooms at its north-west and north-east (correctly north and east) corners that appear to have been used to facilitate circulation but also served to divide the villa into two parts, one the entrance court and the other the house with its appurtenances, including the north range of the peristyle. Looked at in this way it is apparent that there are two successive and substantially identical entrance blocks, each with a square portal or gatehouse reserved for important occasions, leaving the flanking passages, off which open small rooms serving as porters’ lodges, for everyday use. The peristyle itself is interrupted by two pairs of masonry piers marking what may be a stately approach used on important occasions; alternatively they were decorative, forming an impressive frame to the vista from either end. Flanking the north portal are identical suites comprising a large room and a small inner room, and because they appear to have been reached from the inner porticus (but were the doorways found?) they must have been apartments of some consequence; the corresponding rooms in the south range are probably part accommodation, part store- or workrooms.

Figure 51 Montmaurin I, II


— Palaces, Peristyle Houses and Luxury Villas — In the wider east range of the peristyle is what is clearly an important room, essentially a circle inscribed within a square; its purpose is unknown but it shows that this side of the peristyle was more important than the opposite side, which has only plain rectangular rooms and a portal leading to the service court. Whether the scanty remains establish that a service court existed to east, providing perfect symmetry with that to west, is open to doubt. That something stood on that side is clear,34 but it is worth noticing that the narrow passage in the east range has a porter’s lodge on the peristyle side, like those to north-east, whereas the wide opening opposite does not. It demonstrates an intention to control access to whatever lay outside the peristyle, not entry to it, which suggests a private garden rather than a service area. About the house proper little can be said with any confidence. It has a porticus giving access to the principal element of the plan, a long rectangular reception room35 forming an antechamber to what was evidently the most important place of all – presumably a representational room or shrine. If that is correct, the house was of a fairly conventional kind.

Montmaurin II Montmaurin II, like Lalonquette [III], saw an extensive remodelling, preserving only the peristyle and much of the fabric of its buildings around it. The social implications of this change have hardly been explored. The intention underlying the rebuilding was to convert it from a well-appointed large villa, probably with a sizeable home farm nearby, to a luxurious one having no connection with everyday affairs, not even the functions expressive of dignity and lordship. Change was proclaimed at the entrance to the new establishment: instead of approaching the house through a working courtyard, a new semicircular approach, walled off from any trace of useful activity, was created. It would be interesting to know who entered, and how: did everyone make an offering to the protecting deity at the altar to the left of the middle gateway under the eye of the porter in the opposite lodge, and, if so, who visited the polygonal temple before advancing to the house? Passing through the portico and the phase I entrance portal – perhaps now more of a reception room – the visitor saw an impressive vista emphasised by terracing and marble staircases, and terminating in an apsidal walled enclosure containing a piece of sculpture. To the left, between the corner of the peristyle and the wellhidden service quarters, was a large bath suite incorporating a nymphaeum of unusual size and splendour, the whole being devoted to bathing as recreation and enjoyment and no doubt entertainment. Those privileged to go further admired the peristyle garden before climbing a staircase to the marble-paved inner peristyle at the centre of rooms and spaces given over to pleasure, and perhaps principally the pleasures of summer. Two semicircular porticos, pilastered and, like all the rooms in this part of the house, floored and faced with marble, preceded the innermost and highest court. Off the gallery in front of it opened on each side a suite of three small and richly decorated rooms, and backing on to each was an annexe with boldly projecting pilasters. Pleasure being 189

— Chapter Eleven — the sole aim of all this part of the house, everything necessary to minister to it was hidden from view as far as possible. A telling illustration of the villas’s true purpose is the discovery that the six small basins on the east and west sides of the inner court were fish tanks, one of them still full of oysters with their shells firmly closed.36 Montmaurin is certainly a peristyle villa in both phases, and although I reflects an extremely comfortable standard of living, only in II did it become what can truly be regarded as a luxury villa in the strict sense of the term, one from which all the cares of administering great possessions were banished. Whoever the owner may have been in I, in II Montmaurin became a magnate’s pleasure house.

LUXURY VILLAS As the preceding discussion shows, it is not easy to isolate one characteristic by which to distinguish between villas as large and richly appointed as these, and to say which is luxurious and which palatial. Yet the attempt must be made, for until a distinction can be demonstrated, these great establishments will contribute no more than general impressions to the comparative study of social development in different parts of the Empire. Baths enable some distinctions to be made. They are, as Wightman pointed out,37 an early manifestation of the taste for luxury, so that Echternach I combined formal stateliness in the main house with luxuriance of form in the bath block. EuskirchenKreuzweingarten (Rhld-Pf.) and Bad Godesberg-Friesdorf (Nordrh.-Wesf.) appear to share these characteristics. Wightman also noted that the distance of baths from the house was an indication of the status of the owning family. Nennig (Rhld-Pf.), where the bath block was 250 m away, demonstrates this; so, equally well, does the littlenoticed villa of Thuy (Eure-et-Loir),38 where the baths are at some distance from the house39 and are reached by a porticus having a series of niches at intervals, like the ones at Val Catena (Yugosl.). Yet, although Thuy possesses these signs of real luxury, it was clearly a villa in a more utilitarian sense. So, too, probably, was Nennig: its middle room with the great mosaic is, in the useful phrase of a seventeenth-century English writer,40 ‘a piece of state’, not one devoted purely to private pleasure; flanking it on one side are comfortable living-rooms, on the other service rooms; and the detached quasi-wings recall those at Weitersbach II. Nennig has claims to be luxurious but is not, in the same sense as Montmaurin II or Val Catena, a luxury villa. It may be useful, then, to consider first which are luxury villas and which have luxurious elements, and so a selection will be passed under review in the hope of bringing out their differences. A few are clearly luxury villas. Wittlich (Rhld-Pf.), following the curving west bank of the river Lieser for 150 m, is, as Wightman remarked, the only one in the territory of the Treveri than can be regarded as ‘a mere pleasure-ground’.41 In the middle is a block conventionally said to be the dwelling house,42 which, very interestingly, is of palatial inspiration; it has the combination of large apsidal rooms open to the river terrace and flanking an imposing central block which is found in different forms at the governor’s palace at Köln or Aquincum. About 25 m to north, and joined to the middle block by a porticus at the rear and the riverside terrace at 190

— Palaces, Peristyle Houses and Luxury Villas — the front, is what is described simply as a bath block,43 although it has no less than twenty-four rooms, many of which do not fit easily into the usual sequence of functions. To south, joined in the same way and at the same distance, is what is called the working part, no doubt because the basement storey or undercroft produced clear evidence that it was a stable. The ground or principal floor, where there is a large room with two apsidal recesses for which it would be hard to find a work purpose, belies this. Wittlich, as clearly as Montmaurin II, is a luxury villa. Both the situation of Wittlich, overlooking a river, and its tripartite form recur elsewhere, by no means always in luxury villas. For disposition at the waterside the extreme case is the villa of Val Catena, built around the inlet of that name in the island of Brioni Grande near Pula (Yugosl.). On the south side is a large house built around two peristyles and combining comfort with the necessarily considerable working quarters. From it a portico curves around the head of the inlet to run straight for no less than 1.5 km, past various recesses fitted with seats, to a garden, baths and a suite of rooms intended for entertainment. On a more modest scale – and few villas can compare with Val Catena – Pully-Prieuré (Switz.), built overlooking Lake Geneva, had, on the landward side from which it was approached, a porticus at least 60 m long facing an ornamental pool about half that length; and, facing the view, it had a version of the angled front found at Wittlich, with semicircular recesses. Though incompletely excavated, the villa, like Val Catena, shows no sign of any activity other than those conducive to pleasure. That is not entirely true of Graz-Thalerhof (Aus.) (Fig. 52), built on the grand scale to an extended tripartite plan some 160 m long, of which 125 m are largely occupied by wings, rooms and two adjoining porticuses all heated by hypocausts. An attempt has been made to show the circulation pattern and the location of service rooms. At the west end domestic rooms and baths are grouped around a large entrance hall. An architecturally more remarkable block in the middle spanning the porticuses has what is, in effect, an elaborate porch from which the very large house was reached by a remarkably narrow entrance; the same effect as was achieved by circular spaces in Lalonquette [III]. Here two apsidal rooms whose purpose is hard to guess narrowed the opening. To south is an elegant and unusual room of basically cruciform shape.44 These two parts of the house may qualify as luxurious. At the west end, on the north side, is an exercise yard, all service functions being clustered in courtyards at the south side. There, too, is a large hall with a wide shallow apse to one side, giving it a lopsided look. If, as is likely, this was a justice room or audience room, Thalerhof, though certainly luxurious, was not, in the narrow sense, a luxury villa. Grandeur rather than sophistication is the hallmark of Téting-sur-Nied (Moselle).45 The great arc formed by the porticus and terrace and terminating in bowfronted pavilions of slightly different size46 is crowned by an apsidal-ended room, which must be a dining or ceremonial room at the heart of the villa. On the west side are baths and more, of which Grenier remarks that the other rooms ‘have the dimensions usual in triclinia or other rooms intended for social life and the idle conversations after bathing’; and as for the porticus itself, ‘the sun, at every hour of the day and in all seasons, had to light up and warm at least a part’.47 The little that is known about the east or house wing, or about the rest of the establishment, does not detract from this view that Téting is another luxury villa. 191

— Chapter Eleven — These few villas appear, as far as the evidence goes, not to have been directly connected with work of any kind, even (except Thalerhof) with the administration of estates. All have several or many rooms of complex geometrical shapes, a

Figure 52


— Palaces, Peristyle Houses and Luxury Villas —

Figure 52 Luxurious formal villas

proportionately large number of them with hypocausts and mosaics, and all have highly individual plans intended to permit maximum enjoyment of the landscape. Montmaurin II largely shares these characteristics, particularly in the inner peristyle, but is limited by retaining the larger outer peristyle court of I; consequently it has a much greater, though deceptive, appearance of formality. Those points, together with the genuinely palatial elements recognised earlier, may help to discriminate between villas not markedly different in size and the quality of their appointments but of quite varied social significance. The contrast between the quite common peristyle plan and the architectural luxuriance of the inner court found at Montmaurin II can be seen to imply, where that kind of thing occurs in other villas, similar constraints resulting from a change in the nature of the establishment. A way of life necessarily preoccupied with work gave way to one in which thought had to be given to ways of passing the time, and although nearly all men able to indulge in the second engaged also in the first to varying degrees, they had in common the ability to afford, in Thorstein Veblen’s useful phrase, conspicuous consumption. His words imply the desirability of separating elements of extravagance – the essence of the luxury villa – from the costly buildings, rooms and ornament demanded of those in high social or administrative position, and the need to pick out exceptional features from the rising standards of comfort observable over four centuries of imperial rule.

FORMALITY AND LUXURY Several famous villas combine a high level of comfort and conformity to advanced standards of taste, both of which can be taken for granted in luxury villas, with a formality of plan entirely alien to the latter. Woodchester, with an axial approach through two or three courtyards and principal rooms inappropriately large for private entertaining, provided an impressive setting for lordship in which luxury was not


— Chapter Eleven — pursued for its own sake. Comparison with other large and frequently cited villas is instructive on this point. In the elongated villa of Basse-Wavre (Belg.) the largest room is at the east extremity, some 50 m from the second largest room in the middle and at the west end is a luxurious-looking bath suite. Whatever point Basse-Wavre was intended to make, it is not primarily a manifestation of lordship. Mention has already been made of Nennig (Rhld-Pf.), ‘the grandest of all the excavated villas in the civitas Treverorum’, even if Mylius’ reconstruction with a twostorey porticus and wide pavilions of three storeys and basement somewhat overstates the grandeur.48 Among other things, the big mosaic-floored hall in the middle, used for important occasions of a ceremonial nature, is envisaged as being surrounded on all sides by a colonnaded gallery; yet the notion that the great personage at the centre of such observances might be overlooked, or that there might be anyone or anything above him other than a canopy of state, strains belief.49 In general the plan of Nennig is strictly rectilinear and formal, lacking any architectural flights of fancy. The only touch of luxury is the placing of the baths at a considerable distance from the house. For obvious reasons no two large villas are alike. The most that can be expected is to find what appear to be principles of planning which have administrative and social implications. Haccourt II (Belg.) (Fig. 52) incorporates two aspects of formal planning that link it, on the one hand, to Nennig, and on the other to MontrozierArgentelle. It resembles Nennig in having a lofty audience hall (4) in the middle, here looking out over a large ornamental pool (3), with the smaller rooms opening off courtyards: a large one surrounding the audience hall on three sides and a smaller rectangular one (10) at the superior end. Facing south over an ornamental pool is the dining-room (11) and near it a room with an apse (9), under which was a cruciform cellar – a family shrine? It is in the north wing that some resemblance to Montrozier (Fig. 52) can be detected. The parallel blocks of rooms are quite different in detail in the two villas but each shows a distinctive kind of compact planning that mixes service functions and accommodation for those who performed them with the architecturally interesting shapes that must connote superior purposes. At Haccourt a range of rooms (18–23) looking exactly like a row-house faces the semicircular porticus which cuts it off from the rest of the wing except, perhaps, for the elaborate room 16 with its several niches and recesses. Montrozier (Fig. 52), not all of which was excavated – suitably luxurious private quarters are lacking – combined several buildings commonly dispersed more openly in courtyards and wings: two aisled buildings, one of which, by its size and structure, can only be an audience hall; another apsidal room which is perhaps a reception room; a bath block, normally detached in an establishment of this standing; and an exercise yard (palaestra). The whole has the orderliness, on a larger scale, of the buildings grouped at the north-west corner of Badajoz-la Cocosa (Fig. 48), and is generally reminiscent of the close groupings in parts of Chiragan,50 but in a more orderly way. So what is the status of Montrozier? The element that stands out as requiring explanation is the audience hall, which is twice the size of the famous one at Nennig and somewhat larger than that at Haccourt II, though hardly more than half as big as the Fishbourne Palace audience hall. It has a monumental approach 194

— Palaces, Peristyle Houses and Luxury Villas — around a peristyle; it abuts a wide entrance flanked by two large rooms comparable to those at Woodchester; and it has in front what may be a large water feature, the equivalent in a peristyle of the pool at Haccourt. The smaller aisled building may be a reception hall for suitors expecting to proceed to the basilica. In some ways Montrozier resembles Haccourt, yet it has the buildings for the administration of justice so strangely absent in the latter, and the audience-hall complex has a palatial quality. Only further excavation could show whether other official, that is, clearly not domestic buildings are present. Lacking information of that kind, the status of Montrozier remains uncertain. It may be a large, comfortable villa expressive of dominion and occupied by someone who had achieved absolute power over his kin, or it could be a true palace.

LORDSHIP OR JOINT PROPRIETORSHIP? It is easy enough to pick out in the large villa of Fliessem-Weilerbüsch (Rhld-Pf.) (Fig. 53) the elements indicative of the occupants’ high social status. The very grand south porticus with its elaborately planned wings may be coeval with ten or more mosaic floors, hence in Wightman’s words, ‘It may be surmised that a wealthy owner had decided to make the villa his home’.51 No attention is paid to a peculiarity of the main range, that the room facing the main (west) entrance is not a large hall but a comparatively small one (36), and it backs on to a counterpart facing east (35); both are almost completely open on the entrance side, as rooms in such a position usually are. They therefore fulfilled much the same purpose as the middle room at Newport (Fig. 10). Moreover, the east and west porticuses are divided into two unequal parts and consequently the wide entrances are off centre to the two rooms; this deviation from customary practice shows how important the division of the porticus was. It is bound up with the two transverse corridors of unequal width flanking the middle room; that to the south leads to the outer flanking rooms, the function of the other is uncertain and two possibilities are suggested (Fig. 53). The more important point is that there are two entrances of equal size and importance from courtyards on opposite sides of the house, and two entrances can only signify two groups of equal status. That is basically why the house presents the appearance on both sides of a porticus between wings. It bears the same kind of relation to the row type as do houses of the compact row type – a distant derivative, but showing in the core rooms clear traces of its ancestry. How the rest of the house was used and divided cannot be examined in detail except at one particularly instructive point, the block of four rooms (48–51) at the south end. The west room (51) is marginally wider than the east (48); entry from the flanking corridors is at opposite ends, the former to north, the latter to south, and in each case is faced by a doorway into the adjoining service room: an instance of reverse symmetry which exemplifies the minor social differentiation that was always so carefully observed. So not only was Fliessem divided into two parts; each end of the house, north and south, can be subdivided, as was demonstrated summarily by Böttger.52 This is shown both diagrammatically (Fig. 53), with some differences from Böttger, and in more detail (Fig. 53). 195

— Chapter Eleven — About Haccourt I (Fig. 20) much less is known but what was found of the house resembles German examples of the row type. It is like Fliessem in the significant respect of having two porticuses, may also have been of much the same size,53 and so may have been like it in social composition too. The two villas developed in altogether different ways. Haccourt II (Fig. 52: III is

Figure 53


— Palaces, Peristyle Houses and Luxury Villas —

Haut Clocher-Saint-Ulrich 1:3000 Figure 53 Joint proprietorship – grandeur, not luxury

only new baths) was rebuilt on a new alignment a little to the west, on a larger scale and different plan which seems to have provided for two households of considerably different standing, and was dominated by a great hall in the middle, the symbol of unity. Fliessem, on the other hand, enlarged and embellished, retained its original form to the end because the social structure of the community did not change fundamentally; that is why the rooms between the wings, contrary to what often happened, remained unaltered. That the villa was prosperous is shown by the addition of the new south porticus and crypto-porticus, new and larger baths, and large pavilions at three corners, not to mention mosaic floors. The colonnaded south porticus made a statement about social importance implying that whoever occupied that end was more important than those living at the north end, but that seems to be only adding emphasis to what had always been the case. It also provided the elegance of curvilinear rooms, notably in the south-east corner although, surprisingly, that quality was lacking where it might most be expected, in the new baths. It is also surprising that Fliessem did not contrive that other token of luxury, the distant bath reached by a long porticus. That is what Haccourt III achieved, whereas Fliessem’s inhabitants remained content for two baths, neither very remarkable, to remain integral parts of the house; indeed, two baths may have been an essential accompaniment of the continuing kin-group structure. It is necessary to add that the notion lately advanced, that the new baths at Fliessem, furnished with mosaics, were intended for the outdoor servants (Hoffamilie), is bizarre.54 The divergence of these two villas from similar origins reveals contrasting forms of social development. Other villas of very different appearance can be analysed to the same effect. A notable example is the large establishment of Haut Clocher-Saint Ulrich (Moselle) (Fig. 53),55 dug at the end of the nineteenth century with large resources and reported all too summarily. Much the most striking element of the plan is the wall running unbroken from end to end to divide the massive core of the 197

— Chapter Eleven — house (52×28 m) into two equal but not identical parts. Wichmann makes mention of only one doorway discovered in this part of the house, the wide one (4 m) from the east courtyard into the biggest room. All he says about intercommunication between the two parallel ranges is, first, that the two principal rooms facing west and east were certainly connected, although the phraseology suggests this was an assumption,56 as, probably, is the statement that the customary intercommunication was by the flanking corridors 3 and 5, for which no evidence is adduced. Given the existence of two courtyards and the example of Fliessem, Saint-Ulrich may be another villa that began and long continued with some form of joint occupation and proprietorship. If so, two bath suites might be expected. One was in the elaborate complex of rooms in the north-west corner (5) and Grenier thought there was evidence of a second, but his view has been rejected by an expert on the subject.57 There is really little sign of luxury anywhere in the villa despite Grenier’s claims58 and our ignorance of the mosaic floors, which would have made it more comparable to other large villas but went unreported.59 Architectural subtlety and elegance like that at Fliessem is quite lacking. Nevertheless, here, too, joint ownership seems to have been weakening. An indication of this is the entrancehall or reception room to the east, which is not only large but extremely grand, with an internal colonnade on each side.60 It was much superior to its fellow to the west and a showpiece for the whole villa. No doubt these accounts of Fliessem and Saint-Ulrich will appear incredible to many. All that a doubter has to do is to explain why the main ranges of the two buildings are so unlike the general run of their kind.





hus far villas have been categorised without regard to the countries in which they are situated. This approach is equally applicable to south-east Europe, a term which for the present purpose comprises Bulgaria, Hungary, former Yugoslavia and Romania, together with much of Austria. Villas in this vast region are different, broadly speaking, from those in western countries, the houses being usually small and simple and for the most part of distinctive regional types. Moreover, the total number excavated is small compared with Germany, France or Britain, so that a type is established more by common features than by the closeness of resemblance found with halls or row-houses. House types will be discussed first, then the ways in which buildings were grouped. By way of preliminary, though, one considerable limitation must be stated, that ignorance of relevant languages compels me to rely on reports or summaries in German or French; but, that said, it may be possible to improve on the existing villa typologies in these countries.1

HALL-TYPE VILLAS It is remarkable that the one house type that might have been expected in a region of many quite small villas is not at all common, at least in its orthodox western forms. Szentkirályszabadaja-Romkút (Hung.) has as its principal building a hall of the form, rare in western provinces, with an end-entrance; it had two unequal pavilions touching the hall only at the corners, must have had a timber porticus, and has been reconstructed in an impossible way as a yard with corner towers. A similar but rather larger hall at Gyulafirátót-Pogánytelek (Hung.) (Fig. 54), secondary to what is said to be an atrium house, is the subject of a reconstruction as ridiculous as any in the literature: Oelmann’s article on Stahl is quoted for parallels to a villa which is then reconstructed with an open yard.2 How such a house might develop appears from Bistritsa (Bulg.) (Fig. 54), where part of the interior space is occupied by a long lobby and two unequal rooms. How the hall was entered is not known but so large a room is likely to have had a more imposing and more visible doorway than the lobby could provide. This plan 199

— Chapter Twelve —

Figure 54 End-entrance halls

illustrates very clearly how the pavilions were related to two more or less separate parts of the house, as has been suggested for other villas with four. Two endentrance halls of an advanced kind are Budakalász (Hung.) (Fig. 54), which has an apsidalended hall, and Majdan (Yugosl.) (Fig. 54), which is broadly similar although the apse is smaller. In Bulgaria Mikhailovgrad site 2, building 1 is bigger and combines endentrance through a porticus with an apse added to an inner room which is comparable to the transept-like space at Majdan. All these houses are related to Mauren (Bay.) and other end-entrance halls and those with apses imply some considerable change in the way the hall was used; probably a growth of control on the part of an individual or a family and some degree of lordship over others, the apse being the seat and symbol of power. End-entrance halls are also found as subsidiary buildings at Champdivers and Tavaux in the Jura, which were discovered by aerial photography. The sole example of what may be an aisled house occurs in Romania, at Telita (Fig. 54); more likely, it is a derivative because it lacks some of the characteristics of the type. The central part has at the outer (south) end a doorway of domestic width, not one suitable for work purposes as is usual in British examples; and there is no trace of replacement of posts by partitions, or of any division of the large north room along such lines. That room could well be a cross-wing to a hall with outshots. Yet another curious feature of the building is the projection of the aisles/outshots to form rooms; how they were roofed is hard to imagine. However Telita be interpreted, it is so far unique in the region and the true nature of its structure needs to be established.


— The Villas of South-East Europe —

Figure 55 Square halls

SQUARE HALLS Houses having a square hall as their core structure lack the propensity to longitudinal development which is characteristic of oblong halls, as Koerich-Goeblingen 2 (Lux.) (Fig. 30) shows clearly. At Mali Mos?unj (Yugosl.) (Fig. 55) the square hall has on two sides what looks formally like an L-shaped porticus with terminal rooms, yet the hall itself is the point of entrance and it can hardly be for display or to overlook a view – the same problem as that presented by Ovillers and other row-houses.3 It occupies a disproportionately large area relative to the hall and is likely to have been simply another room. Another square house, Lisic?ici 2 (Yugosl.) (Fig. 55),4 suggests what it might be: a comparable L-shaped room heated by a hypocaust is the equivalent of a pavilion or other best room. With these houses can be grouped a third, Stolac 2 (Yugosl.) (Fig. 55), which has some evidence of functional distinction expressed by flooring materials in the same way as at Frocester Court. Here the hall is entered through a porch and into an area (2) paved with stone rubble, the rest of the room (6) having an L-shaped hard plaster floor. The larger subsidiary room 5, by intruding into the stone paving, must have directed attention to the body of the hall where the hearth was. A reasonable conjecture is that it stood opposite the doorway, with the place of honour behind it and behind that the doorway into the larger inner room. The remainder of the hall was given over to some superior use. This kind of plan is not illuminated by early modern comparisons; important Lshaped rooms seem to be unknown. Lisic?ici 2 suggests that rooms like this may at first have been divided according to customary use and only later, in both literal and typological senses, broken up by partitions. It is the kind of process implied by Frocester Court (Glos.) and to be inferred in the difference between Wahlen (Switz.) and Mauren. One or two houses formally of row-type may be better interpreted, because of their size, as a pair of square halls separated by a lobby or shrine-room. BihacZaloz?je (Yugosl.) (Fig. 55) is like this, one of its halls being heated by a hypocaust. It can be compared to Wancourt (Somme) or Niedereschach-Fischbach (Bad.-Württ.) and has no specially south-east European characteristics. That is true also of the double-ended hall houses at Bruckneudorf (Aus.) [I] (Fig. 69) and Rohrbach (Aus.), a single-ended hall like Deva (Rom.) (Fig. 73) or the aisled hall at Kaisersteinbruch (Aus.) (Fig. 67). It is not that this part of Europe lacks types common further west,


— Chapter Twelve — simply that they form a minority of villas and are outnumbered by others which cannot easily be related to them. That point is borne out by the unusual building Sümeg 1 (Hung.) which is an enigmatic instance of the hall-or-yard problem of southern Germany. The internal space, some 18 m wide, is surrounded by rooms on three sides, two of which are two rooms deep. Since there are four pavilions touching corner to corner they require timber verandas, without which they are inaccessible; but these make the lighting of the double-depth sides difficult, and with a hall in the middle, impossible. So in that case the internal space is not a hall but a courtyard which is also hard to explain in detail.5

ROW-HOUSES A few row-houses, most of which are not much like those of France or Britain, can be identified. An exception to this rule is Travnik Rankovicí (Yugosl.), which would not look out of place in Picardy; Ovillers and Mons-en-Chausée are very similar. It would be impossible, though, to match in the Somme basin a building backing on to the yard wall at Drac?evica (Yugosl.) (Fig. 56) which is a row of three large rooms and one or possibly two smaller ones, each entered independently; there is no sign of porticus or veranda; and Lisic?ici 3 (Fig. 60) (following the numbering given by Cremos?nik who dug this site) is a smaller house of the same kind. Kran (Bulg.) comprises four cells entered independently from a veranda and so has a general resemblance to western row-houses, which is as much as can be claimed for an Austrian house, Mautern am Donau, where only in [II] does the characteristic feature of the row-house, the lobby, appear; until then the rooms may have been approached independently from the porticus. Equally scarce is the porticus with pavilions, whose sole representative to date is the partially recovered villa of Sarmizegetusa (Rom.) (Fig. 56). The core building can hardly be anything but a row-house of unorthodox plan; however the whole is reconstructed, the extreme individuality of villas in this part of Europe makes it unlikely to have been absolutely symmetrical, as has been suggested.

ROW-HOUSE EQUIVALENTS Axial Entrance-Halls The few houses in this category are distinguished from the axial-corridor houses described earlier (chapter 7) by having an entrance directly into the corridor and by a different disposition of rooms. In the simplest, Keszthely-Fenékpuszta 8 (Hung.) (Fig. 56), the entrance-hall is flanked by two identical apartments, each comprising two equal rooms and a middle lobby, like the units at Downton (Wilts.). That it is a lobby is implied by the impossibility of reaching the end rooms 2 and 7 from the hall, there being insufficient space for a doorway; access could only have been through 3 and 6 respectively, or 1. The latter was perhaps, by analogy, a shrine room performing part of the function assigned to the middle room in a row-house such as 202

— The Villas of South-East Europe — Newport (I.o.W) or Marshfield N IIIA (Glos.).6 Hardly more complex is Winden am See II (Aus.) (Fig. 56). It differs from the Hungarian house in having unequal rooms in the two rows, all three on the south side having hypocausts, and also a porticus, part of which is cut off to provide a small end room whose purpose is unknown – a porter’s lodge is one possibility, a shrine another. It could be classed as an internal corridor house and is a smaller and socially more closely knit version of MunzenbergGambach. In III, Winden am See changed character by the addition of an apse and the creation of a large room at that end (Fig. 56).7 In that respect it resembled Majdan (Yugosl.), a villa which, otherwise, has equal minor rooms like Keszthely 8 (Fig. 56); the porticus is reduced to fit between the end rooms, as if in imitation of the porticuswith-pavilions. Budakalász has affinities with these villas; it has the equal lesser rooms of Majdan and the porticus and apse of Winden III. All have a focal point in the

Figure 56 Row-house equivalents


— Chapter Twelve — apsidal room facing the entrance, so that anyone coming into the hall would instantly see this dominating feature of the house, whether seat of honour or shrine.

Block Plans What must surely be the extreme limit of compactness was reached at Budapest IIICsucshegy (Hung.).8 Only one other house falls into this category, the very interesting Manerau [II] (Fig. 60).9 It seems to have begun in [I] with a wide entrance-hall equivalent to a transverse lobby in its primary function. To the right was a hall, to the left two large rooms.10 Reorganisation in [II] produced in the east half what looks like a compact row-house; it has a lobby – longitudinal, in relation to this part of the building – from which two room-plus-inner-room units were reached. Alternative interpretations are offered, that of IIB, which bears some likeness to Hummetroth or Lussas-et-Nontronneau, being perhaps the more likely. In the west half the plan is less certain and the interpretation dubious; the two north-west rooms and the lobby may form a unit and the remainder a hall, like the building at Hobita (Rom.) (Fig. 60) or, on a smaller scale, Izola (Yugosl.) (Fig. 60). This interpretation of Manerau [II] rests on the assumption that passage-rooms are unlikely to have found favour in the Roman Empire any more than in later periods, and if that is correct, the resulting likeness between houses so far apart must arise from a common body of theory and a common approach to design.

Houses Essentially of Two Cells An altogether simpler kind of house has a core of two cells, one or both of which may be subdivided, and one or two transverse corridors or lobbies. The nearly square House 1 at Maria Ellend-Ellender Weingarten (Aus.) (Fig. 57), tiny though it is, has a middle room with hypocaust and terrazzo floor and appears to have been the principal house of a farm. It was probably entered by the slightly tapering porticus at the north end, where a small projection on the west side, for which no explanation was offered, may tentatively be interpreted as the base for a statue or shrine. The status of the remaining room is unclear; its position suggests it might be an inner room to the hall in the middle, its size that it was perhaps a second unit like the independent rooms at Dracevica. This is a kind of small house that may have been occupied by a conjugal family or a stem family. Egregy (Hung.) may be a similar plan on a larger scale with the addition of a porticus. Two-cell plans without lobby or entrance-hall appear in a number of houses. One at Sarajevo-Stup (Yugosl.) (Fig. 57) has a large well-built hearth in the middle of the open hall-like living-room – not where it might have been expected, since this is an inner room. Entry to the house is through a larger and apparently unheated room by two adjoining doorways. This, like Sarica (Rom.) (Fig. 57), must be a single-family house of a size and simplicity not paralleled in western Europe. A slightly more complicated plan is found in the small house D at Kaisersteinbruch. It was clearly a house of some quality, all rooms being heated; the two small rooms had hypocausts 204

— The Villas of South-East Europe —

Figure 57 Houses essentially of two cells

and the large one a circular oven. What differentiates it from other buildings of its kind is the close relation between it and the adjoining house,11 D being a superior residence adjoining the work building C; for C[I] is said to have had, under the floor of the small room 3, a pottery kiln.12 Houses of this type may have both cells subdivided. Fischamend-Katherinenhof (Aus.) (Fig. 57), which is less well recorded than Kaisersteinbruch D, is another wellappointed house, with wall-painting and evidence of hypocaust heating, yet if it is a single house its plan is curiously clumsy. The largest room (1), a general living- and workroom, probably, was the point of entrance from which other rooms were reached; yet, unless timber partitions were not noticed, there must have been a second passageroom. An alternative is to suppose that each of the two square rooms 1 and 2 had an inner room on the north side, although that raises an unanswerable question about the function of 2, which can hardly be a second general-purpose room unless the whole formed two separate domestic units. Simple plans lacking informative detail are hard to understand. Cincis (Rom.) (Fig. 57) has some resemblance to Fischamend in its block of four rooms (2–5), yet the presence of a narrower end cell (1) may invalidate the comparison. Room 2 looks like a lobby but, if the house were entered there, room 4 (or, less likely, 3) would be a passage room to 5. Entry into 3 would remove this difficulty without producing a plan that can be paralleled in the number and arrangement of rooms.

HOUSES WITH ONE CROSS-WING One of the most remarkable Balkan houses is Stolac 3 (Fig. 58), justly called by Truhelka, its excavator, the Mosaic House. It comprises a range of two rooms (3, 2) 205

— Chapter Twelve —

Figure 58 Houses with one cross-wing

and a cross-wing (1),13 and at the rear is the largest room (4). All had mosaics, that in 4 being of high quality with animals depicted in the middle panels, yet the three aligned rooms formed a little enfilade, one entered from another, and the rear room 4 can only have been entered from 2. This arrangement rules out Stolac 3 as a richly appointed row-house because effectively it has only two rooms, the other two being successive antechambers to them; indeed, it is not easy to envisage it as a house at all, there being no ancillary service rooms. No real parallel is known; the nearest, but only in its proportion of mosaics to rooms, is perhaps the small house Milton KeynesBancroft 1 V (Bucks.). Stolac 3 must surely have been the centrepiece of a group of buildings, and the enfilade suggests some purpose to do with administration or the reception of inferiors by a local magnate. Two small cross-wing houses at Stolac 5 (Fig. 58) and the smaller villa at Regelsbrunn (Aus.) (Fig. 58) are very similar and fairly conventional compared with Stolac 3. Stolac 5 is the more intelligible of the two because an adjoining building clarifies its function in the little villa complex. It is entered at one end of the porticus, beside which are two intercommunicating rooms – a small one (1) being an antechamber to the larger (2) – forming a unit of some kind. At the other end is a larger room serving work purposes but smaller than the word workhall usually denotes. Standing almost corner to corner with this building is a much larger one, a hall. The position of the entrance, whose precise size is lost through stone robbing but is approximately in the middle, implies a functional division into two squarish spaces separated by a narrow one, as has been envisaged for Bristol-Kingsweston and other halls. Their relation to the rear porticuslike rooms is uncertain; Proboj (Yugosl.) (Fig. 58) is similar in this respect. Stolac 5 comprises the equivalent of the row-house and aisled house found at Mansfield Woodhouse (Notts.) (Fig. 7) or the small square house 2 and the hall 1 at Lauffen am Neckar (Bad.-Württ) (Fig. 41). These comparisons, and they could be 206

— The Villas of South-East Europe — multiplied, show how unusual Stolac 3 is and suggest that it had some important function which, elsewhere, was carried out in part of a larger building. Both the Stolac houses are superior to the larger one at Proboj (Fig. 58), which is formally of crosswing type and gives the impression of a working farmhouse.

HOUSES WITH MULTIPLE SMALL ROOMS Bruckneudorf [F] and Manerau [II] represent in south-west Europe the tendency observable in western provinces to replace halls by houses with more and smaller rooms. Two closely related Romanian houses exemplify the same development, Apahida from the beginning and Ciumafaia II (Fig. 59) after extensive rebuilding. Both are unusual among houses of modest size in having two apses unconnected with baths, something that links them to the far simpler house Keszthely-Fenékpuszta 7 (Hung.) (Fig. 59). The spatially impressive effect of an apse and the kind of use it is likely to have been put to seem to make one enough for any but the largest houses. When comparatively small villas duplicate such a special feature there must be some fundamental reason for it, especially in a province not notable for advanced forms of architectural expression. Ciumafaia underwent at least one major rebuilding which has made it imperfectly intelligible. The plan claimed for the early phase14 bears no resemblance to any other villa and, more important, the rooms neither hang together coherently nor fit easily into the later work.15 Only on [F] (Fig. 59), which has the same awkwardness of an altered plan as Bruckneudorf, can positive comment be made. The north side of the building resembles Budakalasz or Winden a.S. III, and room 2 is apparently related to it in much the same way as the entrance-hall to the apses in those houses except for being off-axis and closed off from it by (presumably) a doorway. On the front of the house a second apsidal room (4) is not, as might be expected, connected with baths, of which there is no other possible trace; instead it provides a room of some architectural quality matching the one on the other side and must reflect a social requirement, not mere luxury, otherwise it would be more natural for two such rooms to form part of a suite. If that be accepted, the rooms of Ciumafaia are grouped around the entrance-hall (2) and the lobby (3) opening off it (to avoid passage-rooms), with 5 and 4 each part of a two-room unit, leaving 11 and the room to north as another.

Figure 59 Houses with multiple small rooms


— Chapter Twelve — Apahida (Fig. 59), only 15 km away, has three apses of which 7a, part of a bath suite, can be disregarded. The house was entered on the south side by the L-shaped entrance-hall (7), the counterpart of 2+3 at Ciumafaia. The remainder falls into two parts: to the west a block of three rooms entered from the hall may comprise a livingroom, an apsidal inner room and a service room; to the east the lobby (5a) leads to the important room 5 and the large room 1. That by no means disposes of all the difficulties, nor does it explain all the small rooms, but interpretation on those lines is strengthened by a comparison with the simpler house Keszthely-Fenékpuszta 7 (Fig. 59) which also has two apsidal rooms. The entrance-hall separates a superior block of rooms (4–6) from an inferior one (2, 3) as noted by Thomas,16 each terminating in an apse. Why both apses should have been set off-axis, though not explicable, demonstrates that they and the rooms were used in a similar way. Most of the houses so far mentioned prove on analysis to be readily divisible into two parts. This is not true of all the two-cell plans unless the assumption be made that passage-rooms are unlikely to have found favour then as now. Though the significance of bipartite division is open to argument, that characteristic, running through so many other kinds of house and expressed in such varied ways, requires some equally comprehensive explanation if the notion of joint family occupation is rejected. As will appear later, two important groups of houses lack this duality, but several larger forms of plan can be interpreted as multi-family establishments.

THE WAYS OF GROUPING BUILDINGS In Line As with the houses, so with the grouping of buildings: the villas of south-east Europe have only a passing resemblance to those further west. Villas like Montacute-Ham Hill or Maillen-Al Sauvenière which were set out in line and linked by porticuses, baths or other rooms do not exist. The nearest approach is Mautern am Donau (Aus.), where a workhall abutted the house to north, and Kaisersteinbruch C (Fig. 60) may be another such. Related to it is Budapest III-Testvérhegy (Hung.), where the principal building of three steps out of line with the others. Other villas which have little or no resemblance to anything in western Europe may be explicable on these lines. Konska (Bulg.) (Fig. 60) has two sizeable halls fronted by a single end-entrance porticus, and abutting this block is another, also of two distinct parts, which looks like domestic quarters. The less well-preserved building at Chichmanovtsi (Bulg.) may be an instance of this arrangement, with two domestic units at the end of a workhall. The point of laying out buildings in line is to demonstrate the relative importance and status of each. This aim was achieved in a different way with two fairly small houses at Weyeregg (Aus.) which stand a short distance apart on opposite sides of a boundary wall joining them. This disposition of two houses, which are plentifully furnished with mosaics and hypocausts but have few service rooms and in that respect have a south-east European air, is comparable to that of the two principal 208

— The Villas of South-East Europe — houses at Collingham-Dalton Parlours (Fig. 68). Its essence is that two houses were independent in all aspects of daily life while acting jointly in working a farm and forming a single unit in political matters. The total number of households within the houseful could vary and the closeness of the houses reflects the degree to which everyday life was integrated. Where the houses were a little distance apart and more or less equal in size and appointments, as at Weyeregg, two parts of a kin-group cooperate economically and politically on equal terms. Where they stand corner to corner and are considerably different, as at Mettet-Bauselenne [I] (Belg.), they are in every sense closer and this leads eventually to a degree of integration in [F]. Where they stand end to end forming an architecturally integrated whole, as at Marshfield (Fig. 66), one house is in a dependent or subordinate relation from the beginning. How the two elements begin and how they develop can be compared with the varying fortunes of their counterparts in other villas, notably with the row-house and aisled hall at Norton Disney and Sparsholt.

As a Block An alternative to housing various domestic and agricultural functions in abutting buildings was to put them together under one roof. That is what seems to have happened at Hobita (Fig. 60). The largest room (1), the hall, is at the south-west corner, with a room (2) to the north which may correspond to the long narrow room sometimes found at the end of a hall.17 In the other half of the building and reached by a corridor (3) is a small bath suite and two rooms. As is apparent from the plan, Hobita is unusual among unified houses in the proportions of its rooms, two of the largest being oblong and only a minor one having the common square shape. Izola (Yugosl.) (Fig. 60) has a general resemblance to Hobita in the position of the hall at the corner of the core rooms and in its room proportions. On the south side, what is said to be a porticus is unbelievably narrow for the purpose;18 more likely the house was entered by an internal porch on the north side that served also as a lobby to the two flanking rooms. Bearing in mind how widely the porticus in its various forms was adopted, it is extraordinary that so many Balkan houses manage without this very useful way of articulating the rooms of even a simple plan. Novi Saher (Yugosl.) is small and so compact that it is hard to imagine how it functioned. It has four rooms, two squarish ones (2 and 4) flanked by two long rooms (1 and 5); both 4 and 5 had hypocausts. Room 1 is an inferior room from which the hypocausts appear to have been fired; 2 may have been the point of entrance. It is difficult to see the point of giving what may have been the best room in the house, 5, such extraordinary proportions as 4:1, yet since it is only a local manifestation of a widespread phenomenon the space must have been usable in practice and desired for social reasons. All that can be said is that such a shape necessitates use as two parts and a function duplicated; that and limited resources may have determined the plan. Several times in the preceding chapters a connection between western and southeast Europe, running contrary to the general impression given by villas of the region, has been noticed. The theme of compactness brings another instance in the likeness 209

— Chapter Twelve —

Figure 60


— The Villas of South-East Europe —

Figure 60 Ways of grouping buildings

between the villas of Deutschkreutz (Aus.) and Friedberg-Pfingstweide 2 (Bad.-Württ.) (Fig. 41). Both are effectively three rooms in depth, combining a row-house, a hall and a porticus arranged as a triple-pile. This is particularly true of Deutschkreuz, where the porticus has, not pavilions but an internal porch which, implicitly, is flanked by rooms. The hall, most unusually for a room of its size (19.5 × 8 m), was heated by a hypocaust and had a mosaic floor.19 Houses like these may represent the same tendency as that which produced compact row-houses, one which runs contrary to that other tendency, predominant in the western provinces, to elongate a house one room deep, whether in a line or around a courtyard.

L-SHAPED PLANS Given the general idea of a house with rooms and buildings set out in a line, a big villa might stretch out very far, Maillen-Ronchinne (Belg.) for almost 300 m, VoerendaalTen Hove (Neth.) (despite its having courtyards) 180 m. Although nothing as big as these has yet been found in south-east Europe, for the more modest continuous total 211

— Chapter Twelve — length of 65 m of Iskar-Gara (Bulg.) (Fig. 60) an L-plan was preferred. The domestic or north range20 comprises, between a bath block and a hall, four rooms, all of different size – a characteristic of row-houses. At the south end of the east wing is a larger hall – a workhall – and between it and the small hall are two unequal-sized rooms, one perhaps a lobby. The whole provides much the same kind of accommodation found at Konska: each villa has two halls and two rooms linked by a porticus, with further domestic rooms – fewer at Konska than Iskar – beyond. And although Iskar is inferior in amenities to Sudeley-Spoonley Wood (Glos.), the elements comprised in the respective plans are not very different – major and minor halls, a bath block and two or three domestic units – so that overall the two houses are sufficiently alike for anyone to have felt at home in both, even to a doorway cutting off the bigger workhall. Kralev Dol (Bulg.) (Fig. 60)21 is bigger, having the considerable total length of 100 m. Although it is not strictly of L-plan, the north range was clearly the least important and will be disregarded. A slight break forward near the middle of the south range shows that, contrary to what its excavator Naidenova supposed, the east part (rooms 1–8) is of a different and probably later phase from the rest. The remainder makes an instructive comparison with Iskar. It comprises, first, four rooms of different sizes looking like a little row-house. In the west wing a small hall (14) adjoins the domestic units and a pair of square rooms separates it from the large workhall. How the remaining room 19 and room 20 in the north wing, and also the oblique east range 21, modified this pattern is impossible to say, but otherwise the resemblance to Iskar is marked. Because Kralev Dol is unusually fully reported, this alternative interpretation is worth discussing further. The added east part of the south range22 was shown by archaeological evidence to be devoted to work activities or storage and, being sealed off from the rest of the house by a wall, was entered separately by a strangely offcentre monumental gateway. This much is well founded;23 the west wing is the bone of contention. This was the residential part, according to Naidenova, because it has a colonnaded porticus, the rooms are more spacious and comfortable, and a hoard of silver coins was found in 16. It is certainly remarkable that the colonnade terminates before covering the supposed row-houses – or did it? But comparison with British villas may extend to the different widths of ranges; for at Chedworth and Spoonley Wood the best rooms are the narrower, and the same appears true of Iskar. Like reverse symmetry, different range widths form part of a code which no doubt applied in Bulgaria as in Britain.

FORMS OF COURTYARD AND FARMYARD Most kinds of yard plan found in western countries appear not at all or very rarely in south-east Europe. While it is hardly surprising, in view of the smallness of so many houses and the fewness of the outbuildings, that signs of regular geometrical planning are rare, even small irregularly shaped enclosures of the kind so common in BadenWürttemberg have not often been discovered. Insufficient exploration of villa sites caused by a combination of lack of interest in the villa as a farm and lack of government 212

— The Villas of South-East Europe — resources as compared with Germany or France must account for much of this blank in the record. Nevertheless, it is strange, in view of the determination implicit in the extent of digging on sites where the lack of mosaics, hypocausts and baths would have discouraged many west European archaeologists, that boundary ditches or banks were either not recognised or not reported, because it is impossible to believe that none such existed.

RECTILINEAR YARDS A few villas have the buildings laid out in so regular a fashion as to imply an equally regular boundary wall or enclosure. One Hungarian villa, Szilágy-Arnyoldal, has a small scatter of buildings reminiscent of Hungen-Bellersheim (Hessen) (Fig. 43), but no enclosure was found. Buildings are laid out parallel at Stolac [6], 24 Szentkirályszabadaja-Romkut, Hobita (Fig. 60) and the early or north villa at Kaisersteinbruch. Nevertheless, plans like these may be accompanied by an only partially rectilinear or geometrical yard, as Hobita shows; conversely, buildings were not necessarily sited parallel to the walls of a (nearly) rectilinear yard, as at the larger villa of Regelsbrunn (Aus.).25 Enclosures which are rectilinear in part only occur several times. Two and a half sides of Kaisersteinbruch look as if a square yard was intended and then the idea was abandoned and the wall completed irregularly (Fig. 60); Lisic?ici 3 (Yugosl.) (Fig. 60) and Hobita give the same impression. It is not obvious why these yards were completed as they were, and what was gained by the change of shape. Whether Deutschkreuz (Aus.) had a truly rectangular yard, as the part uncovered suggests, cannot be known; one point telling against the idea is that the porticus of what is taken to be the principal house has, running obliquely from it, a wall reminiscent of that at Bad Homburg (Hessen) (Fig. 9), as if the yard had been divided. A few of the largest villas have correspondingly large walled enclosures. Nemesvámos-Balácpuszta has inner and outer courts with sides about 200 m long overall. Spanning the wall between them is a large bath building. It must have been intended to serve people of not too dissimilar status living in both yards, for otherwise its siting is pointless. A second house was not found, no doubt because little of the outer yard was explored, but the baths imply some kind of dual occupation of the villa. At Bruckneudorf, in a courtyard as large as this, there is no apparent concern to create a setting for the grand house which, strangely according to the standards of French or German villas, faces a corner. It is curious that so little attention was paid to the relation of buildings to yard in these large villas when the much smaller Bulgarian villa of Chatalka-Delimonyovo Niva is set out with great formality. In its rectangular yard all the buildings are aligned with the sides and the principal house dominates the whole from a position not often paralleled in western Europe, very close to and in the middle of a long wall.26


— Chapter Twelve —

COURTYARD VILLAS AND PERISTYLES Bulgaria has a number of villas built around a courtyard which is far too large ever to have been roofed. Sofia-Obelja, on a trapezoidal plan, differs comparatively little from a British courtyard house. Pleven-Kailuka has a small inner courtyard and a larger outer one, Kadin Most probably the same but with the buildings set out in a more disjointed way; these two have a less familiar look. A local form of courtyard villa has a square enclosing wall against which are buildings little wider than the ranges at Basse-Wavre and, freestanding, a wider hall-like building with two or three rooms. Mogilets is one, the architecturally more pretentious Orlandovtsi (Fig. 60) another.27 Like some other villas of the region, the latter gives the impression of applying features familiar in western contexts in an unfamiliar and even, as with pavilions, in a slightly bizarre way. The porticus-with-pavilions theme was evidently known at some level but whoever commissioned Orlandovtsi dispensed with the element making it so attractive, the porticus which, whether colonnaded or not, by its contrast of height and bulk created a characteristically Roman appearance. Only pavilions, two square, two round and all of markedly different size provide a variation on the four square pavilions of Manderscheid (Rhld-Pf.) and other villas and convey the different standing of their users.28 The distinctive asymmetrical plan of the freestanding building at Orlandovtsi is paralleled at Ljus?ina (Yugosl.) (Fig. 60), where it is almost freestanding. Apsidal asymmetrical buildings like these have been interpreted above as basilicas and these two certainly bear a remarkable resemblance to one at Badajoz (Fig. 48), but that it perhaps too grand a title and function to apply here, or at least to the otherwise modest Bulgarian villa. A virtually identical building at Chatalka-Lambata (Bulg.) adjoins the baths and is interpreted as tepidarium (the apsidal room) and vestibule (the small room adjoining), which, indeed, they may be, but a tepidarium of this size (10×6 m) for an otherwise quite small bath suite seems disproportionate.29 Perhaps it could have accommodated the socialising connected with bathing. This may be another instance of the Roman architectural repertoire being transmuted in a Balkan context. The actual shape of a courtyard seems to be less important for the siting of buildings than in Germany or Britain, so that, for example, Salzburg-Liefering (Aus.) is the nearest place where the principal house is tucked away in a corner like some Rhenish villas. To this generalisation Sofia-Obelja provides an interesting exception in having, around its trapezoidal courtyard,30 four ranges of different width which resemble the differences found at Chedworth, for example, and which form a social hierarchy. The most important range seems to be the narrowest, the west. Its porticus is entered from the yard by a wide opening leading to an equally wide doorway into a shrine room, and there are smaller domestic rooms beside it. The wider and larger rooms of the north range are reached by two openings from yard to porticus of different widths to show their relative status – like Maulévrier; and to south and east the rooms are not of sufficient importance to need a porticus. Two entrances from the yard into a porticus would not have surprised anyone familiar with North Leigh (Oxon.), nor the shrine room anyone who knew Newport; and Chedworth has already been mentioned. Panik (Fig. 60) shows an interesting variation on the double 214

— The Villas of South-East Europe — entrance theme with two porches of different size. The larger one led to the domestic rooms, the smaller to the inferior side and the baths, but the latter seems also to have provided access to the room opening off porticus 19, which has the size and wide entrance of a representational room; and further along the porticus is a room of cruciform shape suggestive of a non-domestic purpose. If this analysis of the component parts of a Balkan villa is valid it shows that the social structure underlying it was not so very different from that revealed by analysis of west European villas. Peristyles will be dealt with briefly because they require more detailed study than has been possible. Several large villas of this type are known in Hungary and Bulgaria but not in former Yugoslavia or Romania. Most are the product of more than one phase of building with Hosszuhéteny (Hung.) as a probable exception. Its unique octagonal pavilions, which were evidently built to impress, are an instance of the change familiar features underwent east of the Alps. Tac-Fövenypuszta (Hung.) may also have been built on a peristyle in I or alternatively was begun to such a design and only completed in II. The row of five rooms at the north end implies a design of that kind; after the addition of three apses in echelon in II this villa had a certain likeness to the Spanish villa of Cuevas de Soria. A much bigger peristyle at Madara (Bulg.) is unusual in being surrounded by ranges of double depth throughout except for a square representational room of about the same size as that at Blankenheim. Madara, though, is highly unusual in having, outside the wall which surrounds it quite closely, several buildings, some large, arranged more or less in rows. In this respect it resembles the big Hungarian villa of Keszthely-Fenékpuszta which, with its fifteen subsidiary buildings and surrounding wall studded with round towers, forms a ‘fortified town-like settlement’.31 Western parallels for villas like these simply do not exist. Montrozier-Argentelle, for example, has as many buildings as Madara but they are set out in lines more like a palace, while very large villas such as Mettet-Bauselenne (Belg.) or Orbe (Switz.) convey much more sense of a grand house dominating regularly set-out lesser buildings than the more sprawling and apparently orderless layout of the Bulgarian and Hungarian villas. Without necessarily denying that the latter are villas, their development and perhaps their beginnings must differ sharply from anything in western Europe.

FORTIFIED VILLAS? Keszthely-Fenékpuszta was certainly fortified. Its walls are strengthened by towers of real defensive value, some 10 m in diameter and far removed from the toy towers (about 1.5 m diameter) which embellish the perimeter walls of Italian villas at Sette Finestre or Cosa. The problem with Keszthely is not whether it was fortified but whether it was a villa. Several other villas said to be fortified have been claimed as such through the confusion of pavilions with corner turrets. In this regard Romania provides a special source of confusion in the fortlets of the type called the quadriburgium, of which Gornea may be the best example.32 It has a square plan of much the same size as Orlandovtsi and has four towers wrapped around the corners, each with a doorway 215

— Chapter Twelve — into the central space. The latter is devoid of buildings but is entered by a big military gateway. The relevance of such obviously military buildings to Romanised farms is questionable. Only Bistrica (Bulg.) Has anything like this treatment of the corners, but it is so much smaller, about 14×14 m – the size of the smallest castellum on the frontier defences in central Germany – that its defensive value is doubtful. Possibly it was a police or signal post, only partly roofed. Those appear to be possibilities, but since a Bulgarian scholar thinks fit to classify this ambiguous site as a villa, Bistrica has been accepted as such (above, p. 200).

CONCLUSION The villas of south-east Europe show marked differences of form from those further west even when, as happens in a considerable proportion of them, the elements – domestic units, workhalls, secondary workhalls – recur at both ends of the continent. That suggests a generally similar social structure throughout Europe which found somewhat different architectural expression while drawing on the same body of ideas. Why that should be can only be answered from a much wider body of evidence and by relating it to the general problem of how ideas in material culture were transmitted.







t is generally accepted that virtually all villas in the European provinces of the Roman Empire were built and occupied by indigenous inhabitants. Why did they choose any particular one of several forms of villa? Various factors must have influenced choice: cost, obviously; the range of forms available; and suitability to the social circumstances and aspirations of the builder. Cost, which is principally in labour, is at present beyond conjecture and the availability of house types will be discussed later (chapter 16). A satisfactory understanding of the third point would entail knowledge of family structure, number of dependants, numbers of servants or slaves, and the pattern of hospitality. It goes without saying that much of the requisite information is hard to come by, yet a beginning can be made by considering the kinds of house in use on the eve of the conquest.

THE EVIDENCE: LIMITATIONS AND PROBLEMS To reconstruct from settlements the social structure of a country under Roman rule is hard but not as hard as understanding its native origins. Many villas have been excavated because a farmer could not, until the 1950s, ignore the difficulties masonry footings create for the plough; any earlier open settlements went unnoticed unless some eye-catching object were unearthed and consequently the body of evidence is much smaller. Germany illustrates a discrepancy in the kinds of evidence available; on the one hand the large number of highly visible cult earthwork enclosures called Viereckschanzen, on the other a paucity of the farms which must have accompanied them. A survey of twenty-five years of excavation on Iron Age sites mentions only a single open settlement, one at Bad Nauheim which was a centre of salt production.1 Even on the most carefully dug open settlements houses have proved elusive, either mutilated by later structures or their fugitive traces indistinguishable in black soil. Thus a survey of the Celts in Baden-Württemberg published in 1981 speaks of this kind of site only in the most general terms,2 and annual summaries of the intense activity during the following decade have thrown less light than might have been hoped for on the immediate predecessors of villas. Nor is the exploration of open 219

— Chapter Thirteen — settlements in Hessen any more advanced,3 and it is tantalising that no late preRoman Iron Age settlement within the frontier has yet been excavated and published completely. Much the most important body of pre-Roman house types has been excavated over the past forty years at the hill-fort of Manching and recently a major study of their measurements and design principles has been published.4 The situation is a little better in Britain and Belgium and much better in Holland, where fuller information is available than anywhere else and with it an unparalleled range and depth of discussion of early house types.5 In all three countries sites have been excavated which begin before or about the time of the conquest and continue until Romanised to greater or less degree. Among them Rijswijk-de Bult stands out as a model of meticulous digging and interpretation and, although Britain can show nothing on this scale, work on house sites and particularly discussion of their interpretation have been enlightening. In recent years French excavations have produced reliable evidence and useful syntheses, accompanied by destruction of the ancient myths about those bottoms of huts (fonds de cabanes) which for so long obscured understanding. As for other parts of the Empire, little seems to be known about immediately pre-Roman houses, even from such widely quoted sites as BaselGasfabriek (Switz.). The strictly pre-Roman evidence can be supplemented from sites with indigenous kinds of house, even though they fall within the Roman period: sites, that is, where the material effects of the new rule are confined to portable objects or the reorganisation of yards and enclosures. Moreover, imprecise dating, coupled with uneven geographical coverage, makes it hard to know, in most countries, just what the houses immediately preceding villas were like on the eve of conquest. Germany illustrates the difficulties. Many excavations have been conducted in that country, especially in Baden-Württemberg and Bavaria. All are reported according to the Hallstatt/La Tène sequence, and only a specialist could redefine their chronology on the lines proposed by Collis6 to show which settlements are most relevant to the present concern. Moreover, because occupation levels have been destroyed, houses have to be discussed almost entirely in structural terms: an unsatisfactory situation occasionally bedevilled by wild interpretations of soil traces that have confused matters further.

TWO-AISLED HOUSES: FORMS AND DISTRIBUTION One widespread type of house has a rectangular plan two or three times as long as it is wide and, on the long axis, the principal element of roof support, comprising a number of posts rising to a ridge-piece to which the common rafters were attached. Since it has no generally accepted name in English the readily intelligible continental term ‘two-aisled house’ will be adopted. Functional interpretation of long rectangular plans has to conform to certain limitations. A point made earlier has to be reiterated: if the entrance is in the middle, the building is virtually divided into two parts; if near one end, the part nearer the entrance becomes to some extent a passage. The considerable variations of wall structure and size of ridge-post recorded in buildings of this type do not affect interpretation of the plan. It is important, as students of vernacular building learned long ago, to treat plan and structure as 220

— The Late Pre-Roman Iron Age Background — independent variables, even in archaeological contexts where both superstructure and plan have to be derived from ground-level evidence. But the roof supports introduce further possibilities. Where the internal posts are closely spaced they are likely to have been incorporated in any partitions that may have existed, and themselves tended to break up the unity of the living-space if it occupied more than one bay of the structure; furthermore, the lack of evidence for partitions, whether or not caused by the loss of occupation levels, does not preclude the drawing of inferences about functional division from differences in bay rhythm. Germany can show a wide scatter of two-aisled houses. Two at Haunstetten (Bay.) (Fig. 61) near Augsburg are among the few firmly ascribed to the Late La Tène period; neither produced hearths or other evidence of how they were used. At the other end of the country the early Roman site of Bedburg-Garsdorf (Nordrh.Westf.) (Fig. 68) has, where a workhall might be expected in a fully-fledged villa, an un-Roman-looking double square building (11×22 m) with four hearths all close to the ridgeposts. Another large ridge-post house among several in a Hallstatt settlement at Eching-Neufahrn (Bay.) has a floor area of 130 sq. m; and the UrnfieldHallstatt settlement at Eching-Autobahn (Fig. 61) produced both house 16 of 16×7 m and the exceptionally large house 1 (21×9.6 m), which was divided into two principal rooms. For the Low Countries we have fuller evidence, albeit largely of the early Roman period, and for the Meuse-Demer-Scheldt area it has been collected.7 The series begins c. 300 BC (Middle Iron Age) at Haps (Neth.) and continues with the OssUssen (Neth.) type, defined by structural characteristics, which appears in the first century BC. Not long afterwards the Alphen-Ekeren type, defined in similar terms, appears. Both persist for a long time and a settlement like Hoogeloon, spanning two hundred years, produced as many as thirty examples. Houses in both groups have an average width of 6–7 m, range in length from 12 to 20 m, and reveal little evidence of how they were partitioned or used. Later still is Beegden 1 (Fig. 61). How widely two-aisled houses are distributed is uncertain. Britain has nothing of the kind that is definitely pre-Roman but, just as the later two-aisled stone building at the villa of Bocholtz-Vlengendaal (Neth.) (Fig. 46) shows its affinity with those described above, so the incompletely explored Building 6 at the Rudston (Yorks.) villa, combining stone walls with a ridge-post, suggest that timber precursors await discovery. A very uncertain example, smaller than those in the Netherlands, may occur at Fishtoft (Lines.) (Fig. 7); it is 11×4 m with clay walls, had a hearth at each end, and looks as if the walls have been considerably rebuilt. It appears to be the largest house of a group including a smaller rectangular one and a circular hut.8 An entirely illusory building of this kind at Pilsdon Pen (Dorset), alleged to be no wider than the Fishtoft house but at least three times and implicitly five times as long (32 m, 55 m!), deserves mention as a curiosity of archaeology.9 It was rightly compared by its excavator to a cross-slotted palisade ditch at BundenbachAltburg (Bad.-Württ.).10 Of the few examples from France two may be cited. Thoraise (Doubs) has a fourbay building of very regular ridge-post and wall-post spacing, and Vix (Côted’Or) is an equally regular two-bay outbuilding.11


— Chapter Thirteen —

Figure 61 Aisled buildings

TWO-AISLED HOUSES: INTERPRETATION What can be said about the uses to which this type of building was put? Only the evidence from the Low Countries is full enough for serious consideration. It can be divided into two main categories, those houses with clear signs of accommodation for cattle or other animals and those without. Despite the lack of positive evidence in the second and more numerous group, those at Haps have, nevertheless, been assimilated to the first group on the basis of their division into two unequal parts. By 222

— The Late Pre-Roman Iron Age Background — analogy with two- and three-aisled houses elsewhere, the larger part served as a byre or (a significant qualification) ‘conceived of more broadly, as a work-room’.12 This may be true but it is not accompanied by discussion of the bay systems of the primary roof supports, the ridge-posts, which require analysis. The largest and best preserved building at Haps, house T (Fig. 61 ),13 has five ridgeposts, one of them exactly in the middle of the building, with a bay rhythm of cacbbc, a being the biggest. The logic of this is that the middle bay c corresponds to the width of the two opposite entrances and demarcates a cross-passage, whether defined by structure or customary use. Next to (west of) the entrance bay the longest bay of all has the characteristic openness, free as far as possible of encumbering posts, that has been remarked in the principal living area of three-aisled domestic buildings of all kinds. How the rest of the building – the lower end, surely – was used its bbc rhythm does not make clear, but the greater number of posts hints at a byre, as in three-aisled houses. One other point: given that both end bays are short, the closeness of the end ridge-posts to the gable-end walls14 must relate to the structural necessities of a hipped roof. In general the end bays are so short that they can only have performed some minor function subsidiary to that of the adjoining larger bay and have to be disregarded in a functional analysis. House T is not typical. Only in one other (V) of twenty-one houses does an axial bay correspond precisely to the entrances and in seven it is slightly longer. In the remainder the house is entered in a longer, and sometimes the longest, bay. All but one house (B) has a ridge-post aligned with the door-jambs nearer the upper end and forming a barrier, probably structural rather than customary, between entrance and upper end. Why was another barrier not thought necessary between entrance and lower end? And why did the lower end, in ten houses, have a bay as long as that in the upper end or longer when indubitable three-aisled byre-houses retain short bays at that end? No doubt the notion of a workroom may account for this situation, and the ambiguity about function was repeated in relation to a house of ‘the beginning of the Middle Iron Age’ at Zijderveld (Neth.) (Fig. 61),15 There, where the lower end bay was double the length of those at the upper end, ‘the intention . . . could only have been to create here (in the storage or cattle area) as much room as possible’.16 If this is true it contrasts sharply with the care taken in the big combined shelters of north Germany to divide the livestock area into stalls. Paradoxically, houses with both more and fewer ridge-posts than the two just discussed present greater difficulties of interpretation. A longer and narrower house at Oss-Ussen (Fig. 61), more lightly built and with a less regular bay spacing,17 is linked to Haps T by having, exactly in the middle, one of the two posts defining the shortest bay, next to which is the longest bay. That the former was the position of the entrance is confirmed by slight indications of extra posts in both long walls. Here, though, spatial organisation is more complicated because the building has a long bay at each end, hence the presumed domestic end comprises the longest and one of the next longest bays as well as the entrance – the equivalents of a hall and inner room, perhaps, and divided by custom rather than a partition. The lower end is more problematic. Whether or not the longest bay needed an open roof truss, the spacing of bays points to some differentiation of function more complex than that at Haps T or Zijderveld. 223

— Chapter Thirteen — Two-aisled houses with this number of axial posts are exceptional. A commoner form, of varied length and sometimes bigger than any of those so far mentioned, has wide-spaced posts making, usually, three or four axial bays. This gives spacing between posts of up to 6.5 m, and as de Boë recognised, intermediate tie-beam trusses carrying king-posts were needed.18 Two points arise from this: first, the roofsupporting structure itself gives no hint of any differentiation of use in the building, nor of the position of the entrance; second, some important purpose must underlie the clearing away of posts by a comparatively complicated technique. That purpose was present already in the first century of our era at Hoogeloon, where the houses have less in common than might be expected with those at Haps. All have ridge-posts in the gable-ends, so floor space throughout could be used more effectively; and only one house of thirty has the post in the middle characteristic of Haps. And there is a tendency for bays to be more equal in size, so that whereas Haps has only five examples with two bays of the largest size a, Hoogeloon has six and three more with three.19 Yet, even in the solitary case of house 22, comprising only three equal bays, the disposition of ridge-posts in such an apparently logical way can hardly have a purely structural explanation because the bay length of 6 m demands intermediate, and technically more difficult, support for a ridge-piece weighed down by rafters and cladding. Variation of bay length in a house requires explanation and, because structural considerations alone demand regular spacing, it has to be sought in functional terms. The obvious division, between living-space and byre, is difficult to apply, given the varying proportions between the two elements in those houses at Hoogeloon where a difference could be recognised: the byre may occupy between a fifth and a third of the whole. Moreover, not a single byre matched a bay length exactly, although the difference might in every case be accounted for by an entrance passage or feeding walk; nor is the relative length of the byre bay constant, being neither the largest nor the smallest. That said, there can be no doubt that small differences of length were important, as the babab rhythm of house X, with no more than a metre as between one bay and another, shows. Explanation is further complicated by a temporal difference in use, the byre-houses being later. The simplest case is presented by two houses (5, 8) which have a single internal post forming two unequal bays differing by 2 m in length. This difference may be related to the point of entrance: whereas the short bay in the middle of Haps and OssUssen is demarcated by a post on each side, here there is only one, forming, it can be assumed, a quasi-passage entrance in the longer bay. The possibilities of these elongated houses are shown by the workhall at BedburgGarsdorf (Fig. 68), a larger building than those so far analysed and no earlier than the mid-first century. It has no less than eight bays, and the post marking one side of the shortest is here not exactly in the middle but about a metre to north; here, by analogy, was the entrance. It was not a passage; on the far side of the two posts was a storage cellar, presumably lined with timber. There were four hearths, each standing beside a ridge-post, not in the bay between them as might be expected. It looks as though two important ones were near the end of the building, to northeast and south-west of the axis, and a third is near the middle on the west side. Possibly this throws light on the Oss-Ussen house, where also two important bays are at the ends, with another to one side or other of the entrance. 224

— The Late Pre-Roman Iron Age Background — This use of Roman evidence to help elucidate pre-Roman houses is undesirable, yet the similarity of the plans and their unlikeness to villa houses may justify this provisional interpretation until fuller early evidence is available. Arguments derived from English medieval halls may also be thought suspect, yet anyone familiar with houses of the period knows how closely varied bay lengths correspond to the relative importance of parts of the hall and – something not recoverable in Roman buildings – how the position of roof trusses reinforced that message. Two-aisled houses with their very varied bay lengths contrast strongly with threeaisled byre-houses in which the variable post spacings of the domestic end are combined with a regular bay rhythm in that part accommodating cattle: a clear and consistent structural difference corresponding to evolving domestic needs at one end of the house and an unchanging agrarian function at the other. The difference between the two patterns may be due to the adaptability of the roof supports in the three-aisled house to the creation of cattle stalls on each side of the middle aisle, something which is lacking in the two-aisled house; yet ridge-posts could certainly have been used to the same end, given suitable wall height and a feeding walk down one side. Why a consistent structural form never evolved in two-aisled houses to meet a constant need is puzzling. It may be that the adaptation of the house type to accommodate a byre, which must have been a response to new circumstances, occurred piecemeal and over what was, by comparison with the long history of three-aisled houses, a short period, so that there was insufficient time to develop a structural response. Two considerations are relevant here. The first is that a satisfactory way of keeping cattle indoors must have been found, whatever the real or apparent structural limitations. The second is that, although traditional building techniques do change in response to needs, roof construction in particular is very slow to do so, and adaptation to anything other than strictly structural purposes would have been difficult. More surprising is the lack of any definite structural demarcation of an entrance passage on the lines of Haps T. Given that some houses have large bays at both ends, it may be that some other use than a byre has to be envisaged, as is the case with the aisled houses of Roman Britain (above, pp. 36–8). If the sequence of the houses were established, a development from more to less regular rhythms might be observable, and certainly the Haps houses appear to have more consistent bay lengths than those of Hoogeloon. Such variation in bay length indicates very fluid socio-economic conditions. The Dutch types may have a wide distribution. The larger of the two Late La Tène houses at Haunstetten (Fig. 61) has, in the middle, a large axial post-pit and, about 2 m from it, another; together they defined the entrance, and the resulting bay rhythm is bacab, with the two presumed living areas about 5 m long by 8 m wide. A faint hint of a similar arrangement can be seen in the other house. Other German two-aisled houses can hardly be said to convey even a hint, although a number of those of only two bays have the intermediate post and the corresponding wall-posts set markedly off the middle, for example at Eching-Autobahn House 13 (Fig. 61) or Dietfurt a.d. Altmuhl (Bay.). Probably they retained an entrance in the middle without its being defined structurally on one side. 225

— Chapter Thirteen — An intermediate development between the two- and three-aisled types is found in the largest building of the Late Hallstatt settlement at Königsbrunn (Bay.) (Fig. 61, D), which combined a ridge-post in the domestic part with pairs of internal posts forming a three-aisled construction in the byre. It occurs also at Oss-Ussen20 in the Roman period. These houses show a clearing away of intrusive axial posts in part of the building where they were inconvenient. In the living-part a similar result was achieved by moving the two ridge-posts the whole width of the building apart and so creating a square living-space.

THREE-AISLED HOUSES More remarkable than anything so far described are two three-aisled buildings which have only their structural form in common. The first, a hall of the Early Hallstatt period at the Heuneburg (Bad.-Württ.) (Fig. 61), has the imposing width of 16.5 m and was at least 20 m long, with a clear middle span of 9 m that necessitated some advanced form of roof construction – all this at c. 500 BC. Its size, the absence of partitions and the complete clearance of axial posts point to its being a representational building, not simply a large house. The second, an aisled long-house which was the principal building at the small hill-fort of Befort-Aleburg (Lux.) (Fig. 61), belongs to the Hunsruck-Eifel culture and is perhaps of 400–300 BC; it was 31×8.8 m, all three aisles were of equal width, and the living-space occupied about a third of the total length. It is one of the comparatively few pre-Roman house in northern Europe to have a fully intelligible ground-plan. For these two houses I must accept the dates given by those who dug them and apparently approved by prehistorians generally. Nevertheless, the 9 m span of the middle aisle of the Heuneburg hall and the stone footings for the internal posts at Befort are unparalleled at their respective periods and indeed for perhaps another three hundred years, so I endorse the conclusion drawn by another writer considering these buildings from a structural standpoint: that in the present state of research they form such an exception that parallels must be awaited before they can be given a secure place in the history of central European houses.21 Other claims for threeaisled buildings at Welwyn-Lockleys (Herts.)22 and Colchester-Camulodunum (Essex)23 are without substance. A three-aisled building found at St Michael-Gorhambury may be pre-Roman but is not provably so.24 Whatever its precise date it was an important building, the largest in the whole complex, and comparable in status, though not in detail, to such aisled byre-houses as 9 and 4 at Rijswijk IIA. It would certainly not have accommodated ‘workers’ families’,25 whatever that term may be supposed to mean around AD 43. Halls claimed to have been aisled occur in France at a much earlier date, notably the Middle La Tène house at Verberie (Oise) (Fig. 61) which was 21.2×11.5 m overall and had aisled structure only near the ends; its width and the apparent lack of internal roof supports in the middle space 12.5 m long present problems not so far solved satisfactorily but do not diminish the importance its size implies. Earlier still, Chassemy (Aisne), of the Early La Tène period and regarded as an aisled building, 226

— The Late Pre-Roman Iron Age Background — has been subject to doubt,26 and certainly the roof-supporting structure with a middle span no wider than the side aisles is problematic. By far the best evidence for three-aisled buildings comes from Holland, mostly of the Roman period. A long series beginning much later occurs at Rijswijk and will be mentioned below (chapter 14); otherwise most lie outside the Roman frontier.

SQUARISH AND TRAPEZOIDAL BUILDINGS Aisled buildings have attracted much attention because of the intrinsic interest of their plan and structure. In some parts of Europe, though, aisleless buildings were the norm but because, in so many, little remains except a much less informative pattern of post-holes or stone footings, less notice is taken of them. Small square timber buildings have been excavated in Britain, northern France and Holland,27 most of them revealed simply by four post-holes. They are generally interpreted as granaries on account of their smallness and the lack of any sign of occupation, and no doubt that is correct in many instances. Exceptions to this general rule have been found in the hill-forts of the Welsh Marches where, at Croft Ambrey (Herefs.) (Fig. 62) and particularly at the Wrekin (Salop) (Fig. 62), hearths were found in several squarish four-post structures, some of them quite tiny, their sides ranging from 1.8 m to 3.6 m long.28 They are of the fourth century BC. Size notwithstanding, it is difficult to gainsay a hearth as proof of occupation, and although these houses are small their interpretation is strengthened by the discovery of not much larger stone-built ones in Languedoc. Martigues-L’Arquet 8 (Bouches-du-Rhone), a trapezoidal house of c. 450 BC whose longest side was 3.5 m, had an area of about 11 sq.m – less than the 13 sq.m of some of their British equivalents – and it may be compared with 5, 10a and 14 (Fig. 62) or a house at Mailhac-Cayla (Aude).29 Their

Figure 62 Square buildings


— Chapter Thirteen — size and hearths, in Britain as in Languedoc, show that they are the houses of conjugal families. Despite the wide date range, which eliminates them as the immediate forerunners of villas, they throw light on the minimum amount of space required by a family, that is, a conjugal unit or unit of consumption. Some timber houses found at Braughing-Skeleton Green (Herts.) (Fig. 62) belong to this category if the very incomplete evidence can be trusted. The most certain were 3, about 6×4.5 m, and 7,30 with ridge-posts, about 5 m square. Slightly larger are three at Eching-Autobahn, all with ridge-posts; another at the Heuneburg lacked internal posts but had a central hearth. A few larger square buildings are known: one at Danebury (Hants.) (Fig. 62), almost divided by axial posts, and another at Hornchurch (Fig. 62) are 9×9 m, and at the latter site is a second one, slightly larger. Among the great variety of houses at Bundenbach-Altburg is a square house (17) of this order of size (8×7.5m). It is earlier than the English examples, belonging to the middle phase, c. 130–80 BC, but lacks ridge-posts and, like others of its kind, presents, as its excavator remarked, considerable problems of roof construction.31 So, too, does the best-known of all these square buildings, Mayen I, a house singularly lacking in sizeable structural posts. Allied to this type by its wide span is a bow-sided variant, the Late La Tène post-built house at Landshut-Sallmannsberg (Fig. 62), which is 10 m square but may possibly have been longer because one wall was not found. A clear span as great as that of an average German hall-type villa demanded roof craftsmanship of a high order, must have created an impressively lofty internal space, and demonstrates the comparatively elevated status of such a house. Some larger square buildings with a row of ridge-posts could be classified as twoaisled houses, but the shape is a more important characteristic. A square room or building, normally a spatial unity, is to an extent divided by ridge-posts, as anyone who has seen a two-aisled Gothic church knows, yet the analogy must not be pressed too far, for the timber building lacks the visual block presented by a stone arcade. Moreover, this kind of square house has no obvious point of entrance, unlike all elongated plans which, whether a doorway is postulated in the long or the short sides, permit logical consequences about use of the interior space to be drawn from the initial assumption. For square two-aisled houses, entrance remains an intractable problem but some customary division seems likely, whether between work space and living-space, inner and outer ‘rooms’, generations of an extended family, or whatever. One such house, Beegden 2 (Neth.) (Fig. 61),32 associated with a slightly more rectangular and probably superior house, is thought, partly by analogy, to be of the first century AD. The relation of the two buildings is reminiscent of that between two round-houses, one large, one small, sometimes found in Britain, or that of a small row-type house with an aisled workhall, as if they represent something fundamental about social organisation over a long period of time.

MONOSPAN OBLONG BUILDINGS Buildings of this kind are rare on Dutch and German sites and small by comparison with two-aisled structures, so that at Eching-Autobahn, for example, where half a 228

— The Late Pre-Roman Iron Age Background —

Figure 63 Monospan halls

dozen of them do not exceed 8×4 m, they were not very important. In Britain, where evidence is usually fragmentary and understanding has been clouded by highly dubious structural interpretations, they do nevertheless occur more than was once thought. One of the best is Babworth-Dunston’s Clump (Notts.) (Fig. 63), built in the first century AD; the principal entrance suggests a division into east and west parts, the former about three times as long as the latter and each having its own gable-end entrance. Buildings of much the same size and of immediately pre-Roman date which have been found at Wickford (Essex) and Canterbury (Kent) are so incomplete that they add little to knowledge. Some more complete small buildings at Silchester which lack evidence of how they were used may have been houses. The few aisleless buildings found in northern France – excluding an improbablelooking one of L-shape at Suippes (Marne) – are for the most part small. Several stonebuilt ones in Languedoc33 are only 4–5 m wide; those with an undivided ground plan are up to 9 m long and the bigger ones are invariably of two rooms. One of the rooms usually has a hearth, the other not; the former is the living-room, the latter a store room. Monospan houses are much rarer in Holland than those with one or two rows of internal posts and belong for the most part, like Kaalheide-Krichelberg (Fig. 63), to the period shortly after the conquest.34 Some, like those at Kethel, are byre-houses. Another type, the monospan house divided into two or more rooms, is rarer still; nevertheless, its mere existence in a house landscape of aisled buildings and open and largely unpartitioned halls demands explanation. Niederzier-Hambach 59 comprises a square hall, entered through opposite porches, and a second room of about the same effective living-space which appears to have been entered from it. Pending full publication and discussion of this house, Rijswijk (Neth.) IIB 2 (Fig. 65) is a more informative example of the kind. It is large (24×6.5 m), aligned east–west, and of three bays about 7, 9 and 7 m long; the middle bay has a hearth placed towards the northwest corner and all three bays were connected by doorways on the southern side. An entrance at the west end is likely from its width, about 2.3 m, to have been for cattle; people entered on the north side, almost in front of the hearth. Most unfortunately for interpretation, the west room was poorly preserved and much about it, particularly whether it was for cattle or storage, remains uncertain. House 2 is as near an approach to the compartmented house of a Roman villa as has been found in an 229

— Chapter Thirteen —

Figure 64 Grouping of houses

indigenous settlement anywhere, yet, even granting that Roman influence was strong in this phase of Rijswijk’s history, in plan and construction alike it is of indigenous inspiration.35

ROUND-HOUSES Round-houses of a variety of structural types are all but universal in the Late Iron Age settlements of the British Isles. Those in the Thames Valley average c. 9 m in diameter,36 and the tendency seems to be for round-houses to get smaller during the Iron Age. Usually a settlement stands within an enclosure formed by a palisade or a bank and ditch, and comprised one to three houses at any one time. Tollard Royal (Dorset) had a single house a little over 5 m in diameter; Harding-Mingies Ditch (Oxon.) had a single much larger one (c. 15 m diameter) which was replaced four times. A small Iron Age enclosure at Draughton (Northants.) (Fig. 64) had within it three round-houses, two small and one large, a pattern sometimes repeated in the grouping of houses in larger settlements. Colsterworth (Lincs.) has a hint of it in the north-east 230

— The Late Pre-Roman Iron Age Background — quarter37 and at Pilsdon Pen three round-houses in close proximity are linked by gullies which are evidence of fences (Fig. 64).38 Groupings of this kind are appropriate to a kin-group larger than a conjugal or even a stem family and appear with variations on many sites. Outside Britain round-houses are comparatively common in Brittany and a scatter has been discovered over a large area of France from Aulnat (Puy-deDôme) to Achenheim (Alsace). Although these instances show that the contrast between Britain, where round-houses predominate, and the continent, where rectangular houses predominate, can be overstated, the extent to which the British tradition is diluted by rectangular buildings has been exaggerated by talk of houses built on a frame of sill-beams or even on the log-cabin principle39 – remarks indicating an unfamiliarity with timber construction. In the generation or so before the conquest St Michael-Gorhambury (Herts.) is important for the round-houses of the Late Iron Age phases IV–V, lying within the bracket AD 20–43. In IV a round-house (6) of medium size (7.3 m diameter) occupied an important position in the inner enclosure B, within which stood a second and very fragmentary house (11). The first rectangular building, three-aisled, appears in V and in VI, the immediate post-conquest period; other aisleless buildings replace all the round-houses. So, too, at Harting-Garden Hill II, one certain and one possible pre-Roman round-house preceded the rectangular buildings that appeared in the post-conquest period III. And the story is not very different at Radley-Barton Court Farm (Fig. 68): in [I], perhaps early first century AD and so approximately coeval with Gorhambury IV–V, a circular ‘hut’ 5 m in diameter within a small enclosure and a house just outside it are followed in [II], late first to mid-second century, by a timber hall some 28–30 m long and 8.5 m wide (Fig. 63).

COMPLEX HOUSES The houses so far described are all of quite simple plan, however complex the customary subdivision of space may have been. In a few, though, structural partitions reveal a more advanced functional or social division. Sigean/Pech-Maho (Aude) (Fig. 64) is remarkable for having a hearth in each of the three compartments into which the trapezoidal plan is divided.40 It is considerably bigger than the small squarish houses and appears to comprise an entrance space from which two elongated rooms are reached. Because it was built against the perimeter wall of the enclosure no light came from the south-east side; this may be why the front part is regarded as an open court rather than a roofed space or room common to both the others, but it is not conclusive. Two similar plans occur at Villeneuve-Saint-Germain (Aisne) (Fig. 64), where the houses are freestanding and thus do not raise any problem of lighting, minimal though that is likely to have been. Here the front ‘court’ has much more the appearance of a small porticus, which was more than simply an entrance, probably a workplace or common room.


— Chapter Thirteen —

CONCLUSION Several types of house found in western Europe after the Roman conquest which were built of stone or at least with stone footings are traceable to pre-Roman origins. Even that most un-Roman type, the British round-house, has some successors with stone footings. The one obvious absentee is the row type; nothing really like it appears in pre-Roman contexts unless Villeneuve-Saint-Germain or Rijswijk IIB be thought to prefigure it. Certainly in Britain, where row-houses are the commonest type in villas, nothing like this is known. Conversely, it is hard to find much evidence of Roman-period halls with ridge-post construction, and tempting to suppose that superior carpentry enabled builders influenced by Roman technology to adopt monospan roofs. Against this is the fact that, in certain parts of Europe, wide buildings, notably some square ones, had roofs which did not need freestanding posts for support and must have demanded quite advanced carpentry techniques. Simple square halls of one building phase, with no additional rooms, are also scarce, though not unknown, and can sometimes be inferred in multi-period structures. Thus, despite the difficulty of tracing a type back from Roman to pre-Roman times on the same site, in typological terms continuity is certain.





ust as the forms of house and the ways in which buildings are arranged in relation to one another vary in innumerable ways, so the transformation of an indigenous pre-Roman society into a more or less Romanised one is a process of infinite variety. Moreover, the various ways it was done are less easy to categorise because both the stages in the process and the end result could be very different. The period at which Roman rule (not simply conquest) became effective and its cultural values began to be accepted is relevant to the types of house available for adoption by subject peoples, and that applies not only as between one country and another but also within countries, due to uneven development. If the concept of kin-groups emerging, as it were, into a Romanised society be accepted, it follows that at that stage they would require some fairly simple house, but those building in the new fashion in the first century of our era would choose from a rather different range of types than those doing so in the early third century, even in the same country, simply because the types themselves changed somewhat over that long period. A further complication is that immediately prior to conquest the various peoples were living in societies at different stages of development and in different relations with the Roman world; consequently the degree of familiarity with Roman culture and receptivity to it also varied, and this factor must have affected the spread of house types. More will be said on these matters in the final chapter, but now some of the ways in which Romanisation came about will be described.

NATIVE TO ROMAN BY EASY STAGES: RIJSWIJK AND ST LYTHANS-WHITTON The long approach to Romanisation is best exemplified in the Netherlands, at Rijswijkde Bult, where the most important excavation since Mayen fifty years earlier has greatly advanced understanding of the social and economic processes leading to the building of a villa (Fig. 65). Settlement virtually coincided in duration with Roman rule in the country. Phase IA comprised a single aisled house and an adjoining enclosure and IB–D saw the addition of other houses scattered at some distance from the first, with their separate 233

— Chapter Fourteen — small enclosures; rebuilding was frequent. Formalisation began in IIA when the components of the settlement were reorganised into two intersecting trapezoidal yards for the two principal houses and an area outside, not wholly bounded by ditches, for two houses of less (and unequal) importance. It continued in IIB with a change from trapezoidal to sub-rectangular yards and the complete enclosure of the larger outer yard. In the latter, two houses were reduced to one and it was subdivided by a fence to create a temple courtyard common to the whole settlement. What may appear a small increase in amenity in this phase, the addition of a hypocaust to house 19, is of great importance. The mere existence of this badge of status established more clearly the superiority of house 19 to the others, and provided a strong reason for improving it piecemeal thenceforth rather than rebuilding it in a slightly different position, as had been usual hitherto. That is what happened in IIIA, when the house was rebuilt with stone footings and a form of roof construction designed to free the interior space of posts, and at the same time the settlement was unified by the building of a large trapezoidal enclosure in which four houses were set out in a hierarchy. The process continued in IIIB. By III Rijswijk had attained a modestly Romanised way of life and a recognisable villa layout which in its final state has a strong resemblance to Walsbetz (Belg.) (Fig. 31). There, too, a principal house is combined with two halls lying parallel and close together but with this difference, that the halls are further from the house and at right-angles to it in order to show their subordination.1 Perhaps the beginning of Romanisation should be seen in IIA, when each of the three principal houses was for the first time set within its enclosure and a separateness was introduced. That sense of formality did not, in material terms, present an obstacle to future reordering of the settlement in the same way as formerly, but in fact the decisive step towards making it permanent came in IIB when the first use of stone in pursuit of Roman luxury was accompanied by making two courtyards a more regular geometrical shape. It is impossible to know whether the kin-group consciously took an irreversible decision or whether its consequences only became apparent later; whichever was the case, the social basis of a villa had begun the process of fossilisation. A British villa with a longer life illustrates this theme. St Lythans-Whitton (Glam.) (Fig. 65) was limited throughout its existence by the circumstances of its immediately pre-Roman origin, when it was built within a squarish protective bank and ditch enclosing an area much smaller than that occupied by Rijswijk ID. Whatever the agrarian potential of the locality – and about the economy of the villa nothing is said in the report – the farm seems to have been ‘at all times a reasonably prosperous agricultural unit’2 – growth on the scale of Rijswijk would have been difficult. Probably, though, social and demographic factors were decisive in restricting growth. Whitton I probably has three round-houses. Two, with porches, face the entrance; the third, B2, in the innermost and thus superior position, turns away from it and was approached at a tangent, rather as early row-houses were approached from one end. In phases II–V the buildings were quite mobile, changing in number and location,3 and it is not easy to know which was the superior house. Unfortunately, the authors pay no more attention to social structure than to the economy and the reader is left to attack the problems unaided.4 The most important change in II was the addition 234

— Modes of Romanisation —

Figure 65


— Chapter Fourteen —

Figure 65


— Modes of Romanisation —

Figure 65 Stages of Romanisation


— Chapter Fourteen — of a large round-house facing the entrance; its position suggests it had a more public function than the smaller houses, perhaps in the nature of a workhall, and its mere existence argues for an increasing population. In III numbers appear to have fallen from that peak and in IV social change is signified by the division of the site into three parts, corresponding to Rijswijk IIA: an important house in the south-west enclosure, the most important one in the inner north-west enclosure, and a third sub-rectangular one and possibly even a fourth, very minor, in the outer yard. In V a complete change takes place with the fencingoff of a yard at the entrance, to limit access, and, within it, a small yard in front of a round-house. This successor round-house to the workhall-like one has some public purpose, otherwise it would scarcely have opposite doorways of equal size. Two squarish buildings now appear, facing each other in a way which expresses a relation not much different from that of the two halls (10 and 11) of Rijswijk IIIA. Unless that siting be regarded as incipient Romanisation, the first real sign of that is the building in VI of a small hall with stone footings, and the curious aspect of this development is a dramatic fall in population. At Whitton, as at Rijswijk, the sequence of changes from loose grouping through defined yards to a unified hierarchical grouping of houses can be interpreted as stages in the transformation of a kin-group or clan. Originally composed of equals working collectively under a head for the time being, it turns first into independent extended family units collaborating in the operation of a farm and then, by the emergence of one pre-eminent group among the three, into a hierarchy. Much the most puzzling aspect of the story is provided by the contradictory evidence of population decline and material advance in VI.

ROMANISATION BY LUXURY A development similar to that at Rijswijk is observable, though not in the opinion of the archaeologist who dug it, at Barnsley Park (Glos.) (Fig. 65).5 In the version preferred here, Barnsley Park is analogous to Rijswijk except that it began with three adjoining yards instead of expanding from one and, of course, the earlier buildings are round, not rectangular. Furthermore, the evidence, being of a very difficult kind – largely scatters of stones – made it harder to achieve certainty about the changes of shape the yards underwent. Even in I, the north yard had two buildings which are likely to have been houses and little change is detectable in II. By III the site seems to have been fenced off more sharply into three main and two minor yards, the three main ones each having two houses – the same population increase immediately before the adoption of stone for building as took place at Whitton. The transformation into a villa began in IV with the building of a bath suite at the end of one of the earlier buildings, a proceeding analogous in terms of both structure and luxury to the addition of a hypocausted room at the end of an indigenous type of building at Rijswijk. When did Romanisation begin at Barnsley Park? Is it observable in the replacement in III and IV of round buildings by more or less rectangular ones, or only when the bath block is built? Probably the latter, even though the change of building shape 238

— Modes of Romanisation — may well be of Roman inspiration.6 It is significant that the first part7 of a sizeable house to be put up was the bath building. By VIII, Barnsley Park, with its house at the head of a small courtyard and the subsidiary building to the sides, looked much like many another British villa. Baths provide the link with the Romanisation of a very different kind of site, the iron-working settlement of Hartfield-Garden Hill (Sussex) (Fig. 65). It was established within the ramparts of a small hill-fort and had, in its pre-Roman phase II, two round-houses. These were followed, early in the Roman phase III, by a hall (E) unusual for Britain in its proportions (3:2) and in having a proto-porticus or timber veranda running round its north and west(?)sides. It is coeval with another timber structure (C), a building on rubble footings measuring 9×3.3 m, proportions which suggest in the light of its siting and subsequent history that it functioned like a wing to an adjoining structure to the east. That seems certain to have been the situation in IV, when whatever had adjoined it was replaced by a hall c. 9×6 m, parallel with which was a small bath house. So when does the site become Romanised? A house with a veranda is not in the native British tradition of building, but if that is excluded, the bath house marks acceptance of the new cultural forms. At Hartfield the hall and baths correspond to building M in the romanised phase of Collingham-Dalton Parlours (Fig. 68) and house C is the latter’s J. Two houses facing in different directions are common in Roman villas, often principal house and workhall, and this arrangement may, like the hierarchical setting-out of Rijswijk IIIB, indicate social relations of a kind familiar in many provinces. Something similar may have been true of München-Denning (Bay.)8 where, in a maze of unintelligible post-holes and pits, a good-sized bath building was found, superior to any so far mentioned. Although the only other identifiable part of a stone building was no more than a pavilion and part of a porticus, it, a few other stone fragments, and some ditches appear to have been laid out on parallel alignments, as if a number of timber buildings partly rebuilt in stone formed a villa-like grouping. The first-phase Romanisation represented by a stone bath-house among the otherwise timber buildings of a native-style farm may account for some of the apparently isolated bath houses that long puzzled archaeologists; dug long ago, the post-holes around them were missed. Fanciful notions of a bath house serving a locality can be abandoned.

HOUSES BUILT OVER BOUNDARIES Many villas were built like Rijswijk-de Bult and München-Denning within complex systems of enclosures. Traces of earthwork boundaries and fences appear frequently and are commonly not explored fully because they are thought to be irrelevant. Sometimes the potential significance of a ditch line is not recognised and as a result it is not traced even within the bounds of a site. Once more Rijswijk provides a starting point for enquiry. When all its buildings are looked at together in a single plan,9 the general course of development based on three domestic sites is clear; successive rebuildings do not move far from the position of the first house on any 239

— Chapter Fourteen — particular spot and new construction is sited to avoid ditches. Why, then, are some other sites so different? Much of the evidence for building across boundaries comes from Britain, something which may in itself signify social differences from other provinces rather than superior excavation techniques. A good example is Marshfield-Ironmonger’s Piece (Glos.) (Fig. 66), where a stone wall appears to have separated two roundhouses and their yards, neither of which was explored extensively. That may not matter much because the subsequent history of the site is clear. The wall was demolished and a villa, comprising two row-type houses built end to end and facing opposite ways, was constructed over it, and the builders carefully sited the new work so that the transverse wall where the two houses joined lies obliquely to the demolished boundary. In the circumstances of a house designed for two separate households replacing two separate houses, the location of the joining wall is certainly symbolic of the unity of equals or, rather, near-equals; for even at the time of union one house was more utilitarian than the other. The short subsequent history of the site reveals that the two were unified by the building of a porticus so that both faced in the same direction, and the greater dominance of the better house corresponded to the decline of its neighbour.10 Marshfield is a different manifestation of the unifying process that is demonstrated at Barnsley Park and Rijswijk by the Romanisation of the buildings. Another instance of simultaneous unification and Romanisation appears to be Condé-Folie (Fig. 66) in the Somme basin11 and there may be another at Trouy (Cher).12 The significant criteria in such cases are that the boundary, usually a ditch, coincides with an important internal division in a building, commonly that between one domestic unit and another; or traverses an important room, commonly the large middle room of a row-house; or has a house abutting it end-on; or runs diagonally across a building, which is unlikely to have been built in that relation to it by chance. Uplyme-Holcombe (Fig. 66) shows the stages by which an elongated row-house might develop. Settlement began in [pre-I] with a small late pre-Roman Iron Age ditched enclosure not much more than a quarter the size of St Lythans-Whitton, within which were two round huts. Some wooden houses of unintelligible plan in I, the first Roman phase, were succeeded in II by others equally unintelligible and also a small single-ended hall of stone or with stone footings which abutted the ditch at its south end.13 When the time came to enlarge the house, it was extended outwards over the ditch, not, as might have been expected had purely structural considerations governed the decision, at the opposite end where vacant space was to be had. Had Uplyme, which was twice lengthened subsequently, been dug in the nineteenth century the unphased plan would have been assigned correctly to the AylesfordEccles class; what would have been missing is the evidence that building over the ditch was a quite deliberate choice, and that maintaining social relations was allimportant. Any risk of structural failure was incidental. What might have appeared as necessity caused by lack of space was a decision taken to perpetuate the social order. Bignor, in its relation to a ditch, began in the same way (Fig. 42) but its development was determined by building along the boundary line, not across it. At both Uplyme and Bignor the ditch was very important. Aylesford-Eccles I (Fig. 66) too was built over what must have been a boundary 240

— Modes of Romanisation — ditch. It runs obliquely underneath the house at the mid-point in the length of the domestic units, excluding, that is, the supposed open-ended shrine at the south end. This is an instance where such a siting seems unlikely to be the result of chance, and the same may by true of the much smaller row-houses at Radley-Barton Court Farm, Little Milton14 and Jublains (Mayenne). Slighter evidence can be interpreted in the same sense. Beadlam N I (Fig. 17), a unit-system house, has, under the floor of the middle room and extending beyond it under the approach road, a trench, whose course is marked by subsidence: ‘This trench may have been for a drain, or it could perhaps have held a water-pipe – it was not excavated.’15 In the light of Marshfield it may have been a boundary over which the new house was built as a symbol of the unification of two houses, each hitherto in its own yard. Henceforth they would be run as one. In this connection Kings Weston (Glos.) (Fig. 66) is interesting. Many years ago Boon pointed out that a ditch appeared to run diagonally under a pavilion and across the hall (then thought of as a yard), because strengthening had been necessary at several points. That being so, the siting can hardly be fortuitous. If the ditch filling had not consolidated sufficiently to prevent subsidence, its presence is likely to have been noticed at the outset by builders to whom variations of soil and vegetation must have been part of their everyday work; and had they found the ditch while digging footings trenches, removal to a trouble-free site would have been easy. Deliberate choice to mark the supersession of a boundary by the new hall is the probable explanation, as it certainly is at Condé-Folie.16 Other kinds of relation between buildings and ditches are observable. At Milton Keynes-Bancroft II (Bucks.) (Fig. 67) six buildings reveal social organisation in the way they abut, overlie or stand tangentially to ditches. The two principal buildings, an aisled house (7) and a hall (2), face each other across a space denned by two eastwest ditches; house 7 is aligned with the edge of one, hall 2 stands parallel to and just over the other. An intention to efface a boundary is very obvious in 2, whose construction necessitated packing the ditch with limestone rubble. On the wall of 2, facing house 7, are what archaeologists invariably describe as buttresses:17 as such they have no conceivable function and are in fact the pilasters appropriate to an important domestic building. To east of 2 a larger single-aisled building (9) abuts the same ditch. One roundhouse barely impinges on a ditch and is, in effect, abutting it; another, 11, overlies one slightly and in that part the floor was reinforced against subsidence. All these points amount to a careful siting of buildings in three kinds of relation to ditches with the quite deliberate intention of obliterating old boundaries. A similar intention is likely at Hemel Hempstead-Gadebridge Park (Fig. 67), built parallel to a ditch with the wings projecting forward over it. Hamblain-les-Prés resembles it in having the porticus built over a ditch, while Faversham I and Brixworth I are built parallel and close to ditches (Fig. 11). Variations on these themes appear at Dragonby (Lines.) and Jublains-La Boissière (Mayenne) (Fig. 67).


— Chapter Fourteen —

Figure 66


— Modes of Romanisation —

Figure 66 Building over boundaries

COLLINGHAM AND RADLEY: CONTINUITY OR DISCONTINUITY? These two carefully excavated British sites have been thought to show a clear break between the indigenous settlement and its Roman successor. When the evidence is examined in the light of Rijswijk and Whitton this conclusion is less clear. It has to be said at the outset that Collingham-Dalton Parlours (Fig. 68) was too damaged by ploughing to provide the stratigraphical clarity of the Dutch site. Given that limitation, the Iron Age saw four phases of growth, from one enclosure with one house to four enclosures and two houses, then successively to five and three and, at its apogee, five and four. In a final native phase the settlement was reduced to a single house and greatly simplified enclosures. The corollary of this rapid shrinkage must be a sharp reduction in the number of inhabitants, just as happened at Whitton between phases V and VI. In the next phase Collingham became a Roman villa comprising three principal structures and, we are told, ‘There is no sign that its erection was symptomatic of the gradual Romanisation of native farmers, as . . . at Whitton. . . . Rather, it represents the plantation on this site of a high-status household with military associations.’18 The dramatic change notwithstanding, ‘there are clear indications that the positions and orientations of the buildings were affected, if not by the ditches, at least by the boundaries which the ditches had once helped to define’.19 Three Roman buildings – J, B and A in order of importance – stand on the site of earlier round-


— Chapter Fourteen —

Figure 67


— Modes of Romanisation —

Figure 67 Building along or over ditches

houses, the most important of them succeeding the last native house. Two (J, M) were set out much like the first house at Uplyme, with the west gable wall aligned on an enclosure ditch, as if the two parts of Marshfield had been staggered, not aligned. In fact, the evidence simply requires continuity and possibly the gradual replacement of round timber houses by rectangular ones with stone footings, and given the impossibility of establishing at all closely the duration of any of them, something like 245

— Chapter Fourteen —

Figure 68


— Modes of Romanisation —

Figure 68


— Chapter Fourteen —

Figure 68 Problems of continuity

the transition observable at Whitton or Barnsley Park, continuing the pre-Roman evolution, seems probable.20 Radley-Barton Court Farm (Fig. 68) is another villa thought to show a break in continuity, in this case of a century or more, between its latest timber phase and the erection of a stone building; yet apart from that its development follows much the same course as that of the villas described earlier, though it has fewer phases. In [I], within a trapezoidal enclosure were two progressively smaller ones and a roundhouse; just outside it was a second round-house. All was swept away in [II], to be replaced by a larger bipartite trapezoidal enclosure, differently orientated, containing, 248

— Modes of Romanisation — in the smaller part, a timber hall house of about the same size as Farmington I overall but without any subdivision. Then, it seems, from the mid-second to the late third century, the site lay vacant, unless some pottery which may be of the later second century ‘may indicate a phase of occupation undetected on the site’.21 this was followed by the construction of a squarish enclosure fenced off into three principal yards which were themselves subdivided by rectilinear ditches and fences, and containing a sub-rectangular ditched enclosure within which the stone-built villa stood. Insofar as it is a progression through increasingly rectilinear enclosures to a stone villa it resembles Rijswijk, from which it differs in the interposition of a slightly Romanised timber hall between the native round-houses and the villa. The time gap may exist but the site develops as if it did not.

BUILT ON DITCHES: KAISERSTEINBRUCH AND RUDSTON Although the villa of Kaisersteinbruch-Königshof (Aus.) (Fig. 67) appears not to illustrate the transition from native farm to Roman villa, it provides an instance of building on ditches so striking and so much in need of explanation as to deserve mention here.22 The story is briefly that an early villa site I was cut through by defensive ditches not definitely associated with any buildings, which themselves were superseded by a wall and towers protecting a second villa II. Problems of I and the significance of the ditches will be ignored here. What matters for the present purpose is that the complex of yards and buildings in II is deliberately sited on that part of the ditches enclosed within the defensive wall, something which can hardly have happened by accident and which nobody has attempted to explain. Three of the elements comprising a villa can be recognised in II: they are the large aisled workhall E and two blocks of domestic buildings (K, H+1) standing on opposite sides of a yard, the whole, with outbuildings, giving the impression of a single building campaign. A most striking aspect of the aisled building is that it straddles the double ditch perfectly symmetrically. Moreover, the south domestic range K is built along the outer ditch which runs along the middle of its entire length, yet, had it and the north block been sited 5 m to south, neither would have been over a ditch. Although the builders were clearly aware of the ditches, all of which must have been visible, they sited two principal buildings upon them quite deliberately. Why? Thomas provides the clue with the comment that the settlement does not give the impression of a country seat or centre of a large agricultural enterprise (latifundia) but rather resembles the buildings appropriate to several families.23 In fact, there are four residential blocks of varied status (including C+D), three of row type and the workhall, which had as much domestic accommodation as, for example, Hölstein (Switz.) (Fig. 7). Looked at in that way the site has some resemblance to Marshfield, and it is difficult to find an alternative to the reorganisation of boundaries to account for building over ditches. The villa at Rudston (Yorks.) (Fig. 67) reinforces the argument. Under the whole length of the principal (east) house and a workhall at right-angles to it a little distance away ran ditches. As Stead, who dug the site, remarked: ‘these two ditches . . . set 249

— Chapter Fourteen — the orientation of the subsequent buildings. Why the buildings were constructed over deep ditches has not been explained: the problems of subsidence must have been obvious, yet buildings were re-built, or refloored on the same site.’24 From Hungary to Britain the siting of buildings over ditches was of the greatest significance to contemporaries and is another element of the code governing property rights within kin-groups.

REMOVAL TO A NEW SITE Commonly the transition from native to Roman is thought of in excessively simple terms as the direct replacement of one kind of house by another. Of course this happened and it is observable at such well-known sites as Ditchley (Oxon.), WelwynLockleys (Herts.) and Mayen. Sometimes, though, the intention to rebuild a settlement in the Roman manner seems to have been achieved by moving it to a new site, hard though it is to establish such a move archaeologically. Some instances are known of a native and a Roman site standing in close proximity, for example Port-le-Grand in Picardy (Fig. 68),25 where a sub-rectangular enclosure in which are two round-houses, one large and one small, stands about 100 m from a villa laid out with a perfectly rectilinear courtyard; the caption to the published plan asks whether there is a link between the two. At Sparsholt (Hants.) (Fig. 68), where a similar juxtaposition occurs, excavation of both has taken place and it is claimed that there is no continuity from one to the other; the gap is some thirty years.26 It is worth considering this statement in conjunction with the inferences that can be drawn from the development of the villa courtyard. The oldest element of the courtyard dated by excavation is a hall on the righthand (north) side; it was aligned with a fragment of wall, all that remained of the original north side which tapered inwards from east to west. The opposite side tapered similarly, from which it is clear that the setting-out of the courtyard is designed to emphasise the importance of the main house opposite and facing east towards the entrance. That the existing house was in that position is not in doubt, but it must have had a predecessor for the shape of the courtyard to have any meaning. For the immediate purpose the importance of Sparsholt is that it was set out as a hierarchy of buildings comprising, probably, the same elements of house and two halls as are found at Rijswijk IIIB but arranged more formally. So if the original tapering courtyard does really represent the removal and initial phase of Romanisation of the kin-group that had lived nearby it expressed a hierarchy from the first; and it would be interesting to know whether any stages towards that social situation were perceptible in the old settlement.

ROMANISED COURTYARDS AND WOODEN BUILDINGS Many villas began with timber buildings of Romanised rectangular form in what was usually a short-lived phase before rebuilding in stone, as Hemel Hempstead-Boxmoor I (Herts.) shows (Fig. 68). 250

— Modes of Romanisation — Where more extensive excavation has been possible, a grouping of timber buildings around a yard, of a kind more familiar when executed in stone, has usually appeared. Bedburg-Garsdorf (Nordrh.-Westf.) (Fig. 68) shows the kind of change that took place on continental sites. About 100 m north-west of the villa site a late pre-Roman indigenous settlement produced, among vestiges of building, the plan of a small hall. In the succeeding Romanised phase three buildings perpetuate in improved form the use of earthfast posts and ridge-post or (implicitly) king-post roofs, and native pottery is found alongside Roman. Here, then, continuity is established and, although pottery evidence pointed to occupation from the first to the fourth centuries, comparatively little change took place, only the principal house being rebuilt in stone with one undivided inner room. So what is meant by Romanisation in the context of Bedburg? In terms of architecture and building construction, not very much. More important are the social implications of organising the settlement on unmistakably villa lines, something which its failure to develop on conventional lines tends to obscure. Comparison with the irregular plans of yards in southern Germany such as Regensburg-Burgweinting (Bay.) shows how big a step this was. As Bedburg is set out it displays the same way of ordering a kingroup as many a British or French villa except that all we see are the beginnings, before increased prosperity sharpened the differences between the component parts through the addition of mosaic pavements, baths and an inner courtyard. Chedworth I, set out on similar but not identical lines, also has scraps of evidence to suggest that, like Bedburg, it was first conceived of in timber in the form later carried out in stone. Much the most remarkable instance of a Romanised courtyard combined with timber buildings (which, like Garsdorf, largely remained so) is Druten-Klepperhei (Neth.), which lies, like Rijswijk, just inside the frontier. Essentially it is a long courtyard aligned east–west which splays out slightly towards the principal house at the west end, like a less pronounced version of Tarrant Hinton (Dorset) or Darenth (Kent). It shows some departures from the orthodox form, for instance in the change of building line in the north range but, had all the buildings been rebuilt in stone, it would not have looked out of place among Picardy courtyards. Druten and Bedburg raise the question of what ‘Romanisation’ really means. They show that the term cannot be limited by the connotations of the word villa as defined early in this book, because both sites would have appeared more familiar to someone coming from central France than, for example, Burgweinting or Ludwigsburg-Hoheneck, particularly when their timber buildings were encased in plaster and embellished, as they no doubt were, with wooden mouldings and painted with classical architectural motifs. Such a visitor would have put them towards the bottom end of the scale by which Romanisation was measured, but a comparison with the indigenous farms of the region, even Rijswijk IIIB, would have left no doubt that these farmyards were a kind of villa, however modest. They conform to Collingwood’s definition, which reads, in part: ‘Any house of the Roman period may be called a villa, provided that it was the dwelling of people, somewhat Romanised in manners, who farmed a plot of land.’27 Whoever set out the plan of Bedburg or Druten was perfectly familiar with Roman ways; and if ‘villa’ be rejected as an inappropriate term, it is hard to think of an alternative which differentiates them from, say, Hartfield-Garden Hill. 251

— Chapter Fourteen — Just how fully Druten conforms to villa norms appears in the siting of the principal house, which is closed off in a typically Roman way. Although it forms the centrepiece of the vista seen from the entrance, it is not truly in the middle but stands somewhat to one side, and the same is true, less markedly, of Bedburg. This is generally the case with long villa courtyards and is particularly noticeable in Picardy.28 Symmetry itself ties Druten to the Roman world; a deliberate partial rejection of it, which is common to a large segment of the western Empire, proclaims some indigenous social imperative which related two of the three principal buildings more closely to one another than to the third.

AN IMPLIED FIRST PHASE IN TIMBER? On present evidence villas with a first phase in timber are a minority although, as excavation to the lowest levels becomes commoner, the proportion is rising rapidly. It may be possible, using the kind of inference developed in the study of British vernacular buildings, to add a few more from old excavations. In Switzerland and Germany a number of hall houses are set out in an unrectilinear way which has not been explained. Neumagen-Dhron-Papiermühle (Rhld-Pf.) (Fig. 68) may be the best illustration of this kind of irregularity. No two walls of the hall are parallel, the porticus is poorly set out, the pavilion at the south-east (correctly south) corner is slightly out of square – all these departures from the right-angles usual in halls are sufficiently pronounced as to suggest that the normal building process was not followed. Explanation can either be specific, such as asserting that the builders were incompetent, or that considerable allowance has to be made for footings (which may in part be what was actually drawn) being more roughly built than the walls they supported; or it can be of a more general nature, founded on the hypothesis that a structure of impermanent materials – walls of clay (cob) or some fairly rudimentary kind of timber-and-clay – of which no trace was found, was rebuilt piecemeal. Obvious objections can be raised to all these explanations; the last is preferred here, as having proved useful in some British contexts. On this assumption Neumagen-Dhron [I] was a timber hall of about the same dimensions as its successor. In [II] the porticus-and-pavilions were added, a process that entailed rebuilding the front wall of the hall – which is the best piece of settingout in the whole building – while leaving the rest standing. In [III] the other three walls of the hall were rebuilt one by one, perhaps leaving the roof intact to provide shelter for the inhabitants who went on living there while the work was proceeding.29 A number of hall houses apparently set out without benefit of right-angles are found in Switzerland. They include Osterfingen I (Fig. 68), Altstetten and Wiesendangen-Steinegg; Öschelbronn (Bad.-Württ.) is similar. The kind of irregularly set-out structure which is assumed to have preceded these villas appeared under another Swiss villa, Laufen-Müschag, although here it did not dictate the shape of its successor. For all the examples quoted, the general explanation suggested above might be considered. Some halls taper markedly without displaying all the imperfections of Neumagen252

— Modes of Romanisation — Dhron. Three such in Switzerland are Bellikon (Fig. 68), Huttwilen and Bözen (Fig. 68) and in Bavaria there is Ehingen-am-Ries, known only from an aerial photograph; they lend credibility to the timber hall preceding the Ditchley villa. This kind of departure from the rectangular norm may be thought deliberate, as some other villa plans certainly are,30 and if so it requires a different explanation.

BUILDING IN STONE: EARLY HALLS The best-known example of the transition from indigenous timber house to Romanised stone hall is Mayen. Two successive pre-Roman buildings gave way first to a hall of mixed construction and then to one in which the stone walls were fully load-bearing. It is a story expounded so clearly by its excavator and so often recapitulated or quoted that it needs no repetition now. A point often overlooked is that it is not the only model of transition from native to Roman house; an alternative one has been largely ignored. It has to be said at the outset that the alternative model is incomplete; of the three stages – indigenous house, first stage of transition, second stage – all of which are observable at Mayen, the first is totally absent at present. In the examples which follow, no clear evidence has been found of the pre-conquest houses which preceded the first transition. The model is based on a few, mostly German, houses, all of them rather fragmentary. Köln-Mungersdorf [pre-I] (Nordrh.-Westf.) is the best example.31 The first building of a Roman kind appears to have been a structurally more sophisticated version of Mayen III, a hall as long but not as wide as Mayen. A second and larger example, Leiwen [pre-I] is of more sophisticated plan with a small apsidal projection at one corner (Fig. 72). Vierherrenborn-Irsch (Rhld-Pf.) is a third example from the same part of Germany; Romanisation began with a hall about 10×9 m, of much the same size as the pre-Roman house at Landshut-Sallmansberg (Bay.). All three were pulled down to be replaced by much larger, highly developed hall houses of three very different kinds. Far away in central France, Noyers-surSerein (Yonne) provides another instance of this short-lived transitional stage before the building of a goodsized villa; here, a square hall with what is probably a porticus gave way to a cluster of row-houses. That is possibly the only certain French example of these early halls. Others may have been found and either dug incompletely or reported in unclear terms, as, for example, the ‘rather eroded building on a different alignment’ which formed Saint-Herblain I (Loire-Atl.)32 or Saint-Germain-lès-Corbeil I (Fig. 74).33 Total demolition as a prelude to rebuilding in the same material is rare in Roman villas, as, indeed, in other cultures. The question of why these early halls were not improved and incorporated in new work involves some very broad social issues and will be taken up in the final chapter.


— Chapter Fourteen —

BUILDING IN STONE: ROUND-HOUSE OR ROUND PAVILION? Britain has produced no such early stone halls destined to be swept away by a second and more thorough wave of Romanisation. If, though, the square stone halls at Vierherrenborn-Irsch and Noyers-sur-Serein be regarded as the translation into stone of wooden predecessors, an analogous change of material applied to the round-house is observable which, as appears to be the case with the continental buildings, is accompanied by a change of status or function. An interesting situation occurs in one or two villas where a circular building coeval with a rectangular main house, both of stone, occupies a very prominent place within the enclosing wall or bank. A northern villa, Manfield-Holme House (Co. Durham) (Fig 68),34 stood inside a walled enclosure built in [III] (and replacing a pre-Roman sub-rectangular earthwork of [I]/[II]), in the centre of which, and occupying the most prominent position opposite the entrance, was a round-house; it was coeval in [III]/ [IV] with a small row-house of three rooms placed only a short distance away.35 It is, on the face of it, strange that a central position should have been occupied by an obsolescent type of house with the new Roman-style house set to one side.36 That dominant position facing the entrance, just where, in most villas, a representational room might be expected, must indicate the pre-eminent social importance of the building, whatever its actual function. At the same time as the row-house was enlarged in [IV] and a bath suite added, the round-house was rebuilt, and of this Harding observed: ‘The fact that the rectilinear villa was displaced to the north . . . while the central location continued to be occupied by a building in the native tradition . . . might argue for an element of tradition in the siting of the latter.’37 Yes, indeed! That perceptive observation is the key to the perpetuation of this pre-Roman kind of building. Anyone approaching through the years saw the conical roof of the round-house matching in height, and probably overmatching, the roof of the adjoining building.38 In a society in which architecture was one of the most important forms of social display the contrast proclaimed the importance of a traditional form of structure. The intention was presumably to emphasise continuity39 or perhaps seniority within a kin-group; even, perhaps, legitimacy of title by a kin-group departing in other respects from the old ways. For although there is a tendency to look upon the transition from round to row-type house as a simple outcome of prosperity or the recognition of superior amenity, the old form must, from its siting, have carried social implications and might ultimately affect claims of inheritance.40 In some such sense the need to emphasise architecturally the continuing importance of the old while establishing the new becomes intelligible. A hardly less striking example, though more difficult to understand, is the roundhouse at Great Weldon (Northants.) (Fig. 70). The site produced ‘not a single scrap of pottery or other clearly pre-Roman material’,41 so any question of confirming ancient legitimacy of title could only relate to this successor to the old site. Moreover, the round-house – and it was a house, with a hearth – first appears in III, when it accompanies a replacement of the row-house which was the only building in I; both were set to one end of the enclosure and consequently neither had the highly visible position of the Manfield round-house. Nevertheless, the older kind of structure had 254

— Modes of Romanisation — a diameter larger than the width of the newer one and must have been a very prominent feature of the yard, however enigmatic the choice of form. These contrasting types may be combined differently by putting a round-house at the end of a porticus instead of a pavilion. West Marden-Watergate (Sussex) is the most remarkable of them.42 The earliest extant building, a round-house (4), may have been originally freestanding; its relation to the villa entrance is unknown, the enclosing wall not having been found. Very likely it accompanied a timber predecessor of the row-house, and its being built over a ditch may be significant. In [II] the putative timber house was replaced by a stone-built row-house of three rooms which intruded into the round-house; a doorway connected them but a timber veranda is virtually certain to have existed and may have extended the whole length of the row-house. In [III] a new front, like a porticus-with-pavilions in conception but not in detail, was added to the row-house and entailed demolition of 4, the latter being replaced nearby by a freestanding square building. Perhaps the most important point about this sequence is that the round-house began as a separate house, when it had about the same floor area as the open hall of a small, late medieval English house. In the succeeding phase, when it became a room equivalent in size to a two-room unit at, say, Brixworth, it had a burnt area or hearth and was large enough for a conjugal family. Whatever social relations were embodied in the conjunction of round and rowhouse at West Marden appear at Ringstead (Northants.) (Fig. 68), where the villa has a somewhat larger round-house standing in a comparable relation to the incompletely excavated main range. So, too, perhaps at Walton on the Hill-The Heath, where a round-house stood very near the rear porticus of the row-house, close enough for a timber porch to have connected them, although nothing of the kind was found in the mid-nineteenth-century excavation. The idea that house 4 at West Marden and the Ringstead round-house were independent houses finds support at Milton Keynes-Bancroft I, where two such buildings (11, 4) spaced well away from others have hearths. Finally, a few Romanised sites such as Overstone (Northants.) show only round-houses with stone walls or footings. It appears that round-houses Romanised by the use of stone for footings or walls present, in their siting more than their form, some of the same problems as the small early halls like Irsch and Noyers-sur-Serein. At Manfield, one long remained in some sense the principal house or building (if it did not have a domestic function). That may well be true of Old Durham too but is rare. Elsewhere round-houses retained an important subordinate position, standing in the same relation to a rowhouse as the hall wings did to the main house at Katzenbach (Rhld-Pf.) (Fig. 74). Sometimes they were equivalent to pavilions, which raises questions about the way the latter were used. Then there is their distribution, which is confined to districts where stone was easily obtainable. If the availability of a building material determines the existence of these houses, the social implications also must be confined to those areas, which is unlikely. In parts of England where timber and clay were the usual building materials the traces of round-houses may commonly be too slight, not for detection but for confident interpretation in the ways advanced above. But if round-houses 255

— Chapter Fourteen — really did not exist there, the problems of these stone-built ones are reduced in significance to a regional difference. In either case, here is another nuance of architectural language awaiting translation into social history.





n the preceding chapters the development of the several kinds of house has been illustrated by reference to phases of individual villas without, usually, recounting their whole history. Now certain kinds of change to which scattered reference has been made will be described in more detail, mainly through the story of particular villas, to bring out their social implications. A suitable beginning will be with the most universal kind of house in the Roman Empire, the hall open from floor to roof-ridge, warmed by an open hearth and ranging in size from very small to very large.

THE OPEN HALL: ITS RISE If we set aside round and very small squarish houses, virtually all pre-Roman houses conform to the definition of a hall, so that it is possible to see their story over the whole duration of the Empire as a tale of steady decline so far as their relation to the growth of the house as a whole and to other house types is concerned. Viewing the hall as a structural type, not limited functionally, the story is one of success, culminating in the huge imperial audience hall, the basilica, at Trier and, of course, this increased technical ability to build roofs with wide spans had a beneficial effect on domestic halls. This kind of improvement, coupled with social changes, led, at certain times and in many places, to an increase in the size and rise in importance of halls relative to other villa buildings. It is not a straightforward story. Halls should be observable first in Italy and, although few have been claimed as such, Francolise-San Rocco I (Fig. 69) has two single-ended halls of c. 100-90 BC set side by side in the sophisticated arrangment of reverse symmetry. The hall 4 was recognised as such from the first; it is almost square, with two smaller rooms to east and a wide ‘terrace, protico or room’1 on the south side. To north, another hall, 2, of almost identical size but inferior in ornament and lacking a terrace, is turned through 180 degrees and has its ancillary rooms – one a reservoir of common use – at the opposite end; this hall was thought to be a courtyard for storing vehicles. Why that purpose should need a ‘hard, smooth and waterproof floor’2 is not explained, nor why a smaller ancillary room, entered past carts, needed a mosaic floor. The antecedents of San Rocco are unknown. Elsewhere in continental Europe 257

— Chapter Fifteen — timber halls were common, and the rise of the open hall can be seen mainly in structural improvements and so to larger buildings of a wider clear span. Increased size must have been prompted by new needs, whether of greater numbers in the houseful, a desire for prestige or some other consideration, and these matters will be

Figure 69


— Patterns of Villa Development —

Figure 69


— Chapter Fifteen —

Figure 69 Rise and decline of the open hall


— Patterns of Villa Development — touched on briefly in the final chapter. Whatever its cause, the building of larger halls was commonly accompanied by the provision of ancillary rooms, sometimes one long, narrow one, often two square ones. In some big German houses there were many more rooms than that but the hall remained dominant. In a few parts of Europe, notably (on present evidence) Britain, halls appeared where none or hardly any were known before. The best-documented case is RadleyBarton Court Farm (Oxon.) (Fig. 68), where the transition from round-houses has been demonstrated, but the same process no doubt took place at many other villa sites. At Barton Court Farm the hall appears to have been undivided insofar as the state of its surviving remains allowed interpretation (Fig. 63) and Otford (Kent) may be another such, whereas at Farmington, Frocester Court (both Glos.) and other halls the first stage included at least one additional room. Evidently houses both large and small needed some space, however used, which was cut off from the body of the hall, and the tendency to increase the number contained the seeds of decline.

THE OPEN HALL: ITS DECLINE Just as Francolise-San Rocco marks the beginning (at present) of the rise of halls, so too it shows the earliest manifestation of their decline. About half a century after it was built the superior hall D was broken up into several small rooms – a change comparable to that which transformed English medieval halls in the hundred years after 1550 – and its neighbour B was improved by throwing the two ancillary rooms into one, filling in the reservoir and covering it with a mosaic pavement. The new room was accesible from what were originally two separate parts of the house, which in II was unified to a considerable extent. The transformation of San Rocco finds parallels in many parts of Europe. In Britain two open-hall houses of the characteristic narrow elongated type, MaidenheadCox Green (Fig. 69) Huntsham ‘main house’ and probably Graux (Fig. 69), undergo a direct transition to row-type. How easy the transition is from one type to the other becomes apparent if alternative paths of development, different from those that actually occurred, be envisaged. If the limits of the various kinds of floor in the hall of Frocester Court I are redefined as walls (Fig. 69), a plausible row-house emerges. And if the hall of Kingsweston [I] be divided into two square rooms flanking a narrower entrance-hall (Fig. 69), a Downton unit or a Wancourt appears. In Germany Blankenheim changes from a large hall in IIB to a smaller representational room in IIIA. Bad Neuenahr-Ahrweiler (Fig. 69) underwent change to similar effect; there the hall, which from the start had other rooms at each end, was divided by light partitions into four small rooms, one of which, opposite the entrance, may also have been an oecus. One simple hall, Bözen (Switz.) (Fig. 68), was divided up to different effect by timber partitions. In all these cases the intention is intelligible; not so in the next one. At the eastern end of Austria, at Bruckneudorf (Fig. 69), occurred an even more radical transformation, in which the large hall of the original building was split up into small rooms, leaving – and this is what is so surprising – not one of any consequence at the entrance to a grand house. As altered in [F], the middle part of 261

— Chapter Fifteen — Bruckneudorf resembles the core of Fliessem (Fig. 53), displaying a change from large hall to small rooms which can be paralleled and would in itself be perfectly intelligible but for the addition at the rear of an apsidal room 1, nearly as big as the former hall. So how did Bruckneudorf [F] function? The porticus, divided in [F], leads to two corridors, 12 and 31, which lead to another, 8 and so, probably, to the apsidal room. On the way the wider hypocausted corridor 31 gives access to two rooms, one large (27) and a corridor-like one (32), also heated. Adjoining these two is the minor bath suite, the major one being at a distance on the other side of the house. At the front, facing out over the porticus, are three small rooms with hypocausts (25, 24, 29), and behind them what may be two service rooms (11, 30). The sizeable square room 9, which must be to some extent a passageroom and can be compared to room 12 of Bocholtz-Vlengendaal [F] (Neth.), may also be a common hall. Fliessem provides a partial key. The hall of Bruckneudorf [I], larger than that of Blankenheim I and differently proportioned, was more of a multi-purpose hall – a cut above Mayen VI (Rh1d-Pf.) and its like but lacking the dignity of the Rhenish villa.3 As split up in [F] it expresses less elegantly the division of the houseful found at Fliessem. That does not explain the new apsidal hall at the rear, which has no equivalent in the German villa and is hard to parallel anywhere. It must have had a different purpose from the old one, as the apse shows; and although it lacks any of the ancillary rooms that might be expected with an audience hall, some element of that function seems to be implied. A kind of change whose social consequences have hardly been recognised is that resulting from the addition of wings to a hall. Weitersbach I (Fig. 69) began as an open hall surrounded by several rooms and porticuses, with the hall itself comprising about two-fifths of the total area. Rebuilding in II involved sweeping away all but the hall and replacing the rest with larger end bays; wings, one a little row-house and the other a bath suite; three rooms at the rear; and, surprisingly, a timber porticus. After all this the hall occupied no more than one-fifth of the whole. A second, minor phase of alteration reduced that still further.4 This unusual way of reducing the importance of the hall obscures the fact that other German hall houses differ from its general development only in the way it was done, Mehring (Fig. 69), with large quasi-pavilions, being a notable example. More generally, the development of secondary rooms around or within halls are all symptomatic of decline, as at Schleitheim [I] (Fig. 69), where what look like two row-houses stand on opposite sides of a corridor. The hall had little more importance than that at Kinheim, and declined further with the building of wings. The same is true on a smaller scale of Friedberg-Fladerlach (Bay.) (Fig. 69): in [I] it was akin to Frocester Court I, but when additional units were added at both ends and baths and also a large wing room, it lost much of its hall character and had points in common with both Sudeley-Spoonley Wood and Kinheim. The extreme case is the complete replacement of a hall by a row-house, as the dwelling house at Langton changed between I and II. The decline of the hall, observable as between Mayen VI and Kinheim [F], finds a parallel at Leutersdorf (Fig. 45). In [I] the hall is all-important; in [II] its function and status are reduced to those of the hall at Köln-Mungersdorf [VI]; and in [III] it declines to comparative insignificance when it is completely overshadowed by the new detached facade. 262

— Patterns of Villa Development —

THE AISLED HOUSE Aisled houses show the decline of the hall in a special form which is implicit in their recognition as a type. The general direction of change is much the same as in other kinds of hall, towards the creation of small living-rooms, and commonly there is encroachment on the lower end by the building of baths and other rooms in the aisles. Generally the changes at the upper end are more complicated than those at the lower and involve cutting off a room from the end of the nave, and sometimes the nave was extended. The purpose is to create an important square room between two groups of rooms in the aisles, as if the whole were the equivalent of a five-room, two-unit row-house: Norton Disney and Denton II (both Lines.) (Figs 75, 7) are examples.5 Looked at like that, these houses bear comparison with Weitersbach II (Fig. 69), a different kind of hall out of which two two-unit wings developed, and also with Villers-Bretonneux (Somme) (Fig. 31), where the proportions of the component units are similar to those of the new aisle rooms. In a variant form, of which Exning-Landwade (Suffolk) and Winterton DII (Lincs.) (Figs 7, 45) are examples, the big room created in the nave has a narrower extension beyond and open to it, so that together these two rooms resemble the double middle room of Sudeley-Spoonley Wood or St Stephen-Park Street. It has to be assumed that all such double rooms had a similar function. Alteration could take an alternative form in which the upper end was cut off from the rest, and where this occured the nave was usually divided by a corridor into two square rooms. East Grimstead (Fig. 7) is an example which, like others of its kind, is complicated by extra rooms. In this case it is the isolated room on the south side and, if this is ignored, the rest falls into place as a porticus off which opens a transverse lobby leading to two-room units. Stroud is the same with a different complication created by the two small, well-appointed rooms at the west end; they or one of them forms part of a superior unit.6 In all these houses the designation ‘porticus’ refers to function, not any particular architectural treatment. The rooms at the upper end were entered from the front aisle, which itself seems in some cases, possibly in all, to have been reached from the usual point of entrance towards the middle of the building, although that does not preclude another doorway opposite the transverse lobby. Quite other social arrangements are implicit in the symmetrical disposition of rooms at West Blatchington (Fig. 7). The two large rooms at the upper end are of equal size and reached by equal long lobbies or porticus-like aisles; the latter may well have had, for part of their length, some secondary use as storage or living-space. Whichever of these two kinds of change was adopted, the aisled house developed row-type characteristics, so that someone coming from Kinheim, Newel or BocholtzVlengendaal would have recognised a social pattern, however much simpler it was and however different its architectural guise from what he was used to.


— Chapter Fifteen —

THE SOCIAL DEVELOPMENT OF A VILLA Blankenheim I The current archaeological concern with theory obscures the continuing usefulness of some older and simpler ideas. To speak of the convergence of types nowadays sounds very old-fashioned, yet it may be worth applying the notion to the development of a few selected villa plans and the kind of society they represent, to see how far their very marked architectural differences correspond to differences of either houseful structure or wealth. The evidence available, small though it is in comparison with the largeness of the subject, and uneven in quality and depth, can throw some light on these matters. Of all the villas so far excavated, the one most suitable for this purpose is Blankenheim, which, eighty years after its publication, remains the best analysed of its kind. In its first phase, IA (Fig. 70), the house has a large hall in the middle and two wings and was, as Oelmann put it, ‘essentially symmetrical’.7 The qualification is important because the architectural evidence is clear: the porticus wall and an offset in the foundations run unbroken from the side wall of the monumental staircase in the middle of the front to the north wing and so preclude a return of the porticus.8 Mylius, in the course of justifying his perfectly symmetrical reconstruction, mentions neither the evidence nor the conclusion drawn from it, nor does he explain why the east porticus wall extends right up to the north wing but not to the south. Instead he states merely that the wall of the porticus returning on the south side of the wing appears in Oelmann’s plan of phase IIB.9 This is unsatisfactory and should be rejected. Blankenheim IA was indeed only essentially symmetrical and the plan explains why. The representational room in the middle was used on grand occasions, when the flight of steps up to it might be used. Everyday access was by the south wing, past what are assumed to be a wide shrine and the adjoining workhall (12/24).10 Facing the entrance is a doorway into a room (22/25) which appears to have been, in effect, a lobby serving the four adjoining rooms (18, 19, 21, 28/29), although this makes 21 a passage-room. Room 22/25 is large to be solely a lobby and probably served as a common hall11 for three sets of people. On the north side of the hall, room 39/40 is only marginally bigger than the corresponding room 28/29 to south12 but, as with row-houses, the smallness of the difference demonstrates its importance; this was the superior room. It was reached from the corridor which led from the porticus to room 42 – a room of indeterminate purpose – and, beyond, to the five rooms which, ‘at least in part’,13 formed an Lshaped bath suite. In the angle of the baths is a large workhall.14 That leaves only the wing room 60/62 whose purpose is unknown. The whole building gives an impression very different from that conveyed by most villas, one of a high proportion of common space compared with the total area: leaving out of account the principal hall (33/38) there are two workhalls (12/24 and 51/67), two fairly large hall/passage-rooms (22/25 and 42), a passage-room (21) and what is probably a little lobby through which room 51/67 was reached from the porticus. Only four rooms (18, 19, 28/29 and 39/40) can really be called livingrooms, and together they occupy a smaller area than the middle hall. Although it is 264

— Patterns of Villa Development — impossible to envisage at all precisely how many households inhabited the villa, whose farm buildings of this period are entirely unknown, the plan suggests seven or eight. There is no indication that any of them occupied more than one room and the houseful may have been like that at Aylesford-Eccles (Kent) on a larger scale. A first step to changing this situation was taken in IB when, in the course of reorganising the baths, the so-called ‘Hof’ (51/67) was replaced by a square room (67) with pillared hypocaust and a corridor (66); and at the opposite end, 21 and 22/ 25 were rationalised.15 In addition, the building was lengthened at both ends, to south by at least three rooms including a porticus,16 and similarly to north where walls had been largely destroyed.

Blankenheim II In IIA (Fig. 70) the most obvious change was a transformation in the appearance of the house to give it a flush front. Demolition of the north wing entailed total removal of the end?porticus and of the south wing, a complete reworking of that end of the house. Now it was entered at the south end through a large hall/kitchen17 and so past the hearth and the cellar steps to the porticus. This is a strange way of entering a house of some architectural pretensions but the obvious point of entry at the end of the porticus – as at Budapest III-Csucshegy, past a workhall – is ruled out by a cellar window there. One other important change in IIA was the addition to 21/22 of a hypocaust and an unheated apsidal extension, which together made it the most important apartment in the house. It was entered through a lobby which was also the entry to 28/29, so there is likely to have been a fairly close link between the people living in both rooms. In IIB room 42 was enlarged by the addition of a square projection matching externally the one to south and implying an enhanced importance and probably a change of function; strangely, though, it still provided the only way into the baths. Phase II changed the shape and importance of some of the rooms remaining from I but left unaltered the middle hall and the shrine room 30, no doubt because the social relations which were expressed from time to time in ceremonies and feasts continued substantially unaltered. Removal of the workhalls from the house must mean that their functions were transferred to buildings in the yard, and although it proved impossible to collate their phases with those of the house, some of them must go with IIA.

Blankenheim III Phase III (Fig. 70) saw a complete reordering of the middle part of the house. Away went the hall, replaced by a much smaller room with an entrance even wider than its predecessor’s. It is more like the middle room of Newport (I.o.W.) than anything in a hall-derivative, and its creation, combined with the building of smaller rooms to


— Chapter Fifteen —

Figure 70


— Patterns of Villa Development —

Figure 70 The social development of villas

north and the retention of those to south, produces something approaching a rowhouse in social terms, though far more complicated than any house built as such. It comprises three suites of rooms grouped around the focal point formed by the new middle room. To south the first apartment next to the hall/kitchen is enlarged by the inclusion of the partially heated room 28/2918 but otherwise remains the same. Between it and the new hall a corridor leads to what looks formally like a Downton-type unit in which a passage-room gives access to two flanking rooms. The south room, in 267

— Chapter Fifteen — which, on the west wall, was a niche, had some special purpose, perhaps in relation to the heated room behind. The third group of three rooms, all opening off the corridor (39), lack hypocausts and, their size notwithstanding, seem to be an inferior apartment. The projecting double room now forms part of the extensive baths (as it may always have been) and is only accessible from them. Blankenheim shows more clearly than any other villa the general direction of social change. In IA it has a few separate, more or less equal rooms and much space used in common. It is impossible to be precise about the activities carried on in the several rooms of common use: religious observance, feasts, ceremonies, cooking, perhaps the making of small wooden or metal tools – all the indoor tasks associated with a properous farm can be envisaged at the outset. In two stages the inhabitants of Blankenheim moved away from this kind of life. The first step in II was to remove some of the work functions to buildings in the yard. The second and bigger one in IIIA was to recognise newly emerged social divisions by building three markedly stratified apartments and a fourth block of rooms including the hall/kitchen in which lived people who may well have been kin but were virtually servants. At the same time, the nature of domestic religious observance changed; altars now appeared in the baths and at the entrance to the hall/kitchen but the shrine in the middle of the house was replaced by a small detached temple outside.19 Greater social rigidity in the house was probably reflected in the yards because, although the date when they were built is unknown, the two inner walls parallel with the perimeter wall are laid out in relation to the phase III house.

SUDELEY-SPOONLEY WOOD No row-house has produced as much evidence of its history as Blankenheim. Few have revealed any doorways and the only British villa to show many, SudeleySpoonley Wood (Fig. 70), which was carefully dug and analysed, has never aroused much interest except in general terms. Like other British courtyard houses, it is really a row-house built around two sides of a square of which a third side is occupied solely or largely by work-related buildings. Since only the final phase is known, evidence needs to be supplemented from multi-phase villas where the kind of inferences already applied to row-houses are the sole path to understanding. The most notable room of the main or east range is the large double room in the middle. This, unlike its counterparts at Blankenheim III and Newport (I.o.W.) has a doorway of normal width from the porticus. Presumably, therefore, it lacked the representational aspect of those and many other rooms so placed, and had become an altogether more private place – for what? A dining-room is often suggested for important middle rooms, but for whom, and on what kind of occasion? If we think in contemporary (and possibly anachronistic) terms of personal friends enjoying a convivial meal, that does not demand a grand room or, usually, a particularly large one. Still less does it demand a bipartite room like that at Spoonley Wood, where the inner and outer parts of the room could have provided for favoured and less favoured guests; that is the kind of distinction to be expected in entertainment designed to forward political ends, in the broadest sense of the term. 268

— Patterns of Villa Development — That does not explain why the wide doorway so common in rooms in this commanding position was unnecessary here. Its absence may have to do with the removal from the room of all religious ceremony which, being of a public nature and not simply private devotions, was performed in public view. These activities, it may be suggested, had been transferred to the porticus, where a masonry foundation blocking access to the north end, and not otherwise explained, was perhaps for a shrine of considerable size; there observances were still in public view. What, then, of Blankenheim IIIA, where the wide doorway to the middle room has to be explained on other grounds? There can be little doubt about the removal of the religious function; the smallness of the middle room compared with its predecessor and the disppearance of the accompanying shrine room establish that as well as archaeological evidence can. The continuing need for a wide doorway is perhaps explicable in relation to the other activity assumed to have taken place in the former hall, the performance of representational acts. Whatever ceremonies were performed to reinforce community are likely to have been accompanied by feasts, and it may well be that this function was perpetuated in a diminished form which still had a public character, rather as in post-Restoration England the king’s public dining days persisted for a few years, long after losing the real importance of earlier times. If an explanation on these lines be accepted, it reveals a difference between Blankenheim and Spoonley Wood, not just of social habits but of social structure and development. To left and right of the principal room at Spoonley Wood are similar suites of rooms, with that to the right, which intercommunicates with the dining-room, being the superior one. Each is a three-room apartment somewhat differently planned but including one room heated by a hypocaust. In the apartment to the left a lobby, which once had a doorway to the porticus,20 leads to a hall of some kind – a minor workhall, presumably. At this end the porticus terminates against the pavilion of the major workhall; in the opposite direction it continues past a kitchen and the entrance to service rooms, returning down the south wing. That part of the south wing nearest the kitchen and service rooms is occupied by a suite of three heated rooms and the rest by the baths, which were entered by a stone-flagged corridor from the west end; there is apparently no other entrance from the wing. On the opposite side of the courtyard a large workhall is dignified by a porticus-with-pavilions front as a sign that it was more than accommodation and a workplace for craftsmen and labourers. This wing has a doorway into the courtyard which forms the only entrance; a doorway facing it on the other side provided both symmetry and convenience. In this analysis Spoonley Wood and Blankenheim III both incorporate a principal room which had something more than a domestic purpose, together with three apartments of graduated importance and baths. They provide for cooking, certain kinds of domestic-related work and entrance in different ways, Blankenheim IIIA grouping them all in the big combined kitchen and entrance hall, Spoonley Wood dividing them between special-purpose kitchen, lesser workhall and three doorways from yard to porticus. Of these, the contrasting manner of entrance would have struck a visitor with most force. Entering Blankenheim IIIA must have conveyed much the same impression as entering a hall like Mayen or Kinheim, except that the 269

— Chapter Fifteen — smaller rooms were not immediately visible and thus the kitchen/hall was a barrier for some. At Spoonley Wood the barriers went up in other places: between major workhall and smallest apartment plus minor workhall, and between the second of those blocks and the more important apartments. A major social difference between the two villas is revealed by the bigger workhall at Spoonley Wood. Its presence in the inner courtyard in the latest phase, as a workhall was in the earliest phase of Blankenheim, implies that the social divisions within the kin-group had not yet advanced as far in Spoonley Wood [F] as they had at Blankenheim II, from which a labour force was excluded entirely. The gradual growth likely at Spoonley Wood was established by excavation in a villa with a similar double room, St Stephens-Park Street (Fig. 70), although not all the plan could be recovered. In its general lines the story is common in Britain. Small beginnings in VI (Fig. 11), the first Roman phase, were followed in VII (here II) by the addition of a lobby-and-room unit to south, probably another to north too, and a narrow south wing difficult to explain. In VIII (here III) further enlargement was overshadowed by the rebuilding of the middle of the house to correspond to new social relations. This is when the large double room appears and several small rooms are replaced by larger ones, both changes being found in other houses at much the same time.21 Thus Park Street and Spoonley Wood in their final phases have in common the same kind of middle room and two adjoining apartments with service rooms. The comparison cannot be pressed further because Park Street is incomplete, but it does suggest that both villages, though one was larger and obviously richer than the other, had a broadly similar social basis.

CHANGE IN A ROW-HOUSE: GREAT WELDON Some recent excavations have uncovered kinds of house development which could not be deduced from a plan of the final phase, in which they have left no trace. One such is Great Weldon (Northants.) (Fig. 70).22 By the final phase VI the house has a quite conventional appearance: fairly obviously, the apse and the rear wing have been added and what is left is a row of five rooms and two lobbies for which many parallels exist. Even an unphased plan would lead to that conclusion; it would give no hint of two previous phases and a less orthodox row-house. Great Weldon I had four rooms of ascending size from south to north and a timber veranda entered at the north end but not extending the full length of the building. The largest room may have been equivalent to a workhall and the smallest to an inner room; if so, the former may have been thought not to need a veranda, that being reserved for the superior houshold rooms. To this core were added in II a north porticus, widened at one end, and a porticus-like range of rooms on the south side, the largest part of which extended for about the length of the timber veranda and no doubt replaced it functionally. The manner of entrance links this with other early row-houses, yet the lack of lobbies makes a confident grouping of rooms in units impossible. In phase III complete rebuilding in effect lengthened the house by one room, the 270

— Patterns of Villa Development — large one in the middle. This formalisation of the plan must correspond to the introduction of a new function not hitherto necessary, and marks the transition from a kin-group operating in fairly informal ways to one in which some degree of authority was required. Subsequently the middle room was made more imposing with an apse but, surprisingly, it never had a mosaic. Embellishment of that kind was confined to the flanking blocks of rooms, and was carefully balanced; both received mosaics but only the south end had a hypocaust as well. The middle room at Great Weldon never became as grand as the one at Park Street. The successive enlargements of a quite modest nature show that no sudden increase of wealth and no considerable increase of population took place. What growth in numbers there may have been depends on how complete the excavated evidence is thought to be. If no other buildings existed than those discovered, and there is no reason to suppose so, the population of the house remained fairly constant and a slow natural increase was accompanied by increasing stratification which populated the outbuildings, not with labourers or hands but with farm workers who had a blood tie to the two principal households. The addition of an apse probably indicates that the middle room had more than a strictly domestic purpose and was important in the life of the whole houseful, but without being either a justice room or a state diningroom.

A STABLE VILLA POPULATION Where the principal house of a villa was enlarged over a long period of time it was due to an increase in population or an improvement in the standard of living or a bit of both. Often there are signs that the way of life changed too, as at St Stephen-Park Street. Where a house survived for a long period with little or no enlargement there may not have been population growth, but only if a villa yard has been excavated or revealed completely from aerial photographs is it possible to be sure of that. Otherwise an increase in numbers could have been accompanied by a loss of social status and relegation to minor buildings for some members of a kin-group. This no doubt occurred but, where it did, it is likely to be accompanied by aggrandisement of at least some part of the principal house and the persons occupying it. The most problematic case is the house which grows in amenity but not in numbers of living rooms, shows no significant social change and displays in its outbuildings no considerable extra accommodation for members of the kin-group who descended the social scale. It is not all that common a case but Köln-Mungersdorf (Fig. 71) is one. Leaving out of account the first small hall in [pre-I], the house in phase I comprised a large workhall at the north end and a block of rooms at the south. The domestic quarters, seven rooms in all excluding a T-shaped corridor, fall into two distinct parts: four rooms at the north end correspond very closely to blocks of rooms at Schupfart III (Fig. 71) and Newel (Fig. 24), while the remaining three resemble the middle block (incorporating the dining-room) of Blankenheim.23 Now the dining-room, facing west, can only have been reached from a veranda or porticus on that side; it is inconceivable that it should have been approached from the workhall. In that case the 271

— Chapter Fifteen —

Figure 71 Social stability in villas


— Patterns of Villa Development — other transverse corridors will also have opened on to the porticus, and the short arm of the T can be explained as going to a south-facing porticus, which perhaps had a pavilion at one end only.24 Moreover, the plan demands a third porticus or veranda, otherwise the north-east porticus could only have been reached by going out of doors. On this analysis Mungersdorf I comprised one two-room and two one-room units carefully differentiated by size and position, a central dining-room and ?service room, and a workhall. Its principal elevation faced west, away from the yard, and the way the house was entered seems to have been perpetuated in III, when gaps at the ends of the west porticus can only be for doorways. To east lay a group of three buildings (Fig. 75), set out with the most important one in front; this was another workhall, its superior status as being also a dwelling house proclaimed, as Fremersdorf noted, by a porticus; he thought it was for indoor and outdoor servants.25 They must have been somewhat different people from those who inhabited the workhall in the main house and may have performed only farm tasks. If this separation of halls is correct it implies that Mungersdorf was a considerably diminished kind of hall-villa even in its first phase. All those who lived in these and other buildings within the perimeter are assumed to be part of a large kin-group. It is extraordinary that in the rest of its long history the house changed not at all in essentials. In the core of the house the only change was the creation of a passage at the north end, the equivalent of a porticus; this is room 10, which is omitted from all the phase plans because it could not be dated properly.26 That can be done approximately by considering its purpose, which is that of a passage by which the bath block added in IV is reached without traversing the hall, and that is the probable time when it was built. When this happened the hall was surrounded by domestic rooms and their approaches and must have lost nearly all its workhall functions while continuing to serve as a kitchen, for which no other obvious place exists – like its equivalent Blankenheim 10/19. Köln-Mungersdorf is remarkable for its stability over a period of three and a half centuries. It had a stable population, showed no marked increase in wealth and underwent no serious social change. The baths represent an important advance in comfort and may even have added a touch of luxury to villa life, but their only social effect was to modify how the hall was used; there is no reason to suppose the relations between the several families in the villa changed. Perhaps any natural increase in population was offset by the opportunities of urban life in the neighbouring colonia, in which case there would have been no pressure on resources and no necessity for new power relations. How far this situation applied in Blankenheim is uncertain for lack of full information about the lesser buildings, but there, too, the house itself does not reflect population growth. It is a story repeated in other villas, for example Newel (Fig 24), where change is confined to improving baths and pavilions; yet why is Leutersdorf (Rhld-Pf.) (Fig. 45), not far from Trier, so different? To explain these contradictory tendencies demands a wide range of evidence and is beyond the scope of this book.


— Chapter Fifteen —

PROSPERITY WITHOUT SOCIAL CHANGE: SCHUPFARTBETBERG Schupfart in Switzerland (Fig. 71) is of the same basic type as Köln-Mungersdorf, having a large workhall at one end and a group of domestic rooms at the other. In its first intelligible phase, II, three rooms are reached from the porticus by a narrow corridor; one has the elongated proportions which are so hard to understand functionally, the others look like a domestic hall or living-room and an inner room (with a hearth). These two blocks of domestic space are comparable to the two similar blocks found at each end of the simpler German halls, so that Schupfart can be regarded as their economic and social equivalent. So what does the difference of plan signify? Probably the position of the hall/workhall depended on its social importance. If it was the focus of daily life, as at Bollendorf or Mayen, a central position was desirable; if a division of labour and differentiation of status had taken place, the end might be preferred.27 In III, Schupfart was much enlarged without change of type. Now there were four rooms grouped as a pair and two singles like those at Newel or Mungersdorf but turned through 90 degrees. Two of these rooms were heated by a small one between them, a hypocauston, and a further improvement was the addition of a pavilion and cellar. The building of one pavilion, not two, may be the only sign that the limited degree of equality they imply at Mungersdorf (where pavilions were of different sizes) was breaking down. Contrary to what happened in so many villas, the work element here grew in importance rather than diminished, so that the proportion of the total space occupied by the workhall shows a marked increase. Even so, the house was turned through 180 degrees; why? Originally the ground in front of the workhall fell sharply away, doubtless towards the fields worked by the farm. It may have proved an inconvenient approach. An easier entrance was provided in III and at the same time the falling ground was utilised to make the cellar and, certainly, to provide a view, for only towards the valley did the best room and the one between it and the porticus have any view at all. Schupfart, then, demonstrates increasing prosperity based on agriculture and, at the same time, a Romanised concern not only for comfort but also for the outlook from the house and the impression it made on those approaching. Population pressure was minimal. The villas discussed in this chapter cover several kinds of development which can be paralleled elsewhere, and it is hoped that they are reasonably representative of the more important variations. The questions posed, which are hardly mentioned in villa reports, and such answers as are offered, may, it is hoped, be thought worth consideration in future.





ypes of house and forms of yard and the ways each developed are not the whole story. Now they will be brought together to show how the various patterns of social relations within villas changed, first under the impact of Roman rule and subsequently through economic stimulus. Because the process about to be described occurred at different times and with different emphasis in the several countries and even in regions of the same country, no precise chronology will be offered; and some outstanding questions posed earlier will be taken up. The form of this chapter follows Peter Medawar’s definition of the ‘kind of explanation the scientist spends most of his time thinking up and testing’; this, he says, is ‘the hypothesis which enfolds the matters to be explained among its logical consequences’.1

THE HYPOTHESIS The hypothesis, which is founded on British evidence, runs as follows. Late preRoman societies in all parts of the Empire are likely to have had certain features in common. The most important is that land was not held allodially, by individuals having absolute right of possesion and disposal, but by kin-groups. By that term is meant a group of families, each of which was a basic reproductive and economic unit – a conjugal family – varying in numbers at different stages of the life cycle. To it might be added elderly parents and unmarried relatives who were not economically independent, producing an extended family. Kin-groups might differ greatly in size, from two families to many more, and in the economic and power relations between component families. Individual families or small sub-groups of families might live in nearby separate houses, in houses or apartments joined in an architectural whole, or communally. What united them into a kin-group was their joint economic activity, which was essentially agricultural but could include some industrial production; the kind of housing and the nature of the mode of support are less important than group co-operation. It need not be assumed that all the families, at any stage in the kingroup’s existence, were of exactly equal status. They could be but commonly were


— Chapter Sixteen — not, and the senior member of the group, or of its leading family, was the recognised head of the kin-group for the time being. Since land was held in common it could not be disposed of without the consent of the group. Moreover, the kin-group can be supposed, like pre-feudal societies generally, to have followed the custom of partible inheritance, by which a deceased person’s estate was divided equally among his heirs. ‘Estate’ here connotes principally land rights, in the sense of an entitlement under customary law to the occupation and use of part of the group lands. In conditions which permit a steady natural increase of population, an inevitable consequence of following to the letter this customary right to share equally in the deceased’s holding is the fragmentation of farms to the point where they are no longer viable even for subsistence; but it seems that, wherever the operation of partible inheritance has been studied, steps have been taken to minimise this adverse effect. Perhaps only when preventive measures break down, as in South Wales between the late fifteenth and the early seventeenth centuries, does the splitting up of landholdings actually occur, although it may have taken place more often at very high social levels, as is suggested by the example, remote though it is from Britain, of Herod the Great dividing his kingdom between his two sons. The equality fundamental to partible inheritance was affected by the needs of the governing class at central or local level, whatever form these took. For those and more general social purposes the kin-group had necessarily to be represented by one person (and similarly for larger groupings such as septs), the one who has been referred to earlier as its head for the time being, and this of itself introduced an element of inequality. This exercise of power by an individual on behalf of his kin could not be shared but had the advantage of bringing honour and prestige to all members of the group. A study of the German aristocracy in the Early Middle Ages has shown that a society with grades of nobility was perfectly compatible with a system of partible inheritance ‘from which neither kings nor dukes nor lesser powers could depart’2 but the need for someone to exercise power and the authority that necessarily went with it was a perennial threat to the solidarity of a kin-group.3 This historical analogy shows that partible inheritance is perfectly compatible with Caesar’s mention of a British aristocracy and explains why a villa plan readily divisible into three or four units has one somewhat larger than the rest. In pre-Roman Britain the division, not clearly discernible in literary sources, appears to have been between nobles and knights or warriors on the one hand and the common people on the other. Any discussion of these matters must take into account recent work on early Welsh society and how it operated, notably in the most recent exposition by Glanville Jones.4 The way hereditary land was held and transmitted in pre-Norman Wales accords with the proposed model, and although Welsh practice cannot be applied in detail it illustrates some possibilities of change over time. Thus the occupation of land for four generations by agnatic (male) descent conveyed absolute proprietorship (by simple appropriation) on the holders, while still subjecting it to partible inheritance. Where such a situation arose the proprietors retained, in proportion to their hereditary lands, rights – which were also subject to inheritance custom – in jointly held land. Descent of property worked thus: during a father’s lifetime the sons would set up homesteads on the family’s appropriated lands; on his death the youngest son would 276

— A Model of Development — succeed to the parental homestead. At that time the family’s appropriated lands might be redistributed (but not necessarily); and this did not apply to homesteads. Later, ‘after the brothers had died, their sons, being first cousins, if they wished, could revise the sharing of the land, presumably so as to remove the rise of marked inequalities’. This redistribution of their grandfather’s lands, applicable also to second cousins and their great-grandfather’s lands, was a right but not an obligation; and with second cousins the right was extinguished.5 There remains the question of how new settlements were created. Welsh sources indicate that as increased population led to progressively smaller subdivision of the arable lands around the primary or ‘old’ settlement of a kin-group (a gwely, rendered as ‘clan’), new gwelyau were established within the group holding, with considerable difference of size between old and new until the latter was well established. It led to a pattern of dispersed settlement6 resembling the scatter that has been noticed as characteristic of late pre-Roman settlements and villas. This hypothesis is used to explain the origin of house types and the ways they were developed to meet the economic and social changes in provincial-Roman society over some four centuries. To apply this model to a large part of the Roman Empire is hazardous, rash even, yet it may prove to explain villas a little better than the present simplistic categories.

BEFORE THE ROMANS So few Late Iron Age settlement excavations have any pretensions to completeness that conclusions about the number of houses coexisting at any given time are hard to draw. On some widely scattered open sites two houses of similar size standing close together imply equality in the working of a farm;7 enclosed settlements with one large and two small round-houses show, if the former belonged to the headman, the beginnings of social stratification.8 In south-east England and in the Somme basin it was common for settlements to comprise two enclosures. The French sites mostly have one enclosure within the other and appear to develop from quite irregular shapes9 to approximately rectangular inner enclosures within a not very much larger and less geometrically shaped one. They are succeeded, probably not until the early Roman period, by sites where the inner enclosure is perfectly rectilinear but the outer one still has a less regular shape.10 Little is known about how many houses stood within these enclosures; in some instances one is discernible in each.11 British sites might be expected to resemble the French but do not. St MichaelGorhambury (Herts.) (Fig. 64) had, in III (the first Iron Age phase) a single subrectangular enclosure, to which was added in IV what became the inner enclosure, both having more than one house; L’Etoile (Somme) resembles it in its courtyards and the position of the later villa. Elsewhere in Britain multiple small enclosures and two or more houses appear to be common.12 At a site like Dragonby the strong impression is left of the constant cutting and recutting of ditches, implying, perhaps, that boundaries were changed frequently in response to the reallocation of land (among other things). All these sites are compatible with occupation by a kin-group. 277

— Chapter Sixteen —

THE EMERGENCE OF VILLAS Amongst the early houses built in any particular province one category stands out, that of residences for members of the ruling or official classes. They are not villas in the ordinary sense but have an important bearing on them. Whether those who lived in them were native to the province or incomers is not known and is in any event irrelevant to their houses, which must have been planted where they are by purchase or expropriation and which, in planning and construction, owe nothing to the indigenous society. The most striking example is Echternach (Lux.), laid out symmetrically on a virgin site in the 70s of the first century to face a symmetrically planned farm enclosure. This is exactly what is to be expected of great houses, whatever their political status, and a reminder that they needed a source of basic food and fodder independent of the market. Beginning as and remaining a luxurious residence, Echternach was also the centre – the caput in medieval terminology – of a territory of unknown size or tenure over which private justice was administered from the basilica at the north-west corner. Imported buildings of this kind include the proto-palace at Fishbourne (Sussex), Southwick (Sussex),13 Euskirchen (Nordrh.-Westf.) and possibly Birkenfeld-Elchweiler (Rhld-Pf.).14 They provided models of building technique, ornament and forms of planning, examples of the last being the two room-and-lobby units in the east range of Fishbourne Palace and the ‘room within a room’ or L-shaped lobby at Echternach. Just as important was the resulting creation of a local labour force capable of disseminating the new ideas and methods wherever there was a demand. Villas appear in Britain within two decades of the conquest,15 so the earliest may antedate the great houses a little. What motive did a kin-group have to merge two farms into one and obliterate the boundary by building this new kind of house across it? Within a few years of the conquest it must have been apparent to even the most powerful kin-group that the old way of life in which its prestige was established by the military strength of its warrior members had gone for ever; indeed, it had probably been weakened a good deal in south-east England before the Claudian invasion. Authority had to be legitimised quickly by the new rulers, so that, once local government had been established, the pressing need for the group to maintain influence and a degree of power by other means was solved by representation, through its head for the time being and other members of the former warrior class, on the civitas council or a perpetuated tribal council. These circumstances diminished the comparatively few individuals in a kin-group who belonged to that class while increasing the importance of the group, thus making unity desirable politically and the muscle provided by collective wealth necessary. This new-found source of power, like the old, required visible expression – the outlook of the time demanded it – and while dress and entertainment kept their importance, that of retinue inevitably declined. Architecture took its place. As trade with army garrisons and the nascent towns developed, the economic benefits of the civilizing process combined with potential political advantage to make the building of a villa and the creation of an agricultural base to sustain it both possible and desirable. It might be regarded less favourably as a familiar manifestation of colonialism, that of currying favour with the new powers to demonstrate assimilation. Those remarks apply to the thirty or so years following the conquest. As the political 278

— A Model of Development — structure and its workings became familiar, smaller kin-groups found two motives to rebuild – the prestige Roman fashion conferred on the whole kin and the enhanced possibility of office-holding for its members. Just this kind of point has been made about another society where partible inheritance was universal – early medieval Germany: It is interesting to see the ways in which the group was to be reconciled with the outstanding position achieved by one of its members, how cousins or more often brothers were to be made to share the heightened dignity and standing an officefief represented.16 Persons prominent in a kin-group who would have been warriors in Caesar’s time and perhaps potential warriors in Claudius’s might, as Black has argued, serve in the Roman army or be officers in the local militia17 but as time went on and military standing yielded in importance to civic office, this aspect of their social position ceased to be important, even though such people may have taken up military service for a time. Tacitus, writing from an imperial point of view, might describe this as enslavement, and so, in a sense, it undoubtedly was. In historical perspective it was a true civilizing process, the exact counterpart of what happened in Europe over a very long period, beginning in the High Middle Ages, as authority became more centralised and fell progressively into fewer hands.18

THE CONSOLIDATION OF SETTLEMENTS Such is the background to every decision to build a villa, whether taken by a kingroup or a part of it wishing to create a new gwely. Probably only a minority of settlements went directly from their Late Iron Age state to a Romanised one, meaning what is commonly called phase I of a villa; increasingly, deep excavation shows that many passed through at least one intermediate stage. What remains of that shortlived stage is usually a hall built of stone or with stone footings, as at Noyers-surSerein (Yonne), Köln-Mungersdorf (Nordrh.-Westf.) (Fig. 71) and Leiwen-Bohnengarten (Rhld-Pf.) (Fig. 72). Why it should have been thought impossible or undesirable to incorporate the building or buildings of that phase into a later one, as invariably happened with those of the next phase (the convential I) is not obvious, nor why it was sometimes accompanied by a change of alignment. At Noyers-sur-Serein a change from hall to row-type houses may explain total abandonment of the old building but there is also a slight change of alignment. The same thing happened in the change from Vieux-Rouen-sur-Bresle (Seine-Marne) I to II, where a courtyard house was replaced by another of virtually the same dimensions. Another change in the [pre-I] to I transition is an apparent increase in the number or size of buildings. A simple stone hall at Noyers-sur-Serein was replaced [II] by two small units, a Downton-type unit and a workhall (Fig. 72); its [pre-I] counterpart at Köln-Mungersdorf was replaced by a large hall-derivative (Fig. 71) incorporating much the same accommodation. On the face of it this kind of change implies a dramatic growth of the villa population in a short period of time, which is highly improbable. It is better accounted for by assuming that the timber buildings of which 279

— Chapter Sixteen —

Figure 72


— A Model of Development —

Figure 72 Second-phase enlargement

traces are sometimes found on sites like these were houses or combined shelters for family and animals or crops, which would allow for a considerable villa population. Consequently, the transition to a villa involved two decisions, not one, and this raises the further question of why the kin-group was unable to go straight to its final objective and had to achieve it piecemeal.19 The answer to that question where sizeable villas are concerned may be that the final objective was inconceivable at the time when the first stage of stone and timber buildings was decided on, and not simply in terms of the resources needed. Both stages involved change in the social structure and the second may well have been influenced by the outcome of the first. In a society where partible inheritance involved the periodic division of land, buildings were of impermanent earthfast construction and replaced comparatively frequently, even if they did not move their location. A house built of stone or with stone footings was less prone to decay and its survival in the hands of the leading family of a kin-group may have tended to perpetuate that family’s position within it, and to limit or eliminate the idea of a ‘head for the time being’. Villas with many buildings around a rectangular or trapezial courtyard such as those in the Somme basin must, if they show no sign of being of two phases, be the product of a single decision to rebuild. The problem is to understand what kind of person or group took such a decision, given the disparity observable in several villas


— Chapter Sixteen — between the various houses enclosed by the perimeter wall. Something may be learnt from the only example which has been dug to its lower levels, Winkel-Seeb (Switz.) (Fig. 72). Winkel I comprised two timber buildings, one large, the other about a fifth of its size, set at an angle to each other rather like Gayton Thorpe (Fig. 39). They are the equivalent of the single hall of Köln-Mungersdorf [pre-I] and lasted perhaps twenty years before being destroyed by fire. Phase II corresponds to Köln-Mungersdorf I, Noyers-sur-Serein II or Hemel Hempstead-Boxmoor II as a period of expansion, the difference between those villas and Winkel being the scale on which the latter was planned from the outset. Those parts established archaeologically were the rectangular hall (Fig. 7) forming the core of the main house, the three subsidiary buildings in the inner courtyard and the west hall of the outer yard; and there is every reason to suppose that the building set perfectly symmetrically to east of the latter represents a second building campaign forming part of the initial design, and the same is likely to be true of H and J. The whole is intended for a far larger and more hierarchically structured population than that in I, and already when II was begun (in AD 45–55) most people must have been living nearby and moving in as buildings were completed. Labelling A the Herrenhaus or manor house and its builder a ?veteran of the 13th Legion20 does not explain where the additional people came from. The present hypothesis postulates a kin-group, one probably, to judge by the lack of pre-villa finds, living in scattered clusters of small buildings. A decision to build a villa produced I but, as in other places, this seems to have been felt unsuitable quite soon and a new set of buildings was planned to accommodate the whole and to realise new political and economic objectives. Explanation of all big regular courtyards has to be in terms of a single decision underlying their planning and setting-out, and the social and economic circumstances which led to it. The matter cannot be pursued here but it may be noted that at Warfusée-Nord the unit-system house (p. 159) implies more clearly than the hall of Winkel II that a kin-group decided, at a certain time, to recognise architecturally the social divisions which had emerged under Roman rule. Warfusée-Nord shows on a large scale the bipartite division apparent in much smaller villas like Maulévrier and Laperrière-sur-Saône, and the wings, one front, one rear, at Winkel-Seeb and Langton Dwelling House III (Fig. 3) are a different way of expressing it. Evidently the kin-groups retained, in big establishments and small alike, a fundamental bipartite structure, within each part of which a number of units of consumption or conjugal families existed. Perhaps the two courtyards of St Michael-Gorhambury should be interpreted similarly.

THE DIFFUSION OF VILLA TYPES When house plans are assembled from Europe and not simply, as usually happens, from a single country, the various types have a very international look, crossing both modern and Roman boundaries. Tholey-Sotzweiler [I] in eastern France and Deva in Romania constitute one striking instance of nearly identical houses found far apart, 282

— A Model of Development —

Figure 73 Diffusion of a type

and Schaanwald (Liechtenstein) is not so very different (Fig. 73). Row-houses appear in Britain, the whole of France down to the Spanish border, Germany and Switzerland. Halls are everywhere except – so far – Spain, but so basic to human shelter is their elementary form that this wide distribution is not surprising. More remarkable is the use of particular architectural means to achieve a social end, notably reverse symmetry; it is found from Italy to Belgium and in Switzerland but not further east and has a distribution similar to that of the porticus-with-pavilions front. Yard types are less widespread, being limited by the different economic base and the varying duration of Roman rule in the several countries, but they are not confined within national boundaries. Regional types hardly exist. Peristyles are confined, with few exceptions, to southern Europe. South-east Europe has a distinctive character with its many small villas of unorthodox plan which are different from those in other parts of the continent and mostly from each other. Many are halls; otherwise the individuality of this part of Europe consists in unusual combinations of elements familiar elsewhere. The hypothesis postulates that the several types of house and the ways they were combined were designed to meet the needs of kin-groups of varying composition and at different stages of development, and that the choice depended on the size and structure of the group. Precisely what decided the choice of house type for any particular kin-group is unknown at present but the general progression is from hall to row type. How did the members of the group know about types? A feature of house planning as distinctive as reverse symmetry cannot possibly have developed independently in several places far apart; it can only have been diffused by the 283

— Chapter Sixteen — architectural profession. The same goes for the very widespread types of lobby. A transverse lobby sounds simple enough to be invented independently wherever it appeared useful; in fact, making it the basis of a house type, especially when used in conjunction with the square lobby, and using that type over a wide area, indicates an ultimate common origin for all of them. How architectural ideas, for example those concerning classical detail, were diffused seems not to be known save by inference from the architecture itself. Presumably architects were summoned to a province as required, primarily to design the public buildings and a few early great houses like Echternach but also to work for private clients. At some level the Roman administration must have been aware of the needs of native societies and made the information available to official architects, since it is hardly credible that individual professionals devised independently the same limited range of solutions, some of them specific to a particular situation. As was said earlier, they worked to a code applicable from the Netherlands to the Pyrenees and from Wales to Switzerland. Eastwards to Romania and Bulgaria a different code applied, one that is much harder to crack and has only tenuous links with the first.

DEVELOPMENT OF THE VILLA SYSTEM A recurrent theme of the preceding chapters has been the many and various expressions of duality in villas. It is not simply a matter of two units and two households in small houses like Sinsheim (Fig. 8) or Maulévrier (Fig. 13); Munzenberg-Gambach (Fig. 31) and a really big villa like Haut-Clocher/Saint-Ulrich (Fig. 53) show it too, even though the number of households in both is certainly more than two. Fliessem, again, shows its dual character in its two entrance fronts, behind which are several apartments. In the story of villa development a division of this kind, however achieved, is fundamental though not universal. Ewhurst-Rapsley (Fig. 41) is an instance of what are effectively two farms united to the extent of being in adjoining yards. Marshfield (Fig. 66), originally two farms, was truly united by the building of two houses end to end. Elsewhere one house faced a divided yard, as at Bad Homburg (Fig. 9) and perhaps, in effect, at Hambach 512.21 At Winkel-Seeb II the hall forming the main house shows, in the ‘portico-like annexes’ at the ends,22 minimal signs of dual occupation which are reinforced in IV by the addition of wings front and rear. The pavilions of Mehring [F] were enlarged to become wings; the wings of Weitersbach II (Fig. 69) are two row-houses developing from a hall house; and Geislingen-Heidegger Hof (Fig. 33) is effectively two houses. These are some of the many kinds of duality running through villas. All the houses mentioned in the preceding paragraph are of more than one period of construction, as is inevitable where occupation extends over two hundred or more years. Where they show continuity of occupation with a pre-Roman settlement, they correspond to the old settlement of the hypothesis; where not, they may be new gwelyau or settlements (but the Welsh term has slightly different implications). If a villa is separated by a considerable period of time from previous occupation it may also be a new gwely, one placed on part of the kin-group land from which the inhabitants had moved away under partible inheritance and redistribution. 284

— A Model of Development — That point is bound up with the question of whether it is possible to determine the extent and boundaries of what British archaeologists, sticking to their Victorian country-house model, call the villa estate. For Roman Britain the matter cannot be proved archaeologically and any area defined by natural boundaries or the proximity of other villas is speculative.23 Moreover, there is an important objection to the idea of an estate as it is currently peddled, that no allowance is made for the effects of land transfers, whether by inheritance and division between heirs or by sale and purchase. For the latter little evidence exists for Britain,24 probably because there was no such market, yet if lands were divided by inheritance there should have been one. But if ‘estate’ boundaries can really be ascertained, and if there is reason to think they were fairly permanent, the hypothesis of a kin-group’s occupation and efforts to avoid fragmentation into smaller and smaller holdings fits well.

SOCIAL CHANGE IN HALL AND YARDS Architectural unification can be observed in broad halls, from a stage in which two or three are grouped loosely in an enclosure to the final one of transmutation into a quasi-row-house. The most basic grouping in which two or three houses turn their backs on one another is exemplified by Koerich-Goeblingen (Fig. 74). Nearer together, but still not positioned in any close relation to one another, are the buildings at Sontheim an der Brenz (Fig. 74). A modest degree of authority is apparent among the three houses, comprising four or five households, at Regensburg-Burgweinting (Fig. 41); the north house-and-hall appears to be superior to the halls or workhalls opposite. At Köngen (Fig. 74) the principal hall is the seat of a kin-group head distinctly superior to those living in the two lesser halls; the one to west, if the line of the enclosure has been correctly ascertained, may well be second in importance, as being more visible and easily approachable from the entrance. This development is carried further at Vierherrenborn-Irsch (Fig. 74), where the central hall house is bigger and better appointed than that at Köngen, and its two households have courtyards (gardens?) to separate them from the farmyard; it may well have been the residence of the hereditary head of the kin-group, which, if that is correct, now showed the disparities of wealth found in eighteenth-century Scottish clans. Hierarchy is rarely expressed as strongly as at Irsch, to the extent that many yards appear to lack an organising principle. Pforzheim-Hagenschiess (Fig. 74) is divided by a road or path into two parts. To east is the principal hall (1), which was probably entered from the south; to west is a workhall (3); baths (2) are sited more or less equidistantly from both; and each has a subsidiary building. The layout is reminiscent of Ober-Grombach with a surrounding wall of more regular plan and control passed to one household. At Bondorf (Bad.-Württ.) (Fig. 74) a very similar house standing roughly in the middle of the yard dominates it completely; nothing else matches it in size, and the two flanking buildings are perhaps workhalls of depressed status, the successors of the minor halls at Köngen. Köngen can be taken as the starting-point of a different line of development. Its principal hall corresponds to that at Katzenbach (Fig. 74) – a building of two or three phases which began, probably, as a simple double-ended hall-derivative. At each end 285

— Chapter Sixteen —

Figure 74


— A Model of Development —

Figure 74 Yards as indicators of social relations


— Chapter Sixteen — of the porticus in front of it is a hall. Their position suggests a degree of independence greater than that of pavilions (Eckrisaliten), as they have been misleadingly called; they are effectively cross-wings, albeit at a slight remove. An advance beyond this stage can be seen in a few houses which have subsidiary halls projecting as long wings, the best example being Ödheim with two. The story continues through changes in the wings, from hall to row type, so that at Flumenthal (Switz.) one wing is a hall and the other is divided into two rooms. That is a step towards the end of the series where the wings are independent row-houses: Weitersbach II has a north wing much like Drax and a south wing like Asthall-Worsham I; Wachenheim (Nordrh.Westf.) an east wing like Plachy-Buyon (Somme) and a west wing like Fontaine-leSec.25 The south wing of Friedberg-Fladerlach (Fig. 69) is yet another example. In these ways halls became steadily more like row-houses. Just how close these apparently contrasting types were is shown by three examples. If the edges of the various floor materials at Frocester Court I be translated into walls, the result is recognisably a row-house (Fig. 69). And if the hall of Kingsweston (Fig. 69) be divided into two squarish end rooms, each with a hearth and separated by a transverse lobby, the result is like Dury or a Downton unit. Although broad halls did not lend themselves to easy total transmutation of this kind, one possible example is the principal house at Graux (Belg.) (Fig. 69). Beginning as an unidivided hall, it was converted into a living-room, transverse lobby and a longitudinal lobby between two small rooms; the latter part resembles in a general way a unit at Ktittingen-Kirchberg or the dometic end of Schupfart I and Newel. It is not surprising, therefore, that in number and type of rooms Weitersbach II (Fig. 69) comes fairly close to Blankenheim I (Fig. 70), which itself, though by no means a row-house, is far removed from a simple hall. Halls pose the problem, just as acutely as row-houses, of why and how an indigenous kin-group came to adopt a hall at any particular stage of development. Someone who knew all the current options must have been on hand to advise them. Such a person must surely have been, if not formally a government architect, one who worked closely with officials. The precise level of co-operation within a kin-group seeking to build the appropriate kind of Romanised house established what kind of plan was best suited to its needs, but the choice was always related to a tightly knit series; the archaeological evidence does not, on present showing, include the unusual or eccentric solutions that might be expected had the matter been left to a kin-group’s whim.

SHRINES AS A UNIFYING DEVICE In many cultures houses incorporate a religious function in the form of a shrine, statue or other cult object. Private recognition of spiritual forces usually involved propitiating gods or spirits with offerings at pools and springs. Devotions of this kind can be inferred in villas through certain minor architectural features and aspects of planning, and many examples have already been suggested. But they reveal more than religious feelings; they were used, and again this has already cropped up, to unify blocks of buildings or separate farmyards. Applebaum put 288

— A Model of Development — the matter succinctly: ‘Cults are required to impose authority not upon slaves and helots, but upon potential equals.’26 A few early villas have, at the entrance to the principal house, a rectangular structure open on the fourth side, the side facing the person entering. Blankheim I has one such, Farningham-Manor House I (Fig. 15) another. Suggestions that the first was a pair of buttresses and the second the house entrance are untenable; a shrine is more likely, for all the complete absence of confirmatory evidence. It is the only obvious use for a room open to all who approached the house, and the same argument applies to a square open-ended room at the south or approach end of Aylesford-Eccles (Fig. 66), looking like a less ambitious version of the temple, replacing the early shrine, at Chedworth (Glos.) (Fig. 74). In an advanced form of end-entrance house like Blankenheim IIIA (and perhaps IIA), a structural shrine was replaced by an altar beside the doorway into the hall (Fig. 70). End entrances soon fell out of favour for houses of any size. Instead of three or more units it became customary to have two flanking a large middle room, to which the shrine was transferred, a move which may imply relegation of one household to a workhall in the yard and increasing social stratification. Newport (I.o.W.) (Fig. 10) had unusual evidence of this religious change in the form of two boulders buried ‘just within the threshold of the entrance’; the larger had a flattened top, the smaller ‘a pivot hole for a metal pin’.27 That it had to do with a grindstone is unlikely, given the obvious importance of the room;28 they are bases for shrines. An entrance 2.4 m wide suggests that the whole room retained something of the character of a little temple, like its equivalent at Aylesford-Eccles. A stone structure in the middle room of Marshfield N IIIB can hardly be anything other than a shrine or altar29 and the room must have been virtually a small temple. The change brought symmetry and a touch of classical dignity to the house front, qualities desirable in provincial societies whose members’ building activities were designed to reinforce their membership of the ruling elite, even if only of its lower levels. But why incorporate a shrine? North Leigh (Oxon.) reveals the intention: when the separate row-house and hall were unified by adding a porticus, an apsidal shrine, filling the space between them, was built. It was not opposite the entrance to the whole; the hall retained its own gateway, the row-house probably did too, and the shrine provided a common focus for two separate households. So too, in a different way, did the shrine placed across yard or farm boundaries, of which Ewhurst-Rapsley (Fig. 41) and Marshfield (prior to the building of the villa) (Fig. 66) are examples.30 Not all structures of this kind have been correctly identified. A squarish masonry building at Voerendaal-Ten Hove II (Neth.), said to be a granary, stands midway between the south-facing hall-derivative house and a large north-facing hall of sufficient pretensions to have a pilastered porticus. In III and IV it was surrounded on three sides by domestic rooms and distant from anything like a barn.31 As a temple it suits its position; as a granary, hardly. Shrines having the same unifying purpose appear in later houses such as Downton, where a direct approach to the middle room across the porticus was blocked by a little (nymphaeum). Verneuil-en-Halatte/Bufosse (Oise) is similar, though unrecognised in publications.32 Not quite the same effect was intended at Brighstone-Rock (Fig. 36), where a circular ‘pier’ flanked the point of entrance on the outside. It is reasonably 289

— Chapter Sixteen — explained as a base for something; that it was for a rotary mill33 is less likely than for a shrine – one apparently not involving water. Its position to one side of the axial approach, not precluding direct entrance, implies that the middle room stood in a somewhat different relation to the flanking units than at Downton. This may signify a greater unity and common action on the part of the households. Bringing a shrine into the house had architectural consequences. Putting it in a small room opening off the grand hall, as at Blankenheim, implies that the devotions and ceremonies were confined to the houseful, hence a separate shrine was required to enable those entering to make an offering to the household deities. To accommodate the first to an imposing exterior was not difficult, the second much more so, and the solution adopted may have proved inconvenient and certainly detracted from a generally symmetrical elevation. A more satisfactory solution from a purely architectural standpoint was found at Aylesford-Eccles by prolonging the house structure as a little temple; it was less satisfactory insofar as domestic observances could only be performed by first going outside. Farningham-Manor House combined the architectural merit of Aylesford with the shallower shrine form of Blankenheim, thereby losing the dignified setting appropriate to solemn group ceremonies which was available in both those villas while retaining the disadvantage of Aylesford. When Blankenheim III was built (and this probably applies to II), an altar not requiring a monumental setting setting sufficed at the entrance to the house and, there being no longer a need to bind the kin-group together by domestic religion, a detached temple was built. A comparable though archaeologically less clearly observable change occurred at Aylesford-Eccles when the original porticus was blocked at the south end, a return added to west and the shrine closed up to make a room. Where a replacement shrine was situated is not obvious. One possibility, since no detached temple was found, is the little lobby-like room opening off the porticus in the middle, which is coeval with the blocking.34 If that last suggestion is correct it is an example of the well-attested link between domestic religion and symmetry and shows how the two were related. Early rowhouses comprise two or more domestic units and usually nothing more; Aylesford is exceptional in its shrine room and Farningham-Manor House in that and the workhall. The only opportunity for symmetry is provided by those two-unit houses which have back-to-back lobbies and could have had adjoining doorways in the middle of the elevation. Whether they did so is doubtful. Symmetry is part and parcel of the need for unity and co-operation, not, initially, something to be desired for its own sake. That shift came as the leadership essential to maintaining the kin-group’s position in the face of governmental and economic pressure became first control and then, in the stage represented by Gayton Thorpe [F], something close to expropriation. Where Aylesford differs from most row-type houses is in never linking religious observance and a representational room. How some large kin-groups managed without such a room is not clear, nor is any obvious alternative to be seen here. It is remarkable that a kin-group managed to prosper without undergoing any fundamental change in what, by [F], must have been an outmoded social structure; but that accounts for its asymmetrical appearance. The increased population of the villa, which may have been of lower status than those families in the core rooms, was accommodated in extensions at both ends. 290

— A Model of Development — All villas will have had a place devoted to religious observance and commonly it appears to have left no archaeologically detectable traces. How easily they could disappear is shown by one of the Hambach villas, Ha77/264, which had, lying in the porticus next to the probable entrance, a millstone and, near it, a bronze statuette of Apollo. There is no proof that the two are connected, or that the millstone was deliberately placed where it was found, and only in the light of the preceding instances does the conjunction of the two appear possible: a base for the figure with a drain for libations. In the same vein explanation may be offered for a very minor architectural feature of some continental villas, for example TreuchtlingenWeinbergshof (Bay.), where the north wall of the proticus has two small projections for which no purpose has been suggested. Similarly with Houthem-Vogelsang (Fig. 9): in the front wall of the porticus are two small projections placed perfectly symmetrically between the pavilions, whereas the imposing staircase is not set axially between them. No explanation or even mention of them appears in the text and they are conveniently omitted from the reconstruction drawing of the front elevation35 – shades of Mylius’ high-handed treatment of Blankenheim! Some correspondence between the pair of projections and the two parts of the main range seems likely, as between those at Treuchtlingen and the double-ended hall behind. I suggest these are the bases for cult statues connected with the two households sharing each villa.

FREESTANDING SHRINES AND TEMPLES In still larger villas a water cult found expression in detached structures of some elaboration which always had a social as well as a religious purpose. Darenth (Kent) (Fig. 42) had, in front of the main building, what has been called a well.36 Nearer the mark is the first explanation offered for it, ‘a domestic chapel or Lararium, and the minute projecting structure the place where the images of the gods were kept’.37 Nearer the house, and dividing the courtyard into two, was a well-built stone channel with a cistern and an arch or other monumental structure at the south end; it was an ornamental pool, so placed to emphasise the importance of the flanking wings and the lack of a direct approach to the middle block; and that is also the primary purpose of a bigger pool at Famechon (Somme).38 The Darenth ‘chapel’ or shrine had a more impressive counterpart at Winkel-Seeb, a round tower-like structure39 paralleled in a temple at Metz-Sablon (Moselle).40 Many villas have, close to the house, a squarish stone building which can be most plausibly interpreted as a little temple. In one such building standing at the yard entrance at Mayen, the terracotta figurines and other votive objects found within established its purpose beyond doubt. Where social division was stronger, as at Blankenheim, the temple stood in its own enclosure near the house entrance. Where, as at Chedworth, the access of several households to the temple was important, it was placed apart in one corner of the inner domestic court and the buildings were skewed towards it, whereas in a large villa such as Oberentfelden with strong social divisions, a temple of commensurate size dominated the view and was presumably accessible from both the inner and the working courtyards. The siting of temples and shrines, like so many aspects of villa planning, was used to express social 291

— Chapter Sixteen — relations as much as religious feeling, whether to unify, as at Mayen (Rhld-Pf.), or to divide, as at Oberentfelden.

WHY DO LARGE VILLAS DIFFER SO MUCH? Two groups of villas provide a strong contrast. On the one hand Woodchester, Nennig, Basse-Wavre and Haccourt II/III are dominated by a large, impressive room in the middle; on the other, Anthée, Haccourt I and Fliessem are not, while HautClocher/Saint-Ulrich is unique in having two rooms of that kind set back to back. According to the hypothesis, kin-groups founded them all. Haccourt I, like Anthée and Fliessem, is a row-house combining a few large rooms of full depth with passageor lobby-like ones. In II it is swept away to be rebuilt as a great country house, yet its plan retains an extraordinary duality: the grand middle room stands isolated, flanked by two equally well-appointed peristyle residences facing in different directions. Haccourt II/III is nevertheless not a truly palatial villa. It lacks the architectural unity of Nennig because the kin-group, having prospered remarkably, preserved its structure and built two separate houses unified by a common ceremonial hall. If the idea of two splendid houses where orthodox opinion expects one appears strange, the example of the two identical palaces built side by side for the sons of King Herod the Great should be remembered,41 and the symmetrical main range at Oberweis (Rhld-Pf.) looks like two sizeable houses, each with a large hall in the middle,42 under one roof. Splendid though Haccourt II is, it has the rather sprawling character of many palaces. It lacks the tightly knit architectural unity of Fliessem (Fig. 53) which retained the ‘simple core of a modest farmhouse’.43 The facades – three according to Boettger, four, to Koepp44 – include two, west and south, which are quite remarkable architecturally and in themselves appear to reflect some internal division of the house. As Boettger notes: ‘This individual variant finally arises through the setting back-to-back of two villas with porticus-and-pavilions’; and even though duplication is not as clear-cut as Munzenberg-Gambach, each of the two residences in both villas comprises two or more households. Tantalisingly, the accompanying plan shows without argument five houses, which may be the correct total but they ought to be disposed differently.45 One crucial point is that the east and west porticuses are divided by a cross-wall, each part being equivalent to a longitudinal lobby from which those entering went into a transverse lobby. Lobbies and porticus divisions were carefully matched to the relative standing of the two parts of the house, the larger leading to the more frequented part (with two sets of baths), the smaller to the part with the grandest facade. Of course the division was not absolute; the occupants of the grander south part entered the ‘new’ baths either along the porticus or, in fine weather, across the terrace in front of it, and another part of the houseful used the smaller or ‘old’ baths.46 The plan fits a kin-group structure perpetuated in a houseful grown rich better that that of a line of proprietors who have gradually become rich,47 and it is precisely because that structure remained intact that a representational room could not be placed in the middle of the house. Another famous villa, Anthée (Fig. 75), also lacks the dominant middle room and 292

— A Model of Development — regular planning of, for example, Oberweis, and shares with Fliessem the characteristic, unusual in so large a house, of a divided middle cell, and its unit-system structure is made clear by the device of a front pool, as at Bocholtz-Vlengendaal. There are four or five units: two, probably three, in the main range and one each in the north-west and north-east wings. It is effectively the largest member of the row-house class. One more example deserves mention in order to illustrate the diversity of the class of very large unit-system houses. Aylesford-Eccles (Fig. 75), despite being extended at both ends and acquiring two semi-detached wings and many small rooms, retained to the end, hardly altered, the orignal room dispositions of [I]. Very significantly, it retained to the end the original pool; no sign here of the change so apparent at Bocholtz-Vlengendaal. And in this context the evolution of a few smaller villas might be considered: Milton Keynes-Bancroft (Fig. 75), for example, kept in V the duality expressed by a pool – here placed at right-angles to the house, like Famechon – and two bath buildings for separate households, like Fliessem; or Norton Disney (Fig. 75), where two buildings are unified by a bath house. This, in the later Empire, can hardly mean that the farm staff commonly supposed to occupy the aisled building shared comparative luxury with the ‘master’ and family; it may indicate that the unit system was not dead – that it was, more likely, flourishing.

THE PROBLEM OF A STABLE VILLA POPULATION Some villas show a remarkably rapid increase in the size of the main house in their early [pre-I] phases, Köln-Mungersdorf being one. Coeval with I is a second sizeable hall, built to accommodate people described somewhat misleadingly (according to the hypothesis) as servants; they must have been quite numerous and were of sufficient standing to merit a porticus at the front of their house and wall-paintings inside it. In the new social structure they formed a third stratum below those in the apartments and in the workhall, but it was not the lowest one: that comprised the labourers, cowherds and others who were lodged in barns, byres (buildings 6, 8) and the various late outbuildings.48 By VI the villa of Köln-Mungersdorf had increased in population, notably when the workhall 3 was built. Nevertheless, without entering into the difficult business of estimating numbers, it seems unlikely that new accommodation matches the probable natural increase over a period of about 350 years. Just as remarkable is the persistence, virtually unaltered, of the social structure of I to the end of the villa’s existence in VI, the late fourth century. Had the numbers of people living at the villa increased markedly, some change would have occurred, either in the direction of Anthée and an enlarged but still comparatively equal social structure, or of Meckel and its much greater stratification and perhaps slave labour. What may have made continuance of the original pattern possible is the siphoning off of surplus labour to the great urban centre of Köln. With little or no pressure of additional numbers the balance between the three strata did not change, added to which no marked rise in living standards is discernible in mosaic floors or other manifestations of luxury. That, of course, gives rise to economic questions which need to be asked of all the houses in the presumed kin-group territory before a satisfactory answer can be given for any particular one. 293

— Chapter Sixteen —

Figure 75


— A Model of Development —

Figure 75 Unit-system villas of the Late Empire

COURTYARDS FOR KIN OR SLAVES? A necessary accompaniment of the development of villas was the steady erosion of their original social basis. How far removed some villas were from a large kin-group or clan is revealed by looking at large courtyards such as that of Anthée (Fig. 76), which, like a number of the Somme villas and a scatter elsewhere, has an outer courtyard lined by two rows of buildings. Who lived in them – slaves, serfs, wage labourers or what? One or two villas have what appear to be slave barracks. Liédena (Navarra) (Fig. 295

— Chapter Sixteen — 76) has such a row near the house and the outermost court is lined with nearly forty more rooms looking like prison cells which may be for slaves working a latifundia.49 Chatalka-Delimonyovo Kale (Bulg.) has a few comparable rooms for the master’s own servants in the small outer court.50 A barrack-like court is different from a yard lined with identical small buildings

Figure 76


— A Model of Development —

Figure 76


— Chapter Sixteen —

Figure 76 Courtyards for kin or slaves?


— A Model of Development — spaced apart just like a Virginia plantation. Levroux-Trégonce (Indre) (Fig. 76) has nine identical buildings, each equal to two of the cells at Liédena, along one side of its 360 m-long courtyard; Oberentfelden (Switz.) (Fig. 76) is similar; Dietikon is the largest and the only one devised as one building phase throughout, with twentyfour virtually identical houses. Noyers-sur-Serein (Fig. 76) shows a variation of the pattern; the outer court, not fully explored, has a row of four uniform halls each with two litle wings. No doubt all these establishments were slave-run. Several rectangular outer yards in Picardy show little sign of uniformity in the buildings lining them; at Warfusée-Sud and Athies (Fig. 43) they show considerable differences, at Estrée-sur-Noye less, and only at Mézières-en-Santerre/Le Ziep and Warfusée-Nord (Fig. 43) are they comparatively uniform. A small but perhaps important difference between these villas and those looking like a Virginia plantation is that buildings in the former are not set out in opposite pairs, which suggests a concern that the inhabitants, whatever their status, should not be too obviously regimented. The best-known large courtyard villa, Anthée (Fig. 76), is the only one of its kind to have been explored fully. Its buildings, laid out in two rows in a long trapezial yard – and the shape itself, a not uncommon one, needs explanation – include houses with porticuses, cellars, mosaics and hypocausts (Fig. 76). The larger and better-appointed ones are nearer the house, except for 15 which seems to have slipped a couple of places. Building 3 is like Bad Rappenau (Fig. 25) in a general way; 15 is a bipartite hall like Nünschweiler or more precisely Neckarrems which also had an enlargement beyond the west pavilion. No slaves or serfs these, but people who would, if their houses stood independently, be called farmers. Whether even those living in the outermost buildings were of a lower status is doubtful. The nature of the connection between yard and main houseful is hard to guess beyond saying that the whole settlement looks much more like an enormous kin-group than a serf- or slave-run establishment, a kin-group which has prospered exceedingly and grown continually without ever hiving off some families to form new settlements – a gwely of extraordinary size. How so many people contrived to farm the land from one spot is a puzzle, the more so now that the idea of a factory making enamelled bronzes and especially brooches has been dispelled.51 Marked differences are observable between the houses standing in the inner court of these enormous yards. Anthée is a row-house on a large scale; WarfuséeNord is divided in the middle by a cross-wall, the two parts of the house matching the two gatehouses; it is like Romegoux, and Ecoust-Saint-Mien is probably the same. Athies (Fig. 43) is like Brixworth (Northants.) or Faversham in having adjacent lobbies in the middle. All three have unit-system characteristics as, less clearly, does Levroux-Trégonce, which is set out like North Leigh. Although Anthée and WarfuséeNord and Athies must all have been organised in somewhat different ways, the factor common to them is that partible inheritance survived among the ruling families while relegating most members of these large communities to various levels of inferior status. None, though, were slaves. The stages by which this situation developed are unknown but the next, the final transition into slavery, is beginning at Liestal-Munzach. There the house (not the yard) has some likeness to the later phases of Rockbourne, the north side of the yard is occupied by workhalls of varying size and comfort, and much of the south side has been built or rebuilt on a changed alignment with 299

— Chapter Sixteen — identical dwellings. Only the villa of Dietikon52 is set out in rigidly geometrical fashion from end to end; this is work discipline approaching that of the barrack square or prison. Dietikon is untypical. Villas with signs of reorganisation and a unit system are the norm and Oberentfelden is alone in having a good-sized house comprising only five large core rooms (plus porticus-and-pavilions overlooking the valley). It is comparable to Hemel Hempstead-Boxmoor VB, and could conceivably be a single-family residence. A rectangular outer courtyard breaking alignment with the inner and earlier one may mark the presence of a definitely inferior class. This is the situation at Liestal and at Estrée-sur-Noye, where the divergence, though very slight, is absent in other Somme villas of its class. It can be inferred at Meckel, where only the yard wall was traced and not the buildings. At Oberentfelden and Belleuse, the way the house lies obliquely to the rectilinear yard and poorly related to it implies that the latter has been completely rebuilt. Crucial to any such inference is the poor relation of house and rectangular yard, otherwise rebuilding need not be invoked, as Chedworth I and probably Vieux-Rouen-sur-Bresle II show. They have much in common. In each the principal (west) house, comprising two units,53 is narrower than the south range, which is narrower than the north range. Neither villa shows any evidence of the uniform habitations which are the hallmark of a depressed class.

THE EMERGENCE OF HEREDITARY LORDSHIP Concentration of power in what begins as a confederation of kinsfolk raises a question difficult to answer from architectural evidence: is it possible to infer a stage at which that superior power becomes hereditary? A development in that direction may ultimately be inevitable once permanent houses are introduced into a community which formerly was accustomed periodically to redistribute the land it worked and move dwellings as necessary. Henceforth, however much the kingroup might seek to maintain redistribution, it would be difficult to change the headship from one household to another. At Aylesford-Eccles the head for the time being was installed at the time of Romanisation in the biggest houe. What mechanism existed, or was created, to deal with this new situation in which, if reallocation of the headship was to be achieved, it would be necessary to change houses? The answer must certainly be none; nobody is likely to have foreseen the long-term effect on the control of power. As long as land was subject to redistribution, hereditary power was impossible to attain, but the element of choice in the matter is likely to have meant that the amount still subject to partibility diminished steadily. How long it was before the operation of customary law finally ceased is shown by the disputed case of inheritance rights which went up to Rome for the emperor’s decision. What C.E. Stevens called ‘the cry over the abyss’54 died at about the same time that the main range at Chedworth came into the hands of a single family, in the early years of the fourth century. The death of the old system and the concentration of wealth in fewer hands coincides broadly with the age of conspicuous consumption in British villas, as evidenced by mosaics. 300

— A Model of Development —

THE END OF THE VILLA SYSTEM Where the social foundation of villas, the kin-group, withered away, the houses designed for it became obsolete. Many were altered to a greater or less extent to suit the needs of rich and powerful men who now controlled what can properly be called estates. Woodchester, Bruckneudorf, Basse-Wavre, Gayton Thorpe and SudeleySpoonley Wood all show this tendency and form an approximate scale of importance in descending order. Some villas disappeared entirely and were presumably incorporated into nearby estates although, if that is the case, it is surprising that accommodation of a simple kind under a bailiff – and for once the word can be used with some probability – was not constructed or improvised in more instances than are yet apparent. North LeighShakenoak A IIIB, the workhall, may at this stage have been partly roofless and had become generally squalid. Piddington (Northants.) declined in the same way even if the details of its final state, when it is said to have had several quite separate rooms, may be open to question. Farningham-Manor House also testifies to the disappearance of the old order of things. It was completley redesigned in III, being reduced in size with fewer and larger rooms. It has two good elevations, one with a porticus on the line of its predecessors, though now embellished with pilasters, the other fronting three rooms with mosaics and facing the river close by; ‘a luxurious riverside retreat’, as its excavator put it,55 may be slight exaggeration but expresses the essence of the situation. A concomitant of the decay of some villas ought to be that others flourish, notably in the sense of becoming luxurious. That may find its principal expression in the large number of mosaics and perhaps, too, in more hypocausts and elaborate bath buildings,56 rather than enlargement. Basilicas or justice rooms may also be related to the growth of lordship over other villas, which would account for the size of the one at Box. Change of this kind is much harder to identify in German villas and particularly in hall-type houses. Decline may have taken more the form of complete abandonment of many small halls, and it may be chance that no yard has yet showed signs of the partial abandonment of a villa, as at Shakenoak. The other side of the coin may be the luxurious elements of such villas as Bad Kreuznach (Rhld-Pf.) and Bad Godesberg (Nordrh.-Westf.). About the end of the villa system is France it is difficult to form an opinion because little evidence of the kinds just discussed has yet been published.

CONCLUSION This chapter aims to provide a better basis for the interpretation of villa society than has hitherto been available. It is not the first time that kin-groups have been suggested as a basis for villa occupancy. Oelmann, without ever enlarging on the notion, had it in mind when discussing who lived in the two smaller houses at Mayen; they were two families dependent on the master (Herr),57 perhaps younger married sons, in any case, families with fewer rights – bondsmen, even. C.E. Stevens interpreted 301

— Chapter Sixteen — RomanoBritish settlements, including villas, in the light of Welsh law58 and, following him, Applebaum made the explicit connection: ‘not a few [capitalistic estates] grew out of kinship groups whose land ultimately came under the control of one man and whose occupants became his clients or tenants, as the patriarchal farmhouse evolved into residence, outbuildings, and hands’ quarters’.59 It has not been possible to include every strand of the argument, for example about the different sizes of hall or the growth of luxury, but there should be sufficient to assess its validity. Among other things, it may permit comparison of the process of Romanisation in different countries and thereby prompt new questions; it may also suggest that some of the explanations now popular, notably those that see villa ‘owners’ as market-oriented entrepreneurs, are wide of the mark and simply do not explain why villas are as they are.



CHAPTER ONE: AIMS AND SCOPE OF THE BOOK 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22

McKay 1975; Percival 1976. Rivet 1969, 237. R. Taylor, Soc Archil. Historians Newsletter, no. 41, 13. Baxter, 1931, 240; Colvin and Newman 1981, 122–3; Latham and Matthews 1970–83, 8, 402 (26 August 1667); de Beer, 1955, 3, 502 (1 December 1667). German nomenclature illustrates this, although the dividing lines between Herrensitz, Gutshof and Bauernhaus, especially the first two, are not at all clear. Harmand 1951. Smith 1992, 106–11. Hingley 1989, 159–61. Hemp and Gresham 1942–3; Fox and Raglan 1953, II, 75–7, III, 135; Smith 1970. de Caumont 1883, 13–15. Schumacher 1896; Haverfield 1990, 296, 302–3, 316; Ward 1911, chapters 6, 7; Haverfield 1906, 299, 326. Anthes 1906, 120; Kropatscheck 1910–11, 78. Swoboda 1918; Oelmann 1921; Oelmann 1928. Fremersdorf 1933; Paret 1932, 38–9. Paret’s incidental enlightening comments on buildings, e.g., on Merklingen (p. 39), make one deplore his unwillingness to reinterpret villas sytematically. Gnomon 1939, 40–1, esp. 40 n.l. Collingwood 1930, chapter 7; Grenier 1934, 784–95, 798–9, 819–21; de Maeyer 1937, 85. Arch Cambrensis 102, 1951–2, 130–1; Proc Dorset 87, 1966; Collingwood and Richmond 1969, 134–5; Branigan 1975, 50–1; Branigan 1977, 125–7. Brack 1975, 49–72; Agache 1978, chapter 6; Castro 1982, esp. Fig. 79, 82–6; Gorges 1979, 113–27, esp. Fig. 19; A.E. Housman, Selected Prose, Cambridge 1961, 49. Vasic 1970; M.-T. and G. Raepset-Charlier 1975; Ternes 1975; Nikolov 1976, 69. Mylius 1933; McKay 1975; Ward-Perkins 1981. Several papers on monumental architecture by T.F.C. Blagg display a rigorous approach which needs to be applied to villas; de la Bedoyère 1991 and 1993 do not advance understanding of villas. J.P. Darmon has set out amusingly the ideal conditions which a comparably large project,


— Notes to pp.10-27 —

23 24 25


a survey of mosaics in the western Empire, demand of the researcher (ANRW II, 12.2, 1975, 268). Rostovtzeff 1957, quoted Percival 1976, 56. Wightman 1970, 139. The changing structure of local government creates problems in Germany, for which I have used, not wholly systematically, Müllers Grosses Deutsches Ortsbuch, 1982–3 edition, and in Britain, for which the Ordnance Survey Map of Roman Britain (7th edn, 1956), has been supplemented by Scott 1993b. France presents less of a problem. For Germany and Britain unpublished work by Dr Friedolin Reutti and Roger Goodburn was not consulted. Regrettably, but hardly surprisingly, no publication comparable to Drack 1975 or Thomas 1964 exists for either country although Scott 1993b provides a comprehensive checklist, as, for northern France and the much wider area of northern Gaul, does van Ossel 1992.


5 6 7 8 9

10 11 12

13 14

Swoboda 1918, chapter 4; Oelmann 1921. See pp. 24–5. Chapter 8. Steiner (1923) was intended to be the first of a series of publications on Roman villas in the Rhineland; the intention failed, probably because of economic circumstances, but was taken up again by Koethe (1934). Thereafter the idea was abandoned, ‘probably on political grounds’. (Letter dated 20 May 1975 to the author from Dr W. Binsfeld of the Rheinisches Landesmuseum, Trier.) TrZft 9, 1934, 44. Fox and Raglan 1951–4; Drury 1982. Smith 1963; Jones and Smith 1963–72, esp. Part I, 5–34, ‘Development of the long-house’. Aydelotte et al. 1972, 77. Hemp and Gresham 1942–3; Fox and Raglan 2, 1951–4, 75–7; Jones and Smith 1963–72. Wills in the National Library of Wales show that equal division among heirs persisted in south Wales 200 years after the Acts of Union, among people who had little or no landed property to bequeath. P. Smith 1975, 167. Edwinsford is described more fully in P. Smith 1974, 72–3. Bull Inst Arch 29, 1992, 165. Raynham Hall, Norfolk, was built with a cross-passage or ‘Entrey’ at both ends of the hall. This apparently unique feature went unremarked when the plan which provides the only evidence of it was published (John Harris, ‘Raynham Hall . . .’, Arch J 118, 180–7, esp. P1. XVI. Webster in Rivet 1969, 245; the late Dr J.B. Ward-Perkins, pers. comm. Reece, Oxf J Arch 3, 1984, 197–210.

CHAPTER THREE: HALL HOUSES 1 Oelmann 1921. 2 Oelmann 1928, 64, 130–1. 3 Doppelfeld (Kölner Jährb 1960–1, 7–35) shows in II a hearth only at the lower end, whereas in all probability the one at the upper end in III had a predecessor; he began a much-


— Notes to pp.30-42 —

4 5 6 7 8 9 10 11 12 13 14 15 16

17 18 19 20 21


23 24 25

26 27 28 29


needed clarification of the phasing but his understanding of the building itself is open to question. Steiner 1923, 5. BJb 133, 1928, 64, 145. Rev Est 1977, 117–35; Uffler 1981, 72–3. I am indebted to Mile A.M. Uffler for lending me a copy of her thesis (Dijon 1980) in which Crain and many other villas are discussed. For this observation I have to thank Professor Amatzia Baram, University of Haifa, who speaks from experience. This is evident in the ideal plan of a Benedictine monastery of AD 820–830 (W. Horn and E. Born, The Plan of St Gall, 3 vols, Berkeley, 1979). One of the cauldrons dated to Final La Tène or the early first century AD contained coins of AD 259–260 (CAG 49, 77). Rahtz and Greenfield 1977, 49. Cf. Neumagen-Dhron (Fig. 68). Rahtz and Greenfield 1977, 53. Leech 1982, 10. For the vicus, Oudhdk Med NS 6, 1925, 40–4, esp. 42. Archiv Fft 3rd series, 12, 1920, 323. English terminology is deficient; ‘nave’ and ‘aisle’ have connotations which make both inappropriate to a secular building having two equal compartments. ‘Schiff and ‘vaisseau’ are better. Smith 1963. ‘Basilican’, coined by Ward in 1911, is used by Swoboda (1918), Fellmann (in Baselb Hmtb 5, 1950, 1–52), Collingwood and Richmond (1969), Hinz (1970). Smith 1963. Virtually every farm building is likely to have been inhabited by the persons responsible for the work associated with it; see above n. 8. Todd 1992. For this phasing see Smith 1963, 3, n. 1. Denton, with a hearth in the middle of the nave, has much less room for cattle than the aisled farmhouses of north Germany. The Hallenhaus may not be a proper parallel, and much of what has been said on this score (Applebaum 1972; Smith 1963) needs to be reconsidered. Applebaum (1972, 135–6) elaborated remarks in the report and was followed by Hingley (1989, 43-5) with two plans showing room use. The stables and byres are identified by floors and unspecified ironmongery for male use. Drack 1990, 40, 265. Drack 1990, 46, 267 (Halle); reconstruction, Gerster 1973, 66–70, partly reproduced Drack 1990, 276–9. Liédena (Fig. 76) (Castro 1982, Fig. 32) is certainly not a basilican building, as claimed by Gorges (1979, 120). Maulévrier (Grenier 1934, 805: note the structural problems arising from different bay lengths); Châtillon-sur-Seiche (which has a hearth) (Gallia-Inf. 1990, 292). FdbaB-W 1977, 405, n. 4. Ibid., 408. Reconstruction, Dmpfl Bad-Württ 1972, 38. Samesreuther, noting that excavation at Laufenburg thus far had provided as many grounds for thinking of an open yard as for a roofed hall, added that, purely intuitively, one would allow more probability to the latter (Germania 24, 1940, 36). Elsewhere he appears to imply a hall (Gnomon 1939,41). This opinion is confirmed (RiBad-Württ 1986, 401). The latest discussion is inconclusive (Rothkegel 1994, Oelmann 1921, 72.


— Notes to pp.42-59 — 31 Schallmeyer, Arch Ausgr Bad-Württ 1986, 155. 32 Two solutions are possible: with tie-beams spanning the whole, as at Neuburg a.d. Munz (Binding 1991, 151–3) and Landshut (Ostendorf 1908, Abb. 124); or with a higher tie-beam over the nave and running the rafters through continuously to the lower aisle eaves, as at the collegiate churches of St Goar am Rhein or Stuttgart (Binding 1991, 151, 155; Ostendorf 1908, Abb. 127). The Roman form of construction would have used king-posts. 33 The alternative, that the bases were for an open roof truss, is unlikely in the absence of other pairs. 34 Bipartite halls like these may go some way to explaining the very curious plan of Schleidweiler. 35 Understanding of Nünschweiler is complicated by its construction, which probably combined clay walls with heavy posts carrying roof trusses.

CHAPTER FOUR: ROW-TYPE HOUSES 1 ‘Oblong villa with rooms strung out in a line’, in three variants (Drack 1975, 56; Smith 1978, 160–2). 2 Jones and Smith (Brycheiniog 16, 1972, 10 = Pt 6). 3 Fox and Raglan 1951, followed by Smith 1978. 4 Smith 1978, 160–2; Drury 1982, 296–9; Black 1987, 23–6. 5 Agache 1978, Fig. 11. 6 Cf. Marshfield, room 6/7, interpreted as a shrine room (Smith 1985, 251). 7 Drury (1982, 295) points out that the common designation of such rooms as passages is demonstrably false. They occur frequently, e.g. Feltwell, Newton St Loe S [I], RamsburyLittlecote Park [I], Whitwell [I], Worplesdon; Carnac E Bldg [I], Ribemont-surAncre; Vicques W Bldg [I]. 8 Thus John Evelyn went to London in 1676 ‘to take order about the building of an house, or rather an appartment, which had all the conveniences of an house’ (Diary, 4, 98). 9 Arch Cant 87, 1973, 2; a ‘vestibule’, p. 3. 10 Some doorways are assumed. That from the lobby to 1 is hinted at in the published plan (Arch Cant 87, 1973, 4). 11 Jones and Smith (Brycheiniog 12, 1966–7, 53–6) for corner-to-corner connection; Hemp and Gresham 1942 and Smith 1992, 72–4, for interconnection in one range. 12 Oelmann 1938. 13 Basler Zft 9, 1910, 81. 14 Kolling (15 Ber Saar, 1968, 14) notes of Bierbach that room 20 at the east end forms a pendant to 2 at the west end, and de Maeyer (1937) may have a comment, which I cannot now trace, on the same dispositions at Jemelle-Neufchâteau and Mettet-Bauselenne. Two rooms (12 and 13) in the north wing of Matagne-la-Petite are arranged like this. 15 A porticus or veranda of some kind must surely have extended to 8. For the splaying porticus of II see chapter 9. 16 See chapter 7 for other examples of narrow rooms. 17 These words should perhaps be amended to:’. . . against thought’. Cf. to the same end the statistical treatment of incommensurables (Clarke 1990). 18 Ferdière 1988, 107–9; ‘great specialist’, (ref. to Grenier), 27. 19 Burgaud, Rev Arch NS 16, 1940, 53; vasarium, which has several different meanings, is here, presumably, furniture or moveables in a bath, and is found in Vitruvius (F.J. Kilvington, pers. comm.); ‘buanderie’ is a wash-house.


— Notes to pp.59-73 — 20 Black 1985. 21 Why the west porch should be the only part of III set out irregularly is not explained, and the significance of the porches is not discussed (Neal 1974). 22 Châtillon-sur-Seiche/La Guyomerais 2/III (Ille-et-Vilaine) may be another example. The symmetrical completion of the plan in V is unconvincing. 23 Mercer 1975, chapters 4, 5. 24 The veranda at Faversham I is very wide (nearly 4 m) but it is difficult to see what else it might have been. The idea that the Lockleys posts might be the only relic of an aisled building (Rodwell 1978) has nothing to commend it; multiplying such evidence multiplies the absurdity. 25 Piddington S [I] (Northants.) and Sudeley-Wadfield [I] (Glos.) are functionally equivalent to Brixworth [I] but socially less advanced. 26 The cellar is unlikely to have lacked a room above it.

CHAPTER FIVE: DEVELOPED FORMS OF ROW-HOUSE 1 Drack 1975, 59. 2 Drack (1975, 19) drew Ferpicloz as an undivided hall and claimed it as such (RiSchw 394). I think it more likely that Ferpicloz [I] resembled Schupfart-Betburg I, although the hall was not necessarily used in the same way. 3 Smith 1992, 72–5, 108–10. 4 Incomplete excavation of Landen-Betzveld creates uncertainty about later additions, but the grouping of rooms at the ends, each prolonging a lobby, appears intended to maintain a balance. This impression might, nevertheless, be mistaken, as the development of Vicques [II] suggests; there, a very large squarish room in the east wing, heated by a pillared hypocaust, has no counterpart facing it. 5 Said to be the tablinum and summer dining-room (Helv Arch 1978–9, 10). 6 Gerster-Giambonini thought the narrower part of the lobby was a staircase with entry to 2 and 4 from the wider part (Helv Arch 1978–9, 10). 7 Cf. Bocholtz and Bierbach. 8 Or a kitchen and ‘from time to time . . . an eating-room’; the work-hall is assigned, without specific evidence, to the north pavilion 1 (Helv Arch 9, 1978–9, 12). 9 Drack 1975, 59; but even if that were correct it would imply the introduction of reverse symmetry, which is itself an indication of the unit system, in a secondary phase. 10 Masonry joints and the off-axis doorway into the middle room show Bierbach was of more than one building phase but the published evidence is insufficient to distinguish them. 11 15 Ber Saar 1968, 14. 12 Ibid., 39. 13 Cf. the small room intercommunicating with the hall of Blankenheim I. 14 The L-lobby at Latimer has a profusion of doorway positions but the only original one seems to have been at the rear of the room; Branigan 1971, 64 and Fig. 17. 15 Friendship-Taylor (Northants Arch 24, 1992, 101 and Fig. 1) argues that one lobby is in fact a small room built after demolition of the villa core; but it happens to fit into the old plan like an L-lobby and has a doorway in just the right place. 16 Rev Nord 66, 1984, Fig. 5. 17 ‘kleinen Einschlussen’ (Keller 1864, 152). 18 Fremersdorf 1933, 22.


— Notes to pp.73-84 — 19 Room 17 is unlikely, in my opinion, to have been a kitchen; the resemblance of the end rooms to those at Newel or Schupfart II is strong. 20 Smith 1992, 120, 53 respectively. 21 Ann Bret 79, 1972, 221–2. 22 Ibid., 222. It is not possible to check this assumption since no indication is given of the height to which walls survived. 23 Ann Bret 79, 1972, 224. 24 This assumes that the lobby 13 is original, not an alteration as in Reutti 1974, appendix 4. 25 Drury 1982, 295–8. A weakness of this article is that no attempt is made to explain rooms not forming part of the proposed grouping. 26 Black 1985. 27 Rooms not numbered as in summary report, Gallia 35, 1977, 333. 28 It was partly dug and published by de Schmidt, Receuil d’antiquités de la, Suisse I, 1761, Berne; not consulted, but cited Keller (1864, 129–31 and Tafel 16). 29 Drack 1975, 60. 30 Also Feltwell, Whitwell and Carnac E [I]. 31 Buchs [I], following Drack (1975, 60), is an enlightening example. To the east and west of the middle room are blocks of three and four rooms respectively in which a short passage leads to the largest room of the unit. The west block resembles the domestic end of Schupfart I, the east block the east block of Newport, both turned through 90 degrees.

CHAPTER SIX: DEVELOPED FORMS OF HALL HOUSE 1 Something similar was found at Beringen-Lieblosental (Anz Schw Altkde, 19, 1886, 332). 2 TrZft 24–6, 1956–8, 512. 3 Bollendorf I (Fig. 2), like many other hall houses, has just such a room, though without the cellar; it, too, had no access to the porticus. 4 FdbaSchw NS 9, 1935–8, 108. 5 Seon-Biswind (Switz.) may be a third (Argovia 57, 1945, 224). 6 Following Drack, Jb Solothurn 40, 1967, 447; but the phases he proposes (1975, 59 and, in more detail, 1988, 407) are not wholly clear. 7 Drack 1975, 56, para. 4. 8 Drack, RiSchw 1988, 535. 9 FdbaSchw 13, 1905, 64. 10 Camb Ag Hist I, 142–5. 11 de Maeyer 1937, 130–1. Applebaum refers to the villa as Al Sauvenière, which is different. 12 Ann Namur 24, 1900, 11–20. 13 de Maeyer, Afb. 8B. 14 Doc Charleroi 7, 1875, 104. 15 Camb Ag Hist I, 144. 16 See above, ch. 3, n. 7. 17 Usually cited by its local name, although TIR 32 Mainz distinguishes it from NeckarzimmernUnteren Au which is invariably cited by the commune name only. 18 Schumacher 1896, 4. 19 E.g., Courcelles-Frécourt (Moselle); and Etalle, Rulles-Chaumont and Nadrin [I] in Belgium. I am indebted to Mr Alain Thomas for information about Rulles. 20 BJb 133, 1928, 62. 21 Wightman 1970, 152.


— Notes to pp.84-100 — 22 Room 8 is claimed as a staircase from its narrowness and position between two larger rooms (TrZft 34, 1971, 154). Much ephemeral evidence was recorded but not a scrap of it testified to a staircase. 23 No compass points are given (Jahresber Schwaben 12, 1847, Beiträge, 25–30). 24 The alleged terracing of the hall – ‘yard’ in the report – is unconvincing; a slightly sloping floor and crypto-porticus, or more properly, undercroft, are more probable. 25 By the only certain doorway, its threshold being still in place (FdbaSchw NS 9, 1935–8, 108). 26 GrossSachsenheim has, in front of the porticus, a projection that does not look, in a smallscale plan, like a porch or the foundation for steps; it may be some comparable feature. 27 This room, interpreted as an open inner court, was presumed to have a doorway between two buttresses (Int Archiv Ethnog 24, 1918, 3). 28 An alteration in room 15 which I do not fully understand is its division by a timber-framed partition which ‘lay crosswise through the room from north-east to south-east’ (Int Archiv Ethnog 24, 1918, 6). 29 Oelmann’s uncertainty as to whether the doorway was original to 30 may now be resolved; if the implications of Kinheim are accepted, it was (BJb 123, 1916, 215). 30 Raversbeuren is relevant to this discussion but its phases are hard to unravel satisfactorily. 31 As the archaeological evidence showed (BJb 123, 1916, 215). 32 An upper storey is postulated on the ground that the height of the adjoining hall demands one, at least over 7–9 (Tr Zft 34, 1971, 154). The reconstructed plan, Abb. 43, showing a second staircase at the north end, conflicts in this point with the text, 216. 33 This is perhaps to follow classical precept too closely in a very un-Roman building. Baselb Hmt b 5, 1950, 22.

CHAPTER SEVEN: PROBLEMATIC HOUSE TYPES 1 Dmpfli Bad-Württ 1, 1972, 38ff. 2 Mühlacker-Lomersheim (Bad.Württ) has a hall only 10.50 m wide which is claimed as a yard (FdbaB-W 16, 1990, 180–1). 3 FdbaB–W 1, 1974, 509–11. Bierlingen is not discussed here because its plan, reconstructed as perfectly symmetrical, is open to doubt. 4 Choisy 1887, 152, 162. A late twelfth-century roof at Angers with tie-beam 18 m long has remarkably small scantlings for its clear span of 15.40 m (Centre de Recherches, Charpentes, Parts 1, 2). Wooden clamps forming hangers from posts to tie-beam appear at Mantes not much later; for Roman metal hangers, Adam 1984, 226. 5 Roofs of 1618 and 1777; H. Götzger and H. Prechter, Das Bauernhaus in Bayern I, Regierungsbezirk Schwaben, Munich 1960, 201, 179; also many without precise dates. 6 Bondorf, Gaubatz-Sattler 1994, Abb. 127; Inzigkofen, Dmpfl Bad-Württ 2, 1972, 38; WestheimHüssingen, Arch Jahr Bayern, 1980, 134. 7 FdbaB-W 3, 1977, 405 and n. 4. 8 FdbaB-W 16, 1990, 446–8. 9 Chapter 9, pp. 126–27. 10 Cf. Winkel-Seeb D, with a clear span of 14.80 m, which was reconstructed with a roof (Gerster 1973, 80). 11 Paret 1932, 38–9. 12 I have used the more detailed plan (Naeher 1885, II), rather than that in the excavation report (BJb 74, 1882, Taf. X).


— Notes to pp.100-12 — 13 Its location in a district prone to avalanches and on a Roman road are reasons for calling Alpnach a mutatio (Pferdewechselstation) (Drack, RiSchw, 320); they are not compelling. It was regarded as a villa by Scherer, who dug it, by Schulthess (8 Ber RGK 1913–15, 106–8) and by Drack himself at first, who thought all four rooms encroached on a hall (Drack 1975, 58). 14 Walls remained to heights which made door positions certain (Mitt Zürich 27, 1909–16, 232). 15 The purpose of the passage(?) west of 7 is unclear. 16 Also Kösching, known only from an aerial photograph. 17 ‘ Gang’: 0.60 m (FbaSckw 19, 1911, 84). 18 Ibid. 19 FdbaB-W 13, 1988, 356. 20 Some villas have, in a position appropriate to a workhall, a secondary structure looking like a hall but devoid of confirmatory detail, e.g., Aulfingen (Bad.-Württ.) – 19.70 × 15.70 m; Cromhall (Glos.), 21 × 12 m; Inzigkofen B, 19 × 13.5 m; Nagold (Bad.-Württ.), 13 × 10 m; Remmingsheim, 20 × 17 m. The hall (or barn?)/yard problem applies equally to them. 21 Agache 1978, 287. 22 Plachy-Buyon is excluded because it does not correspond to the plan of Plachy-Buyon/les Trois Cornets in Agache 1978, 338, nor, I think, to any site in Agache and Bréart 1975. 23 Bouchoir, Démuin, Dury, Fontaine-le-Sec, Vaux, Wancourt. 24 FdbaB-W 13, 1988, 356. 25 Some difference of interpretation is required by the different porticuses – open-ended at L’Etoile and Harbonnières, on all sides at Ovillers. 26 Elongated hypocausted rooms are not uncommon; e.g., Aulfingen (Bad.-Württ.) with a room 9.90 × 3 m and Hohenfels (Fig. 33). 27 TrZft 14, 1939, 249; hearths not shown on plan. 28 Final report awaited; further details proved unobtainable. 29 Hartley and Fitts 1988, 81. 30 Petrikovits’ building sequence (BJb 143–4, 1939, 408–23, esp. Tafel 79) is here reversed. Once the wall of [II], only 1 m from [I], was built, with a narrower doorway somewhat offcentre to that of [I], it is difficult to see any point in an opening so wide and so placed. Therefore the latter is earlier. 31 To regard these structures as later insertions within what is here taken to be [II] (as in Schmidt 1971, 21 and RiHessen, 459) makes the two parallel walls about 0.40 m apart inexplicable. 32 Percival 1976, 81; de Maeyer 1937, 75 respectively. 33 Published as Bertenbreit-Ottenhardt (Schwäb Mus 1930, 159–60); W. Czysz, Die Römer im Ries, 83. 34 Czysz (as n. 33) makes the middle room 1 open to the porticus 2, so producing a compact row-house; cf. Villers-sur-Lesse/Genimont. 35 Smith 1993, 12–13. 36 Sir Roger Pratt (Gunther 1928, 24). 37 Rooms 8/10 are said to have been built as one (Ann Bruxelles 19, 1905, 315). No reasons are offered. 38 The end rooms 41 and 38, 39 – the latter separated only by light partitions – may well have been altered when the baths were added (Ann Bruxelles 19, 1905, 311). 39 Black 1987, 112. 40 The only other example is Zazenhausen (Bad.-Württ). Listed in TIR Sheet 32 Mainz as a farm, it has an incomprehensible plan. In Spain the axial corridor is used in a more limited


— Notes to pp.112-28 —

41 42 43 44 45

way to group several well-appointed rooms together at, e.g., Quintanilla de la Cueza and Comunion-Cabriana. The reconstruction drawing showing it and the wings and added outbuildings as twostoreyed is unbelievable (RiBad-Württ, 318). Wagner 1908–11, II, 133–4. Described by Wagner as ‘Hof oder Gang’. Without the apse 5 and room 6. The scale of the Walsbetz plan is as de Maeyer (1937, Afb. 26) but is doubtful. I have not found it possible to consult the original publication.


9 10 11 12 13 14 15 16 17

18 19 20 21 22

Koepp 1924, 7. The reconstruction in Gaubatz-Sattler (1994, 138–9) is wholly implausible. Ann Namur 24, 1900, 25. de Maeyer 1937, 51. Cf. Stuttgart-Stammheim, Fig. 27. The phase I door jambs and wall are perfectly visible in Spitzing 1988, Abb. 7, and in II the floor level must have been some two courses of masonry higher. The suggestion (Spitzing 1988, 59) that the apse accommodated a staircase is impossible, as consideration of the resulting first-floor arrangements shows. Rothselberg was published as being of one building phase, yet the divorce of cellar and porticus and the lack of alignment between the doorway into the porticus and that into the hall hint at the addition of the porticus-and-pavilions to a simple kind of hall house. These rooms were thought to be a bath suite by Schib (Jb Solothurn 14, 1937, 319–20), a claim rejected by Laur-Bellart (JbSGU 28, 1936, 73). A length of lead water-pipe in the west part suggests this (Drack, Jb Solothurn 40, 1967, 447). As, for example, in the sizes of rooms in row-houses. Smith 1992, 87–8; Smith 1993, 16–18. Smith 1992, 108–10; Smith 1993, 142. Published as having a sinuous outline: I follow Richmond (1969, 55) in doubting this. Swoboda 1918, 107. Württ Vierteljahresh 11, 1888, 30; the overall plan shows the east porticus incorrectly, the detail plan correctly, in this particular. Osterstetten was carefully dug by the standards of the day and the reporting good, so it is unlikely that a stone wall was overlooked. Stratford-upon-Avon-Tiddington itself is a small version of Bilsdorf or the principal house at Ludwigsburg-Hoheneck. Hémecht 22, 1970, 378–80. I envisage a porter in a kin-group as an honoured and important member of the community, in the sense that a monastic porter was, rather than in the modern sense of a menial. Swoboda 1918, 106–7. Smith 1983, 241–2, citing earlier literature. It is unfortunate that Richmond (1969, 54–5), in clarifying the form and phases of Brewood, could not set out his ideas fully; much remains to be explained. Reconstructing the appearance of Brewood-Englewood or even the phases may well be an impossible task because the published plan contains too many structural anomalies.


— Notes to pp.128-40 — 23 See Smith 1982. 24 Gorges 1979, 139, 358, 477. I have not consulted the sources there quoted. 25 Current Arch 122, 1990. It is hard enough to envisage uses for an undivided long groundfloor room, harder still to do the same for a first-floor room above it, entered at one end, and hardest of all to devise a reasonable access to other upstairs rooms. The staircase rises to the end of a long room, and the absence of partitions in the room below implies their absence above. Upstairs partitions in English vernacular houses are invariably placed above ground-floor ones in order to prevent deflection of the binding beams.

CHAPTER NINE: THE PORTICUS-WITH-PAVILIONS: PORTICUSES 1 Oelmann, 1938. 2 Oelmann, BJb 123, (1916, 216) was evidently puzzled by the two short projecting walls and made the implausible suggestion of buttresses. Mylius (1933) in his reconstruction insisted that the plan was perfectly symmetrical and consequently had to ignore them. 3 This interpretation modifies Smith 1978, 160. 4 Agache 1978, 276. 5 Smith 1993. 6 Bramdean (Hants.) looks as if its porticus was built with one open end, otherwise it is hard to see why it should stop short of the other end. Its development is uncertain. 7 Neal 1974–6, 6–7. The suggestion that it was a latrine because it had a drain sounds like a late twentieth-century preoccupation transferred to the second century. 8 This arrangement has something in common with Stolac 5. 9 The room next to (east of) the central room may have been entered from it, according to Thomas (1964, 220), but since the doorway could not be precisely located I am inclined to disregard this idea. 10 Following Drack 1975, 59. 11 I follow Samesreuther’s phasing, which makes this villa structurally and socially intelligible; Rothkegel’s raises more questions on both counts than it answers. The two interpretations can be compared in Rothkegel 1994, 19–21. 12 Swoboda 1969, 54, para. 2. 13 Aus’m Weerth compared the wall stub there to one at Allenz (BJb 39–40, 1866, 259). 14 BJb 39–40, 1866, 260. 15 Halbardier and Thomas 1987, 127. 16 Or quasi-pavilions. The north-east corner block is structurally separate from the rooms behind it and is a pavilion; the south-east block is structurally one with the room behind it and is properly a wing. 17 Gayton Thorpe N may be one; if the grand doorway in [F] was used only on grand occasions, the rear porticus may have met everyday requirements. 18 Bayer Vgbl 22, 1957, 223. Cf. also the slightly different application at Hechingen-Stein (Bad.-Württ.). 19 Köln-Braunsfeld, published with one such empty projection, has been reinterpreted with orthodox pavilions (Kölner Jahrb 5, 1960–1, 9–13). 20 Not all empty projections on villa plans are really so. A large one at Upchurch-Boxted (Kent) was ‘the exterior chamber’ with a cellar or hypocaust (Arch Cant 15, 1883, 105). 21 The completeness of the extant remains and the fullness of architectural detail shown on the


— Notes to pp.141-57 —

22 23 24 25

26 27 28 29 30

plan make it unlikely that the empty projection at the south-west corner is really an undiscovered pavilion. Pepys, Diary, 14 July 1664 (garden); 12 May 1666 (bridge). Drack 1975, 59. RiSchw 418; Swoboda (1918, 92) uses it of [II]. Hamblain-les-Prés (Fig. 67) would be more credible if the term could be applied to what are called angle-turrets (tourelles d’angle). One is only 2.2×1.5 m and the other even tinier at 1.8×1.6 m; and how were they reached? Timber verandas also vary, that at Faversham being unusually wide. An alternative explanation might be found for Hemel Hempstead-Boxmoor I on these lines. Smith 1987. Gaubatz-Sattler 1994, 115. I take Aufenhaltsstatte to imply that the dwelling-places were occupied by those doing the work. Nikolov 1976, Fig. 58.


6 7 8 9 10 11 12 13 14 15 16 17 18 19

Röm-Germ Korrbl 7, 1914, 54. Among the rare instances of end-entrance halls is Dirlewang (Fig. 39). The double-ended hall tapers slightly, as if it had been built against the yard wall. British readers will have a feeling of inevitability on learning that this was the bailiff’s house (G. Steinmetz, Verhand Hist Ver Oberpfalz 67, 1917, 28). As Reinecke pointed out, it can hardly have been part of a bath suite because other bath rooms are lacking (Röm-Germ Korrbl 7, 1914, 55). It is a valid point because, although uncertain evidence of rooms was found, distinctive features such as a plunge-bath might be expected to survive. Confirmed by excavation (Arch Ausgr Bad-Württ 1992, 179–82). FdbaSchw 19, 1911, 100–5; Arch Ausgr Bad-Württ 1992, 181. Nothing like it is to be found in the Trier region, the closest resemblance being to the baths and hall (only) at Meckel (Koethe 1940, 71). Arch Ausgr Bad-Württ 1992, 179. For the division of the villa and an interpretation see Surrey Arch Coll 65, 1968, 1–70 and 72, 1980, 63–8. The idea that building 6 V turned its back on the rest (Smith 1980, 67) is unfounded; what was added was a rear porticus with end-entrance. Agache 1978, 336, 339. A building of less easily explicable development but standing in a somewhat similar relation to a hall is Sontheim a.d. Brenz 2 (p. 105). Spitzing 1988, 65. Horn and Born 1970. Cf. also the brief remarks by Drack 1975, 58. The larger villa at Regelsbrunn is another (Thomas 1964, 258). Cardiff-Ely and Newton St Loe are others. Following Richmond 1969, 57. Archaeologia 71, 1921, 149. The two buttresses in the east wall of 3 suggest it was aisled; this would accord with its being split up by timber partitions.


— Notes to pp.157-71 — 20 The villa was dug in the 1960s by the methods used before 1914. The RCHME 1983 publication (Arch J 140, 1983, 129–50) is a salvage operation. Here the Commission’s lettered phases are replaced by numbers. 21 Llantwit Major, N and S ranges; probably Badbury, E and S ranges; and in slightly different form, at Sudeley-Spoonley Wood, where the different widths of the three ranges resemble those in the three buildings at Aiseau. 22 Many aspects of the Hambach courtyards are dealt with in Gaitzsch 1986. 23 Too little is known about the area to north of the old yard to know whether it had been developed in any way before the enlargement took place. 24 Gaitzsch 1986, 411. 25 If the scale of Agache 1978, Fig. 24, is correct it is 74 m long and it is disturbing to find that the east and west porticuses are 7.5 m and 6.5 m wide respectively. These are incredible figures; the porticuses of the great villa of Fliessem are only 4–4.5 m wide and the front porticus of Köln-Praetorium is only 5.4 m. Probably the scale should be 1:1000, not 1:2000. 26 Agache (1978) has Athies, Estrées-sur-Noye, Mézières-en-Santerre/le Ziep and WarfuséeSud; Namps-au-Mont, Quevauvillers and Ribemont-sur-Ancre lack the enclosing wall. Laboissièreen-Santerre appears only in Agache and Bréart 1975, Fig. 93. Behencourt is excluded because of doubts about the scale. 27 As Koethe suggested (TrZft 9, 1934, 22). 28 There is nothing to be said for Richmond’s notion (1969, 57) that the north block, the workhall, formed [I] and the east range is a later phase. Matagne-la-Petite has some resemblance to Spoonley Wood, particularly in the main range. 29 Hurst argued convincingly for a justice room at Box (Wilts Arch 81, 1987, 29–32). 30 Wellow, Pitney and Keynsham-Cemetery are probable. 31 The wings of Folkestone W[I] are surely an addition – that to S must be – and may denote the conscious abandonment of a unified court of any kind. 32 Port-le-Grand (Fig. 68), Namps-au-Mont, Ville-sur-Ancre, etc. 33 De Caumont’s plan, reproduced in Grenier (1934, 801), differs in the proportions of the east wing rooms from that in CAG 29, 158. I have followed Grenier, as being the more complete. 34 De Caumont’s plan, reproduced and reduced in Grenier (1934, 869), agrees with Jacquemet (1857, 180) but does not correspond with the only dimensions given in the latter’s text: the porticus is 3 m wide and 60 m long. The resulting scale of 1:800 makes sense of the buildings whereas 1:500, which is Jacquemet’s drawn scale, does not. 35 Referring to Bollendorf, Fig. 27 above (Steiner 1923, 5). 36 Goossens thought the plan of this workhall, the lack of heated rooms and the fewness of the finds precluded a dwelling although, in the north ‘corridor’ (Gang), the soil was burnt red in many places (Int Archiv Ethnog 24, 1918, 10). 37 The pilasters on the north side of the workhall can only be for an entrance; the projections to east and west are probably buttresses. I take the porch-like room in the porticus to be a lobby with a room to the east of it. The remainder of the porticus was primarily or solely a corridor. 38 Lischewski regards 5 as the Herrenhaus?, 4 as a house and the granary is also his opinion (RiHessen, 362). 39 Not on plan (Bull Morbihan 1899, 138).


— Notes to pp.172-84 —


Millar 1977, 41. Ibid., 42. Ibid., 30–1. Plan most easily accessible in Wightman 1970, 116. Ward-Perkins 1981, 78–80. Thought by Cunliffe (1971a, 1, 219) to be of AD 75–80 (reiterated, Cunliffe 1991); and by Black (1987, 85) to be of AD 90–110. The Flavian Palace was begun c. AD 75 and inaugurated AD 92. 7 Cunliffe 1971b, 124. 8 The question of whether Fishbourne was a palace or a villa ought not to be linked to its dating and alleged ownership by the client king of the Regni, Cogidubnus. Those who deny that it is a palace in the sense defined above have to explain why it is so unlike any known villa. 9 The suggestion that this was a granary can be dismissed out of hand (Bud Rég 14, 1945, 73). 10 Difficulties presented by cellars and floor levels (Bud Rég 18, 1958, 74) are here ignored; the fact that the two blocks of rooms are a pair is more important. 11 Professor J.J. Wilkes pointed out to me the relevance of the residences of commanders in legionary fortresses and auxiliary forts, as showing the kind of accommodation adequate for a senator or a Roman knight (eques). Unfortunately it proved impossible to follow up this idea. 12 Information from Professor Wilkes. 13 A point also made by Grunewald 1974, 263. 14 H. Hurst and others (Wilts Arch 81, 1987, 28). The discussion at pp. 29–32 is important. 15 Ward-Perkins 1981, 78–80. 16 Hurst et al. 31. 17 Ibid. 18 Castro 1982, 207, 314 and Plan 7. Although ‘arquitectura aulica’ has the connotation of a princely court, the dispensing of justice is a princely activity which may justify the application of Castro’s words. 19 The idea that Aguilafuente began as a row-house or ‘villa linéaire à galerie de façade’ (Gorges 1979, 131) has no obvious foundation and the disposition of rooms makes it unlikely. 20 It is a schematic plan (Congr Arch 41, 1874, 41). 21 Wilson and Sherlock 1980, 16. 22 The suggestion that this range is the original nucleus of Woodchester ignores its extraordinary width, for a British villa (Clarke 1982, 218). 23 Clarke 1982, 199. The suggestion that so grand a room may have had an upper gallery around it (de la Bedoyère 1991, 156; 1993, 114) is inherently unlikely – no grandee would have allowed himself to be looked down on by inferiors in a ceremonial situation – quite apart from the lack of any trace of the considerable staircase that would have been needed. 24 Gallia 31, 1973, 133. 25 Thomas 1964, 80. 26 Gorges 1979, Fig. XXII. 27 Gorges 1979, Figs XXXVI, XXXVII. 28 In the excellent summary report of the excavation the villa was separated into four phases.


— Notes to pp.185-98 —

29 30 31

32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50

51 52 53 54 55 56


I have ventured to divide the first into two on the grounds that the first small temple was soon replaced by a larger one(?), the baths are imperfectly aligned with the west ‘wing’ and mask the workhall to west, and the basilica appears to be part and parcel of the baths/ temple layout. This has entailed renumbering the phases. Cf. de Maeyer 1937, 82; also Smith 1983, 240–2. The room has unusual proportions but its location leaves no doubt about its function. Fig. 26 (‘Villa urbana du IVe siècle’) is mistakenly captioned and referred to (heading, p. 148) as the last state of the villa, and Fig. 20 (‘Etat probable du Ve siècle’) is captioned as the fourth-century plan. The continuity of the baths from Fig. 11, Haut-Empire, to Fig. 26 and their disappearance in Fig. 20, Ve siècle, confirms the point. Ferdière, I, 198–9, repeats the error. ‘Villa de plan diseminado’ (Castro 1982, 64–9). In the light of the difficulty of exploring this phase I, the doorways indicated on the plan are taken to be assumption, not ascertained fact (Fouet 1969, 46). Fouet 1969, 50. Called an ‘oecus?’ (Fouet 1969, 49). Fouet 1969, 76. Wightman 1985, 111. Then spelt Thuit (de Caumont 1870, 388–9). The distance is uncertain; the plan lacks a scale. Roger North, quoted in Mark Girouard, The Making of the English Country House, 1978, 122. Wightman 1970, 153, with plan. It is impossible to say on present evidence whether Welschbillig was a palace or a luxury villa. RiRWa-Pf, 671–2. RiRWa-Pf, 671. Another, more removed from the rest of the villa, is in the north-west wing at Téting. No report was published. A description and discussion appear in Grenier 1906, 159–68. East and west are of 10 m and 12 m diameter respectively (Grenier 1906, 160). The scale published in Grenier (1934, 824) does not match the measurements given on the plan. Grenier 1906, 163, 160 respectively. Great Witcombe is a simpler version. Wightman 1970, 145; Mylius 1924, Abb. 1, Taf. IV–V. A similar reconstruction recently proposed for the great room at Woodchester is even more unlikely (de la Bedoyère 1991, 156, with gallery, and 1993, 58 and Pl. 1, without). Despite its obvious importance I say nothing about Chiragan, having failed to understand its most important parts. A very different kind of villa, Lullingstone, has what is to me an equally incomprehensible plan. Wightman 1970, 145. Gunther and Kopstein 1985, 157. Fliessem I excluding pavilions, 35×21 m; Haccourt I, more than 30×22 m. RiRhld-Pf, 368; that this is the intended meaning of Hof is confirmed by the English edition of the site guidebook which includes the baths in the ‘farm buildings wing’ (Cüppers 1979, 8). Zittersdorf/Haut Clocher-Saint Ulrich in TIR 32, Mainz. ‘Mil dem Saal 17, der genau so in der Mitte der westlichen Haelfte liegt wie 4 in der Mitte der oestlichen, stand er gewiss in unmittelbarer Verbindung’ (Jb Lothring 10, 1898, 172). I take ‘gewiss’ here to mean ‘surely’ in the sense of a strong probability that a statement is correct. The evidence was the circular room 88 with its four niches (Grenier 1906, 150). This was dismissed by Koethe (1940, 128).


— Notes to pp.198-213 — 58 Grenier 1906, 152. 59 Gallia 29, 1971, 28. 60 Wichmann was uncertain on this point (Jb Lothring 10, 1898, 172). Grenier (1906, 145) mentions the discovery of many column fragments.

CHAPTER TWELVE: THE VILLAS OF SOUTH-EAST EUROPE 1 Biro 1974; Dremsizova-Nelcinova 1969; Thomas 1964; Vasic 1970. 2 The pavilions of Szentkirálysabadaja-Romkút, which are drawn as towers and said to be for storing corn, hay and other crops, show absolutely no way of being entered from the yard (or elsewhere), nor any domestic rooms around the ‘yard’ (Thomas 1964, 121). The illustrator seems to have preferred a toy fort as a model to Köln-Mungersdorf with its corner pavilions. For Gyulafirátót, Thomas 1964, 44, n. 112 and Abb. 20. 3 Chapter 7, above, 105. 4 The north wall extending beyond the squarish house is likely to be for a furnace, nothing more, as the lack of any return to the yard wall confirms. 5 Calling it quadriburgium-like and fortifying it, as in the reconstruction drawing, makes matters worse (Thomas 1964, 113). 6 Marshfield (Smith 1987, 250). 7 Following Thomas (1964, 208), using Majdan as a parallel to decide between her alternative sizes. 8 Above, pp. 132–34. 9 As TIR L34; Miclea and Florescu 1980, I, 64 have Minerau. The alleged corridor 0.80 m wide running the whole length of the south side, and with the walls of every adjoining room crossing (or cut by?) it, is omitted in Fig. 56; it is probably the effect of refronting that elevation. Cf. the similar feature at Aiud. 10 It could be argued from the setting-out of the transverse entry that all internal walls are later insertions, but that results in an improbably large hall. 11 As pointed out by Groller (RLiÖ 6, 1905, col. 41). 12 Claimed as such by comparison with one in Pompeii; Groller 1905, col. 40. It appears an unlikely situation for a kiln and no wasters are mentioned in the vicinity. 13 It could have been roofed in line with the adjoining rooms but, in the absence of any definite Roman evidence for such a building practice, a wing is to be preferred. 14 Mitrofan, Acta Mus Napoc 10, 1973, 134; Miclea and Florescu 1980, 87. 15 For example, in the relation between rooms 2 and 3 and the apsidal-ended room 9. 16 The incredible suggestion is made that the entrance-hall was an open yard (Thomas 1964, 65). 17 Weitersbach and Betzingen are examples. 18 Icannot suggest any explanation of these rooms. 19 Strupnic (Yugosl.) has a certain resemblance to Deutschkreuz. 20 Assumed orientation; none is given in Dremsizova-Nelcinova (1969, 505) or Nikolov (1976, 168). 21 I am indebted to Sergei Kovalenko for supplementing the French summary with essential details from the Bulgarian text. 22 Described as the east wing in the report. Sergei Kovalenko furnished the details mentioned here. 23 The idea that this part had an upper storey is unsupported by evidence. 24 Additional to the Stolac numbers in Wilkes 1969.


— Notes to pp.213-24 — 25 Regelsbrunn 1 has two pilasters dignifying the entrance to the porticus of the principal house (1), showing that it was reached from the south gate past 2 – a small-scale instance of the intended effect of in-line planning, to emphasise the importance of the house at the end. House 1 is not in a dominant position in the middle, like Vierherrenborn-Irsch or Köngen, but sufficiently off-centre to show that the second building is of some importance – a habitation of some kind rather than the granary its excavator supposed it to be; Groller, cited Thomas 1964, 257. Thomas expressed no view about its function; the suggestion that it was vaulted and had an upper storey has nothing to commend it. 26 It was, of course, the bailiff’s house (Nikolov 1976, 33). 27 The claim that it was fortified is groundless (Dremsizova-Nelcinova 1969, 510). 28 Bull Inst Arch Bulgare 3, 1925, 299–300; Sergei Kovlenko provided essential details. 29 Nothing like it appears in the Trier district (Koethe 1940). 30 It looks as if the courtyard was originally open to east and was only closed by a new range later. 31 Thomas 1964, 60. 32 Miclea and Florescu 1980, II, 187.


6 7 8 9 10 11 12 13 14 15 16 17

18 19

Schaaf, Ausgr Deutschland, 189–91. Bittel et al. 1981, 84. Herrmann and Jockenhövel 1990, 241. Schubert 1994. I am indebted to Dr F. Schubert for very kindly providing me with a copy of this most important article. It is impossible to do justice to the wealth of material from Holland despite some useful comments by Mr Joris Aarts of the Free University of Amsterdam, and what follows about that country is inevitably inadequate. Manching will be of comparable importance when the houses are fully published. Collis 1986. Slofstra 1991. No report of the excavation has appeared. Mr S.J. Catney provided additional information. Rodwell 1978, 34–7. Gelling, Proc Prehist Soc 43, 1977, 282; ‘quergeschlitze Palisaden Graben’ (Schindler 1977, 26). The Vix building is an ‘à fonction annexe’; Villes 1985, Fig. 5. A large four-aisled building at Mont Beuvray is structurally incomprehensible (Bulliot 1899, 2, 166–72). Verwers 1972, 88. Verwers 1972, 77, 88. Three houses, J, M, P, had upright gables (Verwers 1972, 86). Hulst, Ber ROB 23, 1973, 105. Ibid., 106. The house is either of six or seven bays according to whether one of the axial post-holes, hatched in Sanden and van der Broeke 1987, 58, Afb. 5, open in Slofstra 1991, 142, Fig. 7c, is thought to belong to the building or not; I am inclined to accept it. Slofstra 1991, 141, n. 27, citing an observation by de Boë. These remarks are based on the only source available to me for Hoogeloon, the small-scale plans Figs 12, 22 and also Fig. 26 (Slofstra 1991).


— Notes to pp.226-39 — 20 Van der Sanden 1987, 64. I am grateful for the help of Karen Waugh in obtaining this and other Dutch publications. 21 Trier 1969, 145, n. 15. 22 The posts are for a veranda; above, Fig. 11; discussed, Rodwell 1978, 32–4. 23 Too irregularly set-out to permit reconstruction; Hawkes and Hull 1947, 89–91. 24 Neal et al. 1990, 32. 25 Ibid., 34. 26 Verwers 1972, 93. 27 For example, at Oss-Ussen (van der Sanden 1987, 62 (plan)). 28 Stanford 1974, Fig. 5 (Hut F5) and 123–5; Arch J 141, 1984, 61–90, esp. 71; Stanford 1991, 60–1. Dr Stanford very kindly drew my attention to this material. 29 Dedet 1987 has an importance beyond its immediate concern with Languedoc. The houses mentioned are described and discussed at pp. 185–96 with plans at p. 191. 30 Here as elsewhere Roman numerals in the original publication are replaced by Arabic to avoid confusion with phase numbers, which are Roman throughout in this book. 31 Schindler 1977, 42. 32 Roymans, ROB 1988, no. 356, 355. 33 Dedet 1987, esp. p. 192. 34 van Es 1981, 181. 35 Other settlements with houses of this type are Haamstede-Brabers and, just over the Dutch border, Weeze-Baal (Niedersachsen); van Es 1981, 168–9, 179–80. 36 Hingley and Miles 1984, 63. 37 Grimes in Frere 1960, 21–6. 38 Proc Prehist Soc 1977, 273–5. 39 Harding in Hawkes 1973, 55–6. 40 But it is not always possible to know whether the hearths are contemporaneous or were in use simultaneously, as pointed out by Dedet (1987, 193). If, though, a hearth had fallen out of use for any considerable time, its remains might not have been left occupying useful space, as is evident in the north-east room.

CHAPTER FOURTEEN: MODES OF ROMANISATION 1 The plan redrawn as Fig. 31 follows de Maeyer 1937, 106 but looks too small and I have not consulted the original publication. Nevertheless, it illustrates the point made in the text. 2 Jarrett and Wrathmell 1981, 250. 3 Buildings could not always be related stratigraphically and the views expressed here are not always those of Jarrett and Wrathmell. 4 Jarrett and Wrathmell (1981) failed to discuss the implications of the finds and the specialist reports for the history of the site and since they chose to dismiss the need for an index (p. vi) I may have missed some important evidence or observations. 5 Smith 1985; Webster and Smith 1987. 6 The same development at Radley-Barton Court Farm may be a similar response to Roman rule. 7 The two parts must have been phases of a single building campaign, even if implementation of the second was considerably delayed. 8 This demonstrates again how right Wightman was to underline the importance of baths in early villas or, in the case of Barnsley Park IV, a proto-villa (Wightman 1985, 110–11).


— Notes to pp.239-55 — 9 10 11 12 13 14 15 16 17 18 19 20

21 22 23 24 25 26 27 28 29

30 31 32 33 34 35


37 38 39 40 41 42

Bloemers 1978, Abb. 42. This interpretation of the site appears in Smith 1987. Agache and Bréart 1975; Smith 1987, 247, 249. Leday (1980, 2, Pl. LVI), where what seems to be a ditch approaches a row-type house and may run diagonally across it. Cf. the beginnings of Bignor. J Roman Stud 40, 1950, Pl. VI/2. For examples illustrating this and other themes, and for a valuable discussion, Wilson 1974. Yorks Arch J 43, 1971, 180. Note also Rheinbach-Flerzheim (Hambach 512). Williams and Zeepfat 1994, 142. Wrathmell and Nicholson 1990, 282. Ibid., 279; ‘gradual Romanisation’ is inexact because the native farm of V largely disappeared in VI. The continuity of boundaries and house sites raises doubts about the last phase of the preRomanised settlement; why, if some house sites had been abandoned, was such care taken to perpetuate them by masonry buildings? Miles 1986, 14. I follow Thomas (1964, 152–4) for the history of the site rather than Barb in his reassessment (1937, 153–7) or Groller in his excavation report, RLiÖ 6, 1925, 6–50. Thomas 1964, 157. Stead 1980, 13, 35. Agache 1978, 331, Fig. 24; also Lockington (Leics.) (Wilson 1974, Pl. XXII). David Johnston, lecture, Nottingham 1978; see also Johnston 1978, 89. Collingwood 1930, 113. Agache, 1978 and, further afield, at the Swiss villa of Oberentfelden (Fig. 76). This procedure can be inferred in a few English houses where the roof is older than the outer walls, and I have myself observed the gradual progress of just such a rebuilding at Wareham (Dorset). Chapter 7, above, p. 107. Here and at Leiwen the early pre-villa phase of the site was not numbered. Described as ‘quelques murs d’un bâtiment assez fruste’ (Gallia 41, 1983, 320). Saint-Germain-les-Corbeil I has a combination of hall- and row-type houses comparable to Lalonquette I or North Leigh I. Scott (1993a) gives the parish name. Phases as Harding 1984, 21. It is highly unlikely that the round-house was abandoned in [III]; had that been so its position would have been occupied by the house or ‘primary villa’. It becomes even stranger if Harding’s view is accepted: ‘Doubtless the native building-type was considered adequate as living-quarters for farm-labourers or slaves, to whom in any case native ways may have seemed preferable to the trappings of Romanisation.’ Harding 1984, 8. The notion of an upper storey is fanciful (Harding 1984, 10). Could it have been a shrine or temple? No hearth was found. I have in mind what C.E. Stevens called ‘the cry across the abyss’ (J Roman Stud 37, 1947, 134). Northants Arch 22, 1988–9, 60. I am indebted to James Kenney for information about this villa and discussion of its problems. His provisional conclusions are incorporated in the text but Mr Kenney is not responsible for any misinterpretation that may have crept in.


— Notes to pp.257-70 —

CHAPTER FIFTEEN: PATTERNS OF VILLA DEVELOPMENT 1 Cotton and Métraux 1985, 19. 2 The drains hereabouts were for removing surface water, especially that coming from a steep slope to north (Cotton and Métraux 1985, 17, Fig. 3b). 3 Bruckneudorf [I] has no room which corresponds precisely to Blankenheim [I] 30. 4 Wings of this kind occur more widely than in hall-houses. Oberweningen (Switz.), with a strange double-depth plan, has an east wing planned much like Drax and a west wing, like that at Vicques, resembling Lamargelle. 5 Mansfield Woodhouse (Notts.), Carisbrooke and Brading differ only in having extra small rooms. 6 Elsted-Batten Hanger (Sussex), to which David Rudling drew my attention, is a good example. West Dean (Wilts.) and West Meon-Lippen Wood (Hants.) can probably be grouped with it. 7 BJb 123, 1916, 215. I am much indebted to the kindness of Dr Walter Janssen of the Rheinisches Landesmuseum, Bonn, for providing me with a typescript of the report of 1908 by Constantin Koenen. 8 The report is emphatic on this point, contrasting the way the porticus stops dead (totlauft) against the north wing but turns round the corner of the south wing (BJb 123, 1916, 215). 9 BJb 138, 1933, 15. This wall was probably built in IB at the same time as outer rooms were added to both wings, thus making the house truly symmetrical. Had Mylius superimposed his absolutely symmetrical model based on a hexagon (p. 12) upon the real rather than ideal plan of I the fit would have been less than perfect. If this is the correct geometrical model and it does not fit at all the relevant points it was evidently modified in execution; on no account should it override archaeological evidence where that is unambiguous. 10 Mylius divided the room into two because it is ‘improbably large’ (BJb 138, 1933, 14). Oelmann says that how it was divided can no longer be known (BJb 123, 1916, 215). Cf. Budapest-Csúcshegy for a large workhall in such a position. 11 The point is borne out in IIA when the two functions were separated: 25 is no more than a lobby and 21/22 becomes like a small double-ended hall. This makes it more likely that in IA 21 was a passage-room which it was undesirable to perpetuate in IB. 12 It is incorrect to say they are of exactly the same size, as Oelmann did (BJb 123, 1916, 216). 13 BJb 123, 1916, 216. 14 Interpreted as a yard (BJb 123, 1916, 217). 15 See n. 11. 16 This addition was ‘in part open to the south and so may have formed a porticus [Halle]’ (BJb 123, 1916, 218). 17 A yard in the report (BJb 123, 1916, 219); corrected by Oelmann (Mitt DAI Röm 38/39, 1923–4, 211 n. 1) and treated without remark as a hall (Arch Anz 43, 1928, 240). 18 Only the east part had a hypocaust (BJb 123, 1916, 222). 19 The date of the temple is uncertain; it could fit in II but, even if that were so, it is likely to have gained importance in III. 20 The plan shows what look like two holes, perhaps representing door jambs; cf. also a similar depiction of the opening from the baths corridor to the mosaic room 8. 21 Plan of St Stephen-Park Street VIII, Oxf J Arch 2, 1983, 244; and of VII and VIII, Caesarodunum 17, 1982 = Actes du Colloque Villa Romaine, 336. Hemel HempsteadBoxmoor (Herts.) was greatly simplified in phases VA and VB. 22 I am grateful to Dr D.J. Smith for providing me with an offprint of his excavation report, which I had missed on its appearance, and for saving me from some errors; I am solely responsible for the present interpretation.


— Notes to pp.271-88 — 23 24 25 26 27

‘Speisesaal, oecus’ (Fremersdorf 1933, 22). At both ends according to Fremersdorf (1933, 14–15). ‘Dienerschaft und . . . Gesinde’ (Fremersdorf 1933, 31). Fremersdorf 1933, 21, Tafel 4. This may be relevant to the very strange villa of Saunderton (Bucks.) which has never been adequately explained, despite a valiant attempt by Branigan (1966–70).

CHAPTER SIXTEEN: A MODEL OF DEVELOPMENT 1 2 3 4 5 6 7 8 9 10 11 12 13


15 16 17 18 19

20 21 22 23 24 25

Medawar 1967, 87. Leyser 1968, 38. Goody 1983, 121. Jones 1972, 320–58. Ibid., 321. Stevens 1966, 111. Haunstetten (Bad.-Württ.), Landshut-Salmannsberg (Bay.), Beegden (Neth.), Odell (Beds.). Draughton (Northants.), St Lythans-Whitton (Glam.). Comparable to Hardwick-Mingies Ditch (Oxon.). Agache 1978, 138; CAG85, Vendée, Fig. 76. Tailly-l’Arbre-à-Mouches and Nempont-Saint-Firmin (both Somme) are like this. Collingham-Dalton Parlours (West Yorks.), Lechlade-Claydon Pike (Oxon.) and Barnsley Park (Glos.) are examples. The most useful source for Southwick (Sussex) is Black 1987, 102–4 and Fig. 43. This book is more important than its restricted geographical scope suggests; many of the discussions are enlightening. Birkenfeld-Elchweiler (Rhld-Pf.), despite being very incompletely excavated, is more complex than most fragmentary plans. Nennig (Rhld-Pf.) has some resemblance to Fishbourne Palace but is of much later date; Wightman 1970, 146–7. The long row-house built c. AD 65 at Aylesford-Eccles replaced an earlier structure (Detsicas 1983, 120). Leyser 1968, 38. Black 1994, 107. Norbert Elias, The Civilizing Process, esp. vol. 2, State Formation and Civilization (Basle 1939; English edition, Oxford 1982). As is implicit at Barnsley Park, where IV is incomprehensible unless it is a stage towards V. Barnsley Park IV–VI present problems which need reassessment of the archaeological detail. Drack 1990, 287. Otherwise, building across the yard wall seems pointless. Drack 1990, English summary, 287. Not that the author of this study has any reason to object to other people’s following his own practice. Duncan-Jones 1974; Stevens 1966, 109, 121. There are besides one or two halls in Germany and the Netherlands which, by their proportions and development, look rather like British halls; Friedberg-Fladerlach [I] (Bay.) (Fig. 69) has some resemblance to Frocester Court I. I have not pursued this line of enquiry.


— Notes to pp.289-302 — 26 Applebaum 1972, 46. 27 Ant J 9, 1927, 145. 28 Cf. the corn-drying alleged to have taken place in the best room 9 at Saunderton, in Bledlow-cum-Saunderton (Bucks.) parish. Ashcroft accepted this with reservations (Records Bucks 1934–40, 402), Branigan without comment (Records Bucks 1966–70, 261). 29 Smith 1987, 250–1. 30 Smith 1980, 66; 1987, 254, respectively. 31 Hence the alleged parallel of Köln-Mungersdorf, where the distribution of buildings is very different, is invalid (Braat, Oudhdk Med NS 34, 1953, 59 and n. 14). 32 An alternative explanation offered in a lecture (Abbeville, 1972) is a urinoir – a notion so bizarre as to appear so even in villa studies. 33 Britannia 1976, 369. 34 Arch Cant 86, 1971, Fig. 1. 35 Oudheidk. Med NS 6, 1925, Afb. 44, at 49. 36 Detsicas 1983, 105. 37 Payne, Arch Cant 22, 1897, 77. 38 It may also have been a fishpond (Vermeersch, Cahiers Picardie 1981, 152, citing parallels, and for detail, Bull Trim Picardie 1976, 317–20). 39 Reconstructed as a monumental well by Gerster (1973, 75–7). This interpretation – repeated, Drack (1990, 59–62) – sits uneasily with its distance from the two nearest buildings E and B – 35 m and 47 m respectively. 40 Drack 1990, 61, n. 44. 41 Netzer 1988, 41–2. 42 No evidence is offered for the staircases shown in the reconstructed plan (Koethe 1934, 25, 29). 43 Gunther and Kopstein 1985, 156. 44 Ibid., 156; Koepp 1924, 10. 45 Gunther and Kopstein 1985, 157. Their location if not number no doubt changed when the house was enlarged and a clearer separation of service rooms emerged. 46 Not for the use of those misleadingly called ‘indoor servants’ (Mylius 1924, 122). 47 Mylius 1924, 127. 48 Fremersdorf 1933, 310. Fremersdorf is unusual among archaeologists in perceiving that all the farm buildings would have been dwellings too. 49 I regard this as more likely than a peasant militia (Gorges 1979, 323). 50 Following Nikolov 1976, 14 and Fig. 12. 51 Spitaels 1970. 52 Excluding a timber phase (Ebnöther 1991, 250–1). 53 The paved area (Agache 1978, Fig. 13) of elaborate geometrical outline is an ornamental pond by analogy with the larger one at Welschbillig (Rhld-Pf.). 54 J Roman Stud 37, 1947, 133. 55 Arch Cant 88, 1973, 9. 56 I have not studied bath buildings. A British equivalent to Koethe 1940 is needed, also an examination of the many instances of two baths in a house, sometimes a fairly small one, as Milton Keynes-Bancroft V (Bucks.). 57 A word, Oelmann said, to be taken with a grain of salt (BJb 133, 1928, 136, n. 1). The point made in the text comes from the same page. 58 Stevens 1966. I do not altogether understand this article; fortunately its general significance is clear. 59 Applebaum 1972, 236–7.



AASR Associated Architectural Societies’ Reports and Papers Acad Bulg Fouilles Rech Academie Bulgare, Fouilles et Recherches Acta Arch Acad Sci Hung Acta Archaeologica Academiae Scientiarum Hungaricae Acta Mus Napoc Acta Musei Napocensis (Cluj) Allg Geschfrd Allgäuer Geschichtsfreund Ann Bret Annales de Bretagne Ann Bruxelles Annales de la Société d’Archéologie de Bruxelles Ann Litt Univ Besançon Annales Littéraires de l’Université de Besançon Ann Lux Arlon Annales de l’Institut Archéologique de Luxembourg (Arlon) Ann Namur Annales de la Société Archéologique de Namur Ann Norm Annales de Normandie Ant J Antiquaries Journal Anz Schw Altkde Anzeiger für Schweizerische Altertumskunde Aquitania Aquitania: Revue InterRégionale d’Archéologie Arch Aeliana Archaeologia Aeliana Arch Anz Archäologische nzeiger Arch Aquit Archéologie en quitaine Arch Ausgr Bad-Württ Archäologische

Ausgrabungen in BadenWürttemberg Arch Bret Archéologie en Bretagne Arch Cambrensis Archaeologia Cambrensis Arch Cant Archaeologia Cantiana Arch Dtld Archäologie in Deutschland Arch J Archaeological Journal Arch Jahr Bayern Das Archäologische Jahr in Bayern Arch Korrbl Archäologisches Korrespondenzblatt Arch Limburg Archéologie en Limburg Arch Nachr Baden Archäologische Nachrichten aus Baden Arch Schw Archäologie der Schweiz Arch Wallonne L’Archéologie Wallonne Archaeologia Archaeologia or Miscellaneous Tracts relating to Antiquity Archéologie Archéologie: Chronique Semestrielle Archiv Fft Archiv für Frankfurter Geschichte und Kunst Archiv Hess Archiv für Hessische Geschichte und Altertumskunde Ardenne Ardenne et Famenne, Art, Archéologie, Histoire, Folklore Argovia Jahresschrift der Historischen Gesellschaft des Kantons Aargau


— Abbreviations of periodicals — Bad Fdb Badische Fundberichte Baselb Hmtb Baselbieter Heimatbuch Basler Zft Basler Zeitschrift für Geschichte und Altertumskunde Bayer Vgbl Bayerische Vorgeschichtsblätter Ber RGK Bericht der RömischGermanischen Kommission Ber ROB Berichten van de Rijksdienst voor het Oudheidkundige Bodemonderzoek Ber Saar Bericht der Staatlichen Denkmalpflege im Saarland Berks Arch J Berkshire Archaeological Journal BJb Banner Jahrbücher Bl Heimatkde Blätter für Heimatkunde Bl Schwäb Albver Blätter des Schwäbischen Albvereins Britannia Britannia: A Journal of Romano-British and Kindred Studies Brycheiniog Brycheiniog: Journal of the Brecknock Society BSDSaarld Bericht der staatlichen Denkmalpflege im Saarland Bud RégBudapest Régiségei Bull Ant Calais Bulletin de la Commission des Antiquités Départementales du Pas-de-Calais Bull Arch Bret Bulletin Archéologique de l’Association Bretonne Bull Arch Com Trav Hist Bulletin Archéologique du Comité des Travaux Historiques et Scientifiques Bull Art et Arch Bulletin des Commissions Royales d’Art et d’Archéologie Bull Calais Bulletin de la Commission Départementale d’Histoire et d’Archéologie du Pas-de-Calais Bull Comm Hist Nord Bulletin de la Commission Historique du Département du Nord Bull Comm Royales Art Bulletin des

Commissions Royales d’Art et d’Archéologie Bull Hesbaye-Condroz Bulletin du Cercle Archéologique HesbayeCondroz Bull Inst Arch Bulgare Bulletin de l’Institut Archéologique Bulgare Bull Inst Arch (London) Bulletin of the Institute of Archaeology Bull Liégeois Bulletin de l’Institut Archéologique Liégeois Bull Loire-Atl Bulletin de la Société Archéologique et Historique de Nantes et de Loire-Atlantique Bull Luxembourg Bulletin Trimestriel de l’Institut Archéologique de Luxembourg Bull Luxemb Arlon Bulletin de l’Institut Archéologique du Luxembourg Arlon Bull Mon Bulletin Monumental Bull Morbihan Bulletin de la Société Polymathique du Morbihan Bull Picardie Bulletin de la Société Historique de Picardie Bull Rouennais Bulletin de la Société des Amis des Monuments Rouennais Bull Soc Ant Picardie Bulletin de la Société des Antiquaires de Picardie Bull Soc Normande Etudes Préhist Bulletin de la Société Normande d’Etudes Préhistoriques Bull Trim Picardie Bulletin Trimestriel de la Société des Antiquaires de Picardie Burgenl Forsch Burgenländische Forschungen Burgenl Heimatbl Burgenländische Heimatblätter Cahiers Alsaciens Cahiers Alsaciens d’Archéologie, d’Art et d’Histoire Cahiers PicardieCahiers Archéologiques de Picardie Carinthia I Carinthia I: Zeitschrift für Geschichtliche Landeskunde von Kärnten


— Abbreviations of periodicals — Carnuntum-Jahrb CarnuntumJahrbuch Cheshire Arch Bull Cheshire Archaeological Bulletin Congr Arch Congrès Archéologique Current Arch Current Archaeology Dacia Dacia: Revue d’Archéologie et d’Histoire Ancienne Dmpfl Die Denkmalpflege Dmpfl Bad-Württ Denkmalpflege in Baden-Württemberg Derbyshire Arch J Derbyshire Archaeological Journal DNHAS Proceedings of the Dorset Natural History and Field Club Doc Charleroi Documents et Rapports de la Société Paléontologique et Archéologique de Charleroi Dossiers Alet Dossiers du Centre Régional Archéologique d’Alet Dossiers Arch Les Dossiers de l’Archéologie FdbaB-W Fundberichte aus BadenWürttemberg FdbaHessen Fundberichte aus Hessen FdbaSchw Fundberichte aus Schwaben FdbÖ Fundberichte aus Österreich Forsch u Fortschr Forschungen und Fortschritte Gallia Gallia: Fouilles et Monuments Archéologiques en France Gallia-Inf Gallia-Informations Germania Germania: Anzeiger der RGK Glasnik Bosna Herc Glasnik Zemolskoj Muzeja u Bosni i Hercegovini Glasnik Sarajevo Glasnik Zemaljskog Muzeja u Sarajevu Gymnasium Gymnasium: Zeitschrift für Kultur der Antike Helinium Helinium: Revue Consacrée à l’Archéologie des Pays-Bas, de la Belgique et du Grand-Duché de Luxembourg

Helv Arch Helvetia Archaeologica Hémecht Hémecht: Zeitschrift für Luxemburger Geschichte Herts Arch Hertfordshire Archaeology Herts Arch Rev Hertfordshire Archaeological Review Hist et Arch Histoire et Archéologie Hist Coll Staff Historical Collections for Staffordshire Historia (Stuttgart) Historia. Zeitschrift für Alte Geschichte Int Archiv Ethnog Internationales Archiv für Ethnographie Izv Muz Severozapadna Balgarija Izvestija na Muzeite Severozapadna Balgarija Jahresber Bayer Bodendenkmapfl Jahresbericht der Bayerischen Bodendenkmalpflege Jahresber Dillingen Jahresbericht des Historischen Vereins Dillingen a.d. D. Jahresber Ges Nützl Forsch Trier Jahresbericht der Gesellschaft für Nützliche Forschungen zu Trier Jahresber Hessen Jahresbericht der Denkmalpflege im Grossherzogtum Hessen Jahresber Oberösterr Musver Jahresbericht des Oberösterreichisches Musealvereines Jahresber Schwaben Jahresbericht des Historischen Vereins von Schwaben Jahresber Seetal Jahresbericht der Historischen Vereinigung Seetal Jahresber Stein Jahresbericht der Forderverein zur Erforschung und Erhaltung der Kulturdenkmale im Stein Jahresber Straubing Jahresbericht des Historischen Vereins für Straubing Jb Bern Jahrbuch des Bernischen Historischen Museums Jb DAI Jahrbuch des Deutschen Archäologischen Instituts Jb Liechtenstein Jahrbuch des Historischen Vereins für das Fürstentum Liechtenstein


— Abbreviations of periodicals — Jb Lothring Jahrbuch der Gesellschaft für Lothringische Geschichte und Kunst JbSGU Jahrbuch der Schweizerischen Gesellshaft für Ur- und Frühgeschichte Jb Solothurn Jahrbuch für Solothurnische Geschichte J Brit Arch Ass Journal of the British Archaeological Association J Northampton Mus Journal of the Northampton Museum and Art Gallery JOAI(B) Jahreshefte des Österreichischen Archäologisches Institutes (Beiblatt) J Roman Arch Journal of Roman Archaeology J Roman Stud Journal of Roman Studies Kent Arch Rev Kent Archaeological Review Kölner Jahrb Kölner Jahrbuch für Vorund Frühgeschichte Korrbl WdZ Korrespondenzblatt der Westdeutsche Zeitschrift für Geschichte und Kunst Kurtr Jb Kurtrierischen Jahrbuch Latomus Latomus: Revue d’Etudes Latines Lincs Hist Arch Lincolnshire History and Archaeology Lincs Rep Lincolnshire Architectural and Archaeological Society Reports and Papers Mém Côte-d’Or Mémoires de la Commission des Antiquités du Département du Côte-d’Or Mém Normandie Mémoires de la Société des Antiquaires de Normandie Mém Prés Divers Savants Acad Inscrip Mémoires Présentées par Divers Savants à l’Académie des Inscriptions et Belles-Lettres de l’Institut de France

Mém Soc Ant Fr Mémoires de la Société des Antiquaires de France Mitt Bad Homburg Mitteilungen des Vereins für Geschichte und Landeskunde zu Bad Homburg v.d. H. Mitt Central-Comm Hist Denkmale Mitteilungen der Königliche und Kaiserliche Central-Commission zur Erforschung der Kunst- und Historischen Denkmale Mitt DAI Röm Mitteilungen des Deutschen Archäologischen Instituts, Römische Abteilung Mitt Heddernheim Mitteilungen über Römische Funde in Heddernheim Mitt Hohenzollern Mitteilungen des Vereins für Geschichte und Altertumskunde in Hohenzollern Mitt Pfalz Mitteilungen des Historischen Vereins der Pfalz Mitt Salzburg Mitteilungen der Gesellschaft für Salzburger Landeskunde Mitt Zürich Mitteilungen der Antiquarischen Gesellschaft in Zürich MZ Mainzer Zeitschrift Nassau Ann Nassauische Annalen Norfolk Arch Norfolk Archaeology Northants Arch Northamptonshire Archaeology North Staffs J North Staffordshire Journal of Field Studies Not Scavi Notizie degli Scavi di Antichità Oberösterr Heimatbl Oberösterreichische Heimatblätter Oudheidk Med Oudheidkundige Mededelingen uit het Rijksmuseum van Oudheden te Leiden Oxf J Arch Oxford Journal of Archaeology Oxon Oxoniensia Peuce Peuce: Studii si Comunicari de Istorie si Arheologie Pfalz Mus Pfalzisches Museum


— Abbreviations of periodicals — Pro Alesia Pro Alesia. Revue des Fouilles d’Alise Proc Cambridge Proceedings of the Cambridge Antiquarian Society Proc Cambridgeshire Proceedings of the Cambridgeshire Archaeological Society Proc Devon Proceedings of the Devon Archaeological Exploration Society Proc Dorset Proceedings of the Dorset Natural History and Archaeological Society Proc Hants Proceedings of the Hampshire Field Club Proc Leatherhead Proceedings of the Leatherhead and District Local History Society Proc Soc Ant Lond Proceedings of the Society of Antiquaries of London Proc Somerset Proceedings of the Somerset Archaeological and Natural History Society Proc Univ Bristol Spelaeological Society Proceedings of the University of Bristol Spelaeological Society Proc Prehist Soc Proceedings of the Prehistoric Society Publ Limbourg Publications de la Société Historique et Archéologique dans le Limbourg Quartalbl Hist Ver Hessen Quartalblätter des Historischen Vereins für das Grossherzogtum Hessen Records Bucks Records of Buckinghamshire Reports Lincs Reports of the Lincolnshire Architectural and Archaeological Society Reports Northants Reports of the Northamptonshire Architectural and Archaeological Society Rev Arch Revue Archéologique Rev Arch Narbonnaise Revue Archéologique de Narbonnaise Rev Arch Picardie Revue

Archéologique de Picardie Rev Est Revue Archéologique de l’Est et du Centre-Est Rev Nord Revue du Nord Röm-Germ Korrbl RömischGermanisches Korrespondenzblatt Saalb-JbSaalburg-Jahrbuch Sammelbl Ingolstadt Sammelblatt des Historischen Vereins Ingolstadt SargetiaSargetia: Acta Musei Devensis Schaff Beitr Schaffhauser Beiträge zur Geschichte Schr Baar Schriften des Vereins für Geschichte und Naturgeschichte der Baar Schr Württ Alt-Ver Schriften des Württembergischen AltertumsVereins Schwäb Mus Schwäbische Museum Soc Préhist Nord Société Préhistorique du Nord South Midl Arch Newsletter of the CBA South Midlands Group Surrey Arch Coll Surrey Archaeological Collections Sussex Arch Coll Sussex Archaeological Collections Trans Birmingham Transactions of the Birmingham Archaeological Society Trans Bristol Transactions of the Bristol and Gloucestershire Archaeological Society Trans London and Middx Transactions of the London and Middlesex Archaeological Society Trans Thoroton Transactions of the Thoroton Society Trans Woolhope Transactions of the Woolhope Naturalists’ Field Club Tr Grab Forsch Trierer Grabungen und Forschungen Tr Jahresber Trierer Jahresberichte TrZft Trierer Zeitschrift für Geschichte und Kunst des Trierer Landes und seiner Nachbargebiete Verhand Hist Ver Oberpfalz


— Abbreviations of periodicals — Verhandlungen des Historischen Vereins für Oberpfalz und Regensburg Vom Rhein Vom Rhein: Monatsblatt des Wormser Altertumsvereins Wavriensia, Wavriensia: Bulletin du Cercle Historique et Archéologique de Wavre et de la Region Westd Zft Westdeutsche Zeitschrift für Geschichte und Kunst Wilts Arch Wiltshire Archaeological and Natural History Magazine

Wiss Arb Burgenld Wissenschaftliche WMBH Wissenschaftliche Mitteilungen aus Bosnien und der Herzegovina World Arch World Archaeology Württ Vierteljahresh Württembergische Vierteljahreshefte für Landesgeschichte Yorks Arch J Yorkshire Archaeological Journal Zft Hist Ver Schwaben Zeitschrift des Historischen Vereins für Schwaben



All books and articles other than excavation reports on single sites

ABBREVIATIONS OF BOOKS AND SERIALS Actes Balkaniques 1969 = Actes du Premier Congrès International des Etudes Balkaniques et Sud-Est Européennes, vol. 2, Sofia AFdS = Archäologische Führer der Schweiz ANRW = Aufstieg und Niedergang der Römischen Welt Arch Belg = Archaeologia Belgica (reprints) Ausgr Deutschland = RGK, Ausgrabungen in Deutschland 1950–75, Mainz 1975 Ausgr Rhld = Ausgrabungen in Rheinland 1981–2, Bonn BAR = British Archaeological Reports, Oxford: Brit Ser = British Series Int Ser = International Series Suppl Ser = Supplementary Series CAG = Carte archéologique de la Gaule 01 Ain 03 Allier 14 Calvados 16 Charente 18 Cher 19 Corrèze 23 Creuse 27 Eure 28 Eure-et-Loir 29 Finistère 35 Ille-et-Vilaine 36 Indre 37 Indre-et-Loire 38/1 Isère 40 Landes 41 Loir-et-Cher 44 Loire-Atlantique


— Bibliography — 45 Loiret 46 Lot 48 Lozère 49 Maine-et-Loire 50 Manche 53 Mayenne 62 Pas-de-Calais 64 Pyrénées-Atlantiques 71 Saône-et-Loire 87 Vienne Camb Ag Hist = Cambridge Agrarian History CBA Res Rept = Council for British Archaeology Research Report Déchelette = J. Déchelette, Manuel d’archéologie préhistorique, celtique et gallo-romaine, 8 vols EAA = East Anglian Archaeology Forsch u Ber Bad-Württ = Forschungen und Berichte zur Vor- und Frühgeschichte in BadenWürttemberg, Stuttgart Kent Mon = Kent Monograph Series, Dover NAR = Nederlandse Archaeolgische Rapporten Neue Ausgr Dtld = RGK, Nette Ausgrabungen in Deutschland, Frankfurt, 1958 ORL Obergermanische-Rätische Limes Paret = Paret, O. 1932, Die Siedlungen des Römischen Württembergs, Stuttgart = Goessler, P., F. Hertlein and O. Paret, Die Römer in Württemberg, vol. 3 RCHME Royal Commission on the Historical Monuments of England RGF Römisch-Germanische Forschungen RiBad-Württ Filtzinger, P. 1986 Die Römer in Baden-Württemberg, Stuttgart RiHessen Baatz, D. and F.-R. Herrmann (eds) 1982 Die Römer in Hessen, Stuttgart RiNordrh-Westf Horn, H.G. 1987 Die Römer in Nordrhein-Westfalen, Stuttgart RiRhld-Pf Cüppers, H. 1990 Die Römer in Rheinland-Pfalz, Stuttgart RiSchw Drack, W. and R. Fellmann 1988 Die Römer in der Schweiz, Stuttgart RLiÖ = Der Römische Limes in Österreich, Vienna ROB = Rijksdienst voor het Oudheidkundige Bodemonderzoek, Overdrukken SKF = Schweizerische Kunstführer Soc Ant Res Rept = Society of Antiquaries of London Research Report TIR = Tabula Imperii Romani: Sheets L32, L33, L34, M31, M32, M33; not numbered – CondateGlevum-Londinium-Lutetia; Britannia Septrionalis Tr Grab Forsch = Trierer Grabungen und Forschungen VCH = Victoria History of the Counties of England Adam, J.-P. 1984 La Construction romaine, Paris. Agache, R. 1970 Détection aérienne, Amiens. —— 1978 La Somme pré-romaine et romaine (= Mémoires de la Société des Antiquaries de Picardie, vol. 24, Amiens). Agache, R. and B. Bréart 1975 Atlas d’archéologie aérienne de la Picardie, vol. 1: Le bassin de la Somme à l’époque protohistorique et romaine, Amiens. Alfoldy, G. 1974 Noricum (transl. A. Birley), London. Allen, T.G. 1993 Excavations at Roughground Farm, Lechlade, Gloucestershire, Oxford. Anthes, G. 1906 ‘Römische Landhäuser in Deutschland’, Dmpfl 8, 117–22. Applebaum, S. 1972 ‘Roman Britain’, Camb Ag Hist I/ii, 3–277. Aucher, M.-R. and M. Aucher 1984 Villa-gallo-romaine des Cassons, Poitiers.


— Bibliography — Aydelotte, W.O., A.C. Bogue and R.W. Fogel 1972 The Dimensions of Quantitative Research in History, London. Barb, A. 1937 ‘Das angebliche Kastell Ulmus’, RLiÖ 18, 153–8. Barker, W.R. 1901 Account of Remains of a Roman Villa discovered at Brislington, Bristol, December 1899, Bristol. Baumann, V.H. 1983 Ferma Romana din Dobrogea, Tulcea. Baxter, R. 1931 Autobiography (Everyman edn.). de la Bedoyère, G. 1991 The Buildings of Roman Britain, London. —— 1993 English Heritage Book of Roman Villas and the Countryside, London. de Beer, E.S. (ed.) 1955, The Diary of John Evelyn, 6 vols, Oxford. Binding, G. 1991 Das Dachwerk, Munich. Binsfeld, W. 1979 ‘Zu treverischen Kultdenkmälern’, Festchrift 100 Jahre Rheinisches Landesmuseum Trier, Trier, 263–9(= Tr Grab Forsch 14). Biro, M. 1974 ‘Roman Villas in Pannonia’, Acta Arch Acad Sci Hung 26, 1–57. Bittel, K., W. Kimmig and S. Schiek 1981 Die Kelten in Baden-Württemberg, Stuttgart. Black, E.W. 1985 ‘Hypocaust Heating in Domestic Rooms in Roman Britain’, Oxf J Arch 4, 77–92. —— 1987 The Roman Villas of South-East England (= BAR Brit Ser 171). —— 1994 ‘Villa-Owners: Romano-British Gentlemen and Officers’, Britannia 25, 99–110. Blockley, K. 1985 Marshfield-Ironmongers Piece Excavations 1982–3 (= BAR Brit Ser 141). Bloemers, J.H.F. 1978 Rijswijk (Z.H.) ‘de Bult, een nederzetting van de Cananefaten’ (= Nederlandse Oudheden 8). Bodewig, R. 1906 ‘Römische Gehöfte zwischen Limes und Rhein’, Nassau Ann 36, 133–57. Braat, W.C. 1934 ‘Nieuwe Opgravingen van Romeinsche Villae’, Oudheidk Med NS 15, 4–37. —— 1941 ‘Nieuwe Opgravingen van Romeinsche Villa’s in Limburg’, Oudheidk Med NS 22, 39–51. Bradley, R. 1984 The Social Foundations of Prehistoric Britain, London. Branigan, K. 1971 Latimer: a Belgic, Roman, Dark Age and Early Modern Farm, n.p. —— 1977 The Roman Villa in South-West England, Bradford-on-Avon. Branigan, K. and P.J. Fowler 1976 The Roman West Country, Newton Abbot. Brodribb, A.C.C., A.R. Hands and D.R. Walker 1968–73 Excavations at Shakenoak,3 vols, privately printed. Bulliot, J.G. 1899 Fouilles de Mont Beuvray,2 vols, Autun. Carandini, A. and S. Settis 1979 Schiavi e padrone nell’ Etruria romana, Bari. Castro, M.C.F. 1982 Villas romanas en Espana, Madrid. de Caumont, A. 1838 Cours d’antiquités monumentales, Paris; 1st edn 1830–1. —— 1870 Abécédaire d’archéologie française; ère gallo-romaine, Paris; 2nd edn, Caen. Centre de Recherches sur les Monuments Historiques, Charpentes,7 vols, Paris, n.d. Choisy, A. 1873 L’art de bâtir chez les Romains, Paris; repr. 1984, Bologna. Christlein, R. and O. Braasch 1982 Das Unterirdischen Bayern, Munich. Clarke, G. 1982 ‘The Roman Villa at Woodchester’, Britannia 13, 197–228. Clarke, S. 1990 ‘The Social Significance of Villa Architecture in Celtic North-West Europe’, Oxf J Arch 9, 337–53. Clemen, P. (ed.) 1900 Die Kunstdenkmäler des Kreises Euskirchen, Düsseldorf. Collingwood, R.G. 1930 The Archaeology of Roman Britain, London. Collingwood, R.G. and I.A. Richmond 1969 The Archaeology of Roman Britain, London. Collis, J.R. 1986 ‘Adieu Hallstatt, Adieu La Tène’, Aquitania, Supplement I, 327–30. Colvin, H. and J. Newman (eds) 1981 Of Building: Roger North’s Writings on Architecture (Oxford). Corder, P. (ed.) 1954 The Roman Town and Villa at Great Casterton, Rutland, Nottingham. Corder, P. and J.L. Kirk 1932 A Roman Villa at Langton, near Malton, East Yorkshire, Leeds.


— Bibliography — Cotton, M.A. and G.P. Métraux 1985 The San Rocco Villa at Francolise, Rome. Coutil, L. 1898–1921 Département de l’Eure: archéologie gauloise, gallo-romaine, franque et carolingienne, 4 vols . Cunliffe, B. 1971a Excavations at Fishbourne 1969, 2 vols (= Soc Ant Res Rept 26). —— 1971b Fishbourne: a Roman Palace and its Garden, London. —— 1991a Danebury Excavations 1979–88, vol. 4: The Site (= CBA Res Rept 73). —— 1991b ‘Fishbourne revisited; the site in its context’, in J Roman Arch 4, 160–9. Cunliffe B., and D. Miles (eds) 1984 Aspects of the Iron Age in Central Southern Britain, Oxford. Cunliffe B., and T. Rowley, 1978 (eds) Lowland Iron Age Communities in Europe (= BAR Int Ser 48). Cüppers, H. 1979 Otrang Roman Villa, Mainz. Czysz, W. 1974 Der römische Gutshof München-Denning und die römerzeitliche Besiedlung der Münchner Schottenerbe, Munich. Darmon, J.P. 1981, ‘Les mosaïques en Occident I’, ANRW II/12.2, 266–319. Dedet, B. 1987 Habitat et vie quotidienne en Languedoc au milieu de l’Age de Fer (= Rev Arch Narbonnaise, Suppl. 7). Detsicas, A. 1983 The Cantiaci, Gloucester. Down, A. 1979 Chichester Excavations, vol. 4, Chichester. Drack, W. 1975 ‘Die Gutshöfe’, in Ur- und frühgeschichtlichen Archäologie der Schweiz, vol. 5, 49–72, Basle. —— 1976, Die römische Kryptoportikus von Buchs ZH und ihre Wandmalerei (= AfDS 7). —— 1990 Der römische Gutshöfe bei Seeb, Gem. Winkel, Zurich. Dremsizova-Nelcinova, C. 1969 ‘La villa romaine en Bulgarie’, Actes Balkaniques 1969, 503–25. Drury, P.J. (ed.) 1982a Structural Reconstruction (= BAR Brit Ser 110). —— 1982b ‘Form, Function, and the Interpretation of the Excavated Plans of Some Large Secular Romano-British Buildings’ in Drury 1982a, 289–308. Duncan-Jones, R.P. 1974 The Economy of the Roman Empire, Cambridge. Ebnöther, C. 1991 ‘Die Gartenanlage in der pars urbana des Gutshofes von Dietikon ZH’, Arch Schw 14, 250–6. Eiden, H. 1982 10 Jahre Ausgrabungen an Mittelrhein und Mosel 1963–76, Koblenz (= TrZft, Supplement 6). van Es, W.A. 1972 De Romeinen in Nederland, Bussum; 2nd edn 1983, Haarlem. Fallue, M. 1836 ‘Sur les antiquités de la forêt et de la presqu’Ile de Bretonne, et sur la villa de Maulévrier, près Caudebec’, Mém Normandie 10, 369–94. Fokkens, H. and N. Roymans (eds) 1991 Nederzettingen nit de Bronstijd en de Vroege Ijzertijd in de Lage Landen (= NAR 13), Amersfoort. Fouet, G. 1969 La villa gallo-romaine de Montmaurin, Haute-Garonne (= Gallia, Supplement 20). Fox, Sir C. and Lord Raglan 1951–4 Monmouthshire Houses, 3 vols, Cardiff. Fremersdorf, F. 1933 Der römische Gutshof Köln-Müngersdorf (= RGF 6). Frere, S.S. (ed.) 1960 Problems of the Iron Age in Southern Britain, London. Friendship-Taylor, R.M. and D.E. Friendship-Taylor 1989 Iron Age and Roman Piddington, an interim report, Northampton. Gaitzsch, W. 1986 ‘Grundformen römischer Landsiedlungen im Westen der CCAA’, BJb 186, 397–427. Gaubatz-Sattler, A. 1994 Die Villa rustica von Bondorf (= Forsch u Ber Bad-Württ 51). Gerster, A. 1973 ‘Der römische Gutshof in Seeb; Rekonstruktionsversuche’, Helv Arch 15, 62–81. —— 1983 Vicques: une grande villa gallo-romaine, Porrentruy.


— Bibliography — Girouard, M. 1978 The Making of the English Country House, New Haven and London. Gnirs, A. 1908 ‘Istrische Beispielen für Formen der antik-römischen villa rustica’, Jahrb Altkde 2, 124–43. Goodburn, R. 1972 The Roman Villa, Chedworth, London. Gorges, J.G. 1979 Les villas hispano-romaines, Paris. Grenier, A. 1906 Habitations gauloises et villas latines dans la cité des Médiomatrices, Paris. —— 1934 Manuel d’archéologie gallo-romaine, pt 2: L’archéologie du sol, Paris (= Déchelette VI/2). Grunewald, M. 1974 ‘Spätantike Herrschaftsvillen in den nordwestichen Provinzen des Römischen Reiches und ihr gesellschaftlichen Hintergrund’, Dissertation, University of Vienna . Gunther, R.T. 1928 The Architecture of Sir Robert Pratt, Oxford. Gunther, R. and H. Kopstein (eds) 1985, Die Römer an Rhein und Donau, 3rd edn, ViennaCologne-Graz. Gurney, D. (ed.) 1986 Settlement, Religion and Industry on the Roman Fen-edge, Norfolk (= EAA 31). Halbardier, B. and A. Thomas 1987 ‘La villa gallo-romaine de Chaumont à Rulles’, in Lambert 1987, 125–34. Harding, D.W. 1984 Holme House, Piercebridge: Excavations 1969–70 (= University of Edinburgh Department of Archaeology Project Paper no. 2). Harmand, J. 1951 ‘Sur la valeur archéologique du mot “villa”’, Rev Arch, 2, 155–8. —— 1961 ‘Les origines des recherches françaises sur l’habitat rural gallo-romain’, Latomus 51. Hartley, B. and L. Fitts 1988 The Brigantes, Stroud. Hartmann, M. 1975 Der römische Gütshof von Zofingen (= AFdS 6). Haverfield, F. 1990 ‘Romano-British Hampshire, pt 5: The Villas and Farm Buildings’, VCH Hampshire I, 293–318. —— 1906 ‘Romano-British Somerset, pt 4: Villas and Rural Dwellings’, VCH Somerset I, 298– 334. Hawkes, C.F.C. 1947 ‘Britons, Romans and Saxons . . . in Cranborne Chase, pt 4: Iwerne ’, Arch J 104, 48–62. —— (ed.) 1973 Greeks, Celts and Romans, London. Hawkes, C.F.C. and M.R. Hull 1947 Camulodunum, London. Hemp, W.J. and C. Gresham 1942–3 ‘Park, Llanfrothen, and the Unit System’, Arch Cambrensis 97, 98–112. Herrmann, F.-R. and A. Jockenhövel 1990 Die Vorgeschichte Hessens, Stuttgart. Hettner, F. 1878–81 ‘Drei römische Villen bei Leutersdorf, Mechern und Beckingen’, Jahresber Ges Nützl Forsch Trier 1878–81, 52–63. —— 1883 ‘Zur Kultur von Germanien und Gallia belgica’, Westd Zft 2, 1–26. Hingley, R. 1989 Rural Settlement in Roman Britain, London. Hingley, R. and D. Miles 1984 ‘Aspects of Iron Age Settlement in the Upper Thames Valley’, in Cunliffe and Miles 1984, 52–71. Hinz, H. 1969 ‘Die Ausgrabungen auf dem Kirchberg in Morken, 4, Die römische Siedlung’ (= Rheinische Ausgrabungen 7, 21–62), Düsseldorf. —— 1970 ‘Zur Bauweise der villa rustica’, Germania Romana, Heft 7/3, 15–25 = Gymnasium, Beihefte. Hoare, Sir R.C. 1821 Ancient Wiltshire, 2 vols. Hommages Millotte 1985 Elements de pré- et proto-histoire Européenne, Hommages à JacquesPierre Millotte (= Ann Lift Univ Besançon 299). Horn, W. and E. Born 1970 The Plan of St Gall, 3 vols, Berkeley. Jacquemet, M.H. 1857 ‘Etablissements gallo-romains du Morbihan’, Bull Man 23, 175–85. Jarrett, M. and S. Wrathmell 1981 Whitton: an Iron Age and Roman Farmstead in South Glamorgan, Cardiff.


— Bibliography — Jeannin, N. and G. Chouquer 1978 ‘Elements pour une géographic historique du Finage (Jura) à l’époque gallo-romaine’, Rev Est 29, 267–91. Johnston, D.E. 1970 ‘Villas . . . and Other Substantial Buildings’, Current Arch 21, July 1970, 280–6. —— 1978 ‘Villas of Hampshire and the Isle of Wight’, in Todd 1978, 71–92. Jones, G.R.J. 1972 ‘Post-Roman Wales’, Camb Ag Hist 1(2), 281–382. Jones, S. and J.T. Smith 1963–72 ‘The Houses of Breconshire’, 7 parts, Brycheiniog, vols 9–13. Joullin, L. 1901 Les établissements gallo-romaines de la plains de Martres-Tolosanes, Paris. Keller, F. 1864 ‘Statistik der römischen Ansiedelungen in der Ostschweiz’, Mitt Zürich 15, 63–331. Kellner, H.-J. 1971 Die Römer in Bayern, Munich. Kimmig, W. and E. Sangmeister 1983 Die Heuneburg am oheren Donau, 2nd edn, Stuttgart. Koepp, F. 1924 ‘Die villa von Odrang’, Germania 8, 6–13. Koethe, H. 1940 ‘Die Bäder römischer Villen im Trierer Bezirk’, 30 Ber RGK, 43–131. Kolling, A. 1977 ‘Zum Bau- und Siedlungswesen im südlichen Teil der Provinz Belgica’, 24 Ber Saar, 29–67. Kropatscheck, G. 1910–11 ‘Das romische Landhaus in Deutschland’, 6 Ber RGK, 51–78. Kubitschcek, W. 1926 Romerfunde von Eisenstadt (= Sonderschriften des Osterr Archaol Inst, Bd 11). Lambert, G. (ed.) 1987 Archéologie entre Semois et Chiers, Brussels. Latham, R. and Matthews, W. 1970–83 The Diary of Samuel Pepys, II vols, London. Leday, A. 1980 Rural Settlement in Central Gaul in the Roman Period, 2 parts (= BAR Int Ser 73). Leech, R. 1982 Excavations at Catsgore 1970–3, n.p. (= Western Archaeological Trust Excavation Monograph no. 2). Leyser, K. 1968 ‘The German Aristocracy from the Ninth to the Late Twelfth Century’, Past and Present no. 41, 25–53. Lucas, R.N. 1993 The Romano-British Villa at Halstock, Dorset: Excavations 1967–85 (= Dorset Natural History and Archaeological Society Monograph no. 13). Lugli, G. 1966 Villa d’Orazio, Rome. Luik, M. and F. Reutti 1989 Der Römerpark in Köngen, Stuttgart (= Führer zu den archäologischen Denkmälern in Baden-Württemberg 12). McKay, A.G. 1975 Houses, Villas and Palaces in the Roman World, London and New York. Mackreth, D. 1978 ‘Orton Hall Farm, Peterborough: a Roman and Saxon Settlement’, in Todd 1978, 209–28. de Maeyer, R. 1937 De Romeinsche Villa’s in Belgie, Antwerp. Marres, W. and J. J.F. W. van Agt 1962 De Nederlandse Monumenten van Geschiednis en Kunst, Deel 5, iii, Zuid-Limburg, The Hague, 15–20. Meates, G.W. 1979 The Roman Villa at Lullingstone, Kent, Maidstone. Medawar, P. 1967 The Art of the Soluble, London. Mercer, E. 1975 English Vernacular Houses, London. Metzler, J., J. Zimmer and L. Bakker 1981 Ausgrabungen in Echternach (Luxembourg), Luxembourg. Miclea, I. and R. Florescu, 1980 Stramosii Romanilor: Daco-Romanii, 2 vols, n.p. Miles, D. (ed.) 1982a The Romano-British Countryside, 2 vols (= BAR Brit Ser 103 i–ii). —— 1982b ‘Confusion in the Countryside: Some Comments from the Upper Thames Region’, in Miles 1982a, 53–78. —— 1986 Archaeology at Barton Court Farm, Abingdon, Oxon (= Oxford Archaeological Unit Report 3). Millar, F.G.B. 1977, The Emperor in the Roman World, London. Miln, J. 1877 Fouilles faites à Carnac, Paris. Modrijan, W. 1971 Der romische Landsitz von Löffelbach, 3rd edn, Graz.


— Bibliography — Mylius, H. 1924 ‘Die Rekonstruktion der römischen Villen von Nennig und Fliessem’, BJb 129, 109–28. —— 1933 ‘Zwei neue Formen römischer Gutshäuser (Villa bei Blankenheim)’, BJb 138, 11–21. Naeher, J. 1885 ‘Die römische Bauanlagen in der Zehntlanden badischen Anteiles’, BJb 79, 28– 104. Neal, D.S. 1974 The Excavation of the Romano-British Villa in Gadebridge Park, Hemel Hempstead (= Soc Ant Res Rept 31). —— 1974–6 ‘Three Roman Buildings in the Bulbourne Valley’, Herts Arch 4, 1–135. Neal, D.S., A. Wardle and J. Hunn 1990 Excavation of the Iron Age, Roman and Medieval Settlement at Gorhambury, St Albans (= English Heritage Archaeological Report 14). Netzer, E. 1988 ‘Architecture in Palestina prior to and during the Days of Herod the Great’, Akten des XIII. International Kongresses für Klassische Archäologie, Berlin. Nikolov, D. 1976 The Roman Villa at Chatalka, Bulgaria (= BAR Suppl Ser 17). Oelmann, F. 1921 ‘Die Villa rustica bei Stahl und Verwandtes’, Germania 5, 64–73. —— 1928 Hauptvortrag, ‘Römische Villen in Rheinland’, Arch Anz 43, cols 228–50. —— 1938 ‘Zur Aedicula von Mainz-Kastel’, in Festschrift für August Oxé, Darmstadt 183–91. van Ossel, P. 1992 Etablissements ruraux de l’antiquité tardive dans le nord de la Gaule, Paris. Ostendorf, F. 1908 Die Geschichte des Dachwerks, Leipzig; reprint, Hanover 1982. Palmer, N. 1982 Roman Stratford: Excavations at Tiddington 1980–1, Warwick. Partridge, C. 1981 Skeleton Green: a Late Iron Age and Romano-British Site, London. Percival, J. 1976 The Roman Villa, London. Philp, B. 1968 Excavations at Faversham, 1965 (= Kent Mon 1). —— 1984 Excavations in the Darent Valley (= Kent Mon 4). Philp, B. et al. 1991 The Roman Villa Site at Keston (= Kent Mon 6). Pollard, S. 1974 ‘A Late Iron Age Settlement and a Romano-British Villa at Holcombe, near Uplyme, Devon’, Proc Devon 32, 59–161. Potter, T.W. 1987 Roman Italy, London. Precht, G. 1973 Baugeshichtlichen Untersüchung zum römischen Praetorium in Köln, Cologne. Price, J.E. and F.C. Hilton Price 1883 A Guide to the Roman Villa . . . Brading, Ventnor. Price, J. and P.R. Wilson 1988 Recent Research in Roman Yorkshire (= BAR Brit Ser 193). Raepset-Charlier, M.-Th. and G. Raepset 1975 ‘Vingt-cinq ans de recherches historiques et archéologiques’, ANRW II Principal 4, 3–299. Rahtz, P.A. and E. Greenfield 1977 Excavations at Chew Valley Lake, London. Ramm, H.G. 1978 The Parisii, London. RCHME 1976 Gloucester I: Iron Age and Romano-British Monuments in the Gloucestershire Cotswolds, London. Reece, R. 1987 Coinage in Roman Britain London. Reece, R. 1993 ‘British sites and their Roman coins’, Antiquity, 863–9. Reuter, K. 1876 Zur Geschichte des römischen Wiesbadens, Bd 2, Römische Ansiedlung in der Umgebung von Wiesbaden (= Nassau Ann 5). RGK 1958 Neue Ausgrabungen in Deutschland, Frankfurt. Richmond, Sir I. 1969 ‘The Plans of Roman Villas in Britain’, in Rivet 1969, 49–70. Rippengal, R. 1993 ‘“Villas as a Key to Social Structure?” Some Comments on Recent Approaches to the Romano-British Villa and Some Suggestions toward an Alternative’, in Scott 1993a, 78–101. Rivet, A.L. 1969 The Roman Villa in Britain, London. Rodwell, W.J. 1978 ‘Buildings and Settlements in South-East Britain in the Late Iron Age’, in Cunliffe and Rowley 1978, 25–41.


— Bibliography — Rodwell, W.J. and K.A. Rodwell 1986 Rivenhall: Investigations of a Villa, Church and Village 1950–77 (= CBA Res Rept 55). Rostovtzeff, M.I. et al. (eds) 1952 The Excavations at Dura-Europos. Preliminary report of the ninth season of work, Part 3: The Palace of the Dux Ripae and the Dolicheneum, New Haven. Rothkegel, R. 1994 Die römische Gutshof von Laufenburg (= Forsch u Ber Bad-Württ 51). Roth-Rubi, K. 1986 Die Villa von Stutheien/Huttwilen TG, Basle. Roymans, N. and F. Theuws (eds) 1991 Images of the Past, Amsterdam. Sanden, W.A.B. van der 1987 ‘Oss-Ussen: de Nederzettingen’, in van der Sanden and van der Broeke 1987, 53–67. Sanden, W.A.B. van der and P.W. van der Broeke (eds) 1987 Getekend Zand’, Waalre (= Bijdrage tot de Studie van het Brabantse Heem, Deel 31). Saria, B. 1951 Der Römische Gutshof von Winden am See (= Burgenl Forsch 13). Schindler, R. 1977 Bundenbach-Altburg, Mainz (= Tr Grab Forsch 10). Schmidt, R.A. 1971 Die Villae rusticae ‘Oher der Pfingstweide’ in Ober-Ramstadt und ‘Am Zahl’, Kreis Darmstadt (= Ober-Ramstädter Hefte 2). Schubert, F. 1994 ‘Zur Mass und Entwurfslehre spätkeltische Holzbauten im Oppidum von Manching. Untersuchungen zu Grundrisstypen, Bauten und Baustrukturen’, Germania 72, 133–92. Schumacher, K. 1896 ‘Römische Meierhöfe im Limesgebiet’, Westd Zft 15, 1–17. —— 1923, Siedelungs- und Kulturgeschichte der Rheinlande, Bd II, Die römische Periode, ‘Die Meierhofe’, Mainz, 193–209. Scott, E. (ed.) 1993a Theoretical Roman Archaeology: First Conference Proceedings, Aldershot. —— 1993b A Gazetteer of Roman Villas in Britain (= Leicester Archaeology Monograph 1). Slofstra, J. 1991 ‘Changing Settlement Systems in the Meuse-Demer-Scheldt Area during the Early Roman Period’, in Roymans and Theuws 1991, 131–99. Smith, J.T. 1963 ‘Romano-British Aisled Houses’, Arch J 120, 1–30. —— 1970 ‘Lancashire and Cheshire Houses: Some Problems of Architectural and Social History’, Arch J 127, 156–81. —— 1978 ‘Villas as a Key to Social Structure’, in Todd 1978, 149–85. —— 1980 ‘The Roman Villa at Rapsley: an Interpretation’, Surrey Arch Coll 72, 63–8. —— 1982 ‘Structural Reconstruction’, in Drury 1982, 7–19. —— 1983 ‘Flight of Capital or Flight of Fancy?’, Oxf J Arch 2, 225–37. —— 1985 ‘Barnsley Park Villa: Its Interpretation and Implications’, Oxf J Arch 4, 341–51. —— 1987 ‘The Social Structure of a Roman Villa: Marshfield-Ironmongers Piece’, Oxf J Arch 6, 243–55. —— 1992 English Houses: the Hertfordshire Evidence, London. —— 1993 Hertfordshire Houses, London. Smith, P. 1974 ‘Historical Domestic Architecture in Dyfed’, in J. Barnes and N. Yates, Carmarthenshire Studies: Essays Presented to Major Francis Jones, London, 43–109. —— 1975 Houses of the Welsh Countryside, London. Spitaels, P. 1970 ‘La villa gallo-romaine d’Anthée’, Helinium 10, 209–33. Spitzing, T. 1988 Die römische Villa von Lauffen a.N. (= Forsch und Ber Bad-Württ 43). Sprater, F. 1929 Die Pfah unter den Römern, 2 Teile, Frankfurt. Staehelin, F. 1948 Die Schweiz in römischer Zeit, 3rd edn, Basle. Stanford, S.C. 1974 Croft Ambrey, privately printed. —— 1991 The Archaeology of the Welsh Marshe, privately printed. Stead, I.M. 1976 Excavations at Winterton Roman Villa 1958–67, London. —— 1980 Rudston Roman Villa, Leeds.


— Bibliography — Steiner, P. 1923 Römische Landhauser im Trierer Bezirk, Berlin. Steinhausen, J. 1932 Ortskunde Trier-Mettendorf, Bonn. Stevens, C.E. 1947 ‘A Possible Conflict of Laws in Roman Britain’, J Roman Stud 37, 132–4. —— 1966 ‘The Social and Economic Aspects of Rural Settlement’, in Thomas 1966, 108–28. Sumner, H. 1924 Excavations at East Grimstead, Wiltshire, London. Swoboda, K. 1918. Römische und romanische Paläste, Vienna; 3rd edn, Vienna-CologneGraz, 1969. Taylor, M.V. 1927 ‘Romano-British Kent, pt 4: Country Houses and Other Buildings’, VCH Kent III, 102–26. —— 1939 ‘Romano-British Remains, [Section] E: Country Houses and Farms’, VCH Oxfordshire I, 306–24. Ternes, C.-M. 1975 ‘Provincia Germania Superior im Bilde der jüngeren Forschung’, ANRW II, Principat 4, 722–1260. Thomas, B. 1964 Römische Villen in Pannonien, Budapest. Thomas, C. 1966 Rural Settlement in Roman Britain (= CBA Res Rept 7). Thomas, C.L. 1903 ‘Das römische Villengebaude bei der Guntersburg und die Bornburg’, Festchrift des Städtischen Historischen Museums in Frankfurt-am-Main, Frankfurt, 83– 101. Todd, M. 1978 Studies in the Romano-British Villa, Leicester. —— 1992 ‘Aisled Halls in Roman Gaul’, Oxf J Arch 11, 369–72. Trier, B. 1969 Das Haus im Nordwesten der Germania Libera, Munster. Uffler, A.M. 1981 ‘L’habitat rural en Gaule centrale’, Archéologie, November 1981, 71–5. Vasic, M. 1970 ‘Romische Villen vom Typus der Villa rustica auf jugoslawischen Boden’, Archaeologia Jugoslavia 11, 45–81. Verwers, G.J. 1972 Das Kamps Veld in Haps im Neolithikum, Bronzezeit und Eisenzeit, Leiden. Villes, A. 1985 ‘Que savons-nous des structures d’habitat des Ages du Bronze et du Fer en France septentrionale?’, in Hommages Millotte, 650–68. Wagner, E. 1908–11Fundstätte und Funde im Grossherzogtum Baden, 2 Teile, Tübingen. Walters, B. and B. Phillips 1978 Archaeological Excavations in Littlecote Park, Wiltshire, privately printed. Wanner, G. 1899 Die römischen Altertümer des Kantons Schaffhausen, Schaffhausen. Ward, J. 1911 Romano-British Buildings and Earthworks, London. Ward-Perkins, J. 1981 Roman Imperial Architecture, Harmondsworth. Webster, G. and L. Smith 1987 ‘Reply to J.T. Smith’s Suggested Reinterpretation of Barnsley Park Villa’, Oxf J Arch 6, 69–89. Welter, T. and H.E. Heppe 1908 ‘Die gallo-römischen Villen bei Loerchingen und Saaraltdorf’, Jahresber Ges Nützl Forsch Trier 20, 152–76. Widrig, W.M. 1987 ‘Land Use at the Via Gabina Villas’, in Ancient Roman Villa Gardens, Dumbarton Oaks, 225–60. Wightman, E. 1970 Roman Trier and the Treveri, London. —— 1985 Gallia Belgica, London. Wilkes, J.J. 1969 Dalmatia, London. Williams, R.J. and R.J. Zeepfat 1994 Bancroft: the Late Bronze Age and Iron Age Settlements . . . and the Roman Villa, 2 vols, Aylesbury. Wilson, D.R. 1974 ‘Romano-British Villas from the Air’, Britannia 5, 251–61. Wilson, D.R. and D. Sherlock 1980 North Leigh Roman Villa, London. Winbolt, S.E. 1925 Roman Folkestone, London. —— 1935 ‘Romano-British Sussex, III: Country Houses’, VCH Sussex III, 20–9. Woodward, S. (ed.) 1992 The Archaeology of Chichester and District 1991, Chichester.


— Bibliography — Wrathmell, S. and A. Nicholson, 1990 Dalton Parlours: Iron Age Settlement and Roman Villa (= Yorkshire Archaeology 3), Wakefield. Zahn, B. 1991 Die Basilika in Trier, Schriftenreihe des Rheinischen Landesmuseums Trier no. 6.



Only the principal sources used for the plans are listed. Achenheim Aguilafuente-Santa Lucia Ahrweiler Aiseau Albert Albesa-El Romeral Alcala de Guadaira, Casa de Pelay Cornea Allenz Almenara de Adaja-Villa Romana Alphen-Ekeren Alpnach-Dorf Altstetten Alresford Ambresin Anthée Apahida Apethorpe Arradon-Lodo Ash Ashtead Asthall-Worsham Athies Atworth

Hawkes 1973, 48–9 Gorges 1979, 355 see Bad-Neuenahr de Maeyer 1937, 63 Agache 1978, 287 Gorges 1979, 278–9 Gorges 1979, 358 see Mayen Gorges 1979, 437–8 Roymans and Theuws 1991, 137–45 Mitt Zürich 27, 1909–16, 227–57 JbSGU 57, 1972–3, 353–5 VCH Essex III, 37–8 Bull Comm Royales Art 15, 1876, 253–67 Ann Namur 14, 1877, 165ff.; 15, 1881–3, 1–40; Helinium 10, 1970, 209–41 Acta Mus Napoc 10, 1973, 130–3 RCHME Northamptonshire 1, North-East 8–10; AASR 5, 1859–60, 97–107 de Caumont 1870, 385–6, 388 VCH Kent III, 103–4; Kent Arch Rev 20, Summer 1970, 13–20 Surrey Arch Coll 37, 1926, 144–64; 38, 1930, 132–48 VCH Oxfordshire I, 319–21 Rev Nord 1967, 715–20; Agache 1978, 323 Wilts Arch 49, 1940–2, 46–95; Britannia 7, 1976, 362 340

— List of villas — Aulfingen Aylesford-Eccles Babworth-Dunston’s Clump Badajoz-Dehesa de la Cocosa Badbury Bachenau Bad Dürkheim-Ungstein Badgeworth Bad Godesberg Bad Homburg vor der Höhe Bad Kreuznach Bad Neuenahr-Ahrweiler Bad Rappenau-Zimmerhof Bancroft Bargen im Hegau Barnsley Park Barton Court Farm Basel-Gasfabrik Basse-Wavre Beadlam Beckingen Bedburg-Garsdorf Beegden Befort-Aalburg Bellersheim Belleuse-les Mureaux Bellikon Bennwil Bergen-Holzhausen Beringen-Lieblosental Bertenbreit Betting Betzingen Biberach Biberist-Spitalhof Bierbach Bierlingen Bignor Bihac-Zaloje

Schr Baar 8, 1893, 61–7 Arch Cant 78, 1963–88, 1973, annually; Detsicas 1983, 120–6 Current Arch 85, 1982, 43–8 Gorges 1979, 189–90 Wilts Arch [M5] Schumacher 1896, 13–17 RiRhld-Pf, 317–19; Mitt Pfalz 79, 1981, 40–1 RCHME 1976, 5–6 BJb 158–9, 1959, 380–2 RiHessen, 235; Mitt Bad Homburg 18, 1935, 16–30 RiRhld-Pf, 321–3 RiRhld-Pf, 324–5; van Ossel 1992, 227–8 FdbaBad-Württ 3, 1977, 474–84; RiBad-Württ, 215–16 see Milton Keynes Bad Fdb 1, 1925–8, 170–4 Trans Bristol 99, 1981, 21–78; 100, 1982, 65–190 see Radley JbSGU 58, 71–5 Ann Bruxelles 19, 1905, 303–29 Yorks Arch 43, 1971, 178–86 Jahresber Ges Nützl Forsch Trier, 1878–81, 59–63 BJb 159, 1959, 382–4 ROB no. 356 Germania 26, 1942, 26–34; Hémecht 21, 1969, 37–50 see Hungen Agache 1978, 328, Fig. 21 Zft Schw Arch 5, 1943, 86–122 JbSGU 32, 1940, 128–30; RiSchw, 361–2 Bayer Vgbl 49, 1984, 99–112 Anz Schw Altkde 19, 1886, 331–3; Wanner 1899, 31–3 RiSchw, 362–3 see Gunzenheim Grenier 1906, 79–86; 1934, 810 FdbaSchw 13, 1905, 63–70 RiBad-Württ, 244–5 JbSGU 69, 1986, 199–220 15 Ber Saar, 1968, 7–39 see Sturzach Britannia 13, 1982, 135–95; Black 1987, 153 and passim Wilkes 1969, 404 341

— List of villas — Bilsdorf Birkenfeld-Elchweiler Bistrica Blankenheim-Hulchrath Bledlow-cum-Saunderton Blieskastell-Altheim Bocholtz-Vlengendaal Böckweiler Bollendorf Bondorf Boos-Le Bois Flahaut Börstingen Bouchoir Box Boxted Bözen Bradley Hill Brain-sur-Allonnes Bramdean Braughing-Skeleton Green Bray-sur-Somme Bregenz Brewood-Engleton Brighstone-Rock Brioni Grande-Val Catena Bristol-Brislington Bristol-Kingsweston Brixworth Broichweiden Brombach Bruchsal-Ober-Grombach Brücken Bruckneudorf-Parndorf

Buchs Buchten Budakalász Budapest-Aquincum Budapest-Csúcshegy Budapest-Testvérhegy Bundenbach-Altburg

Ann Lux (Arlon) 45, 1910, 354–66 Westd Zft 6, 1887; Korrbl WdZ, no. 187 Dremsizova-Nelcinova 1969, 510 BJb 123, 1916, 136–7, 210–26; 1932, 281–2; Mylius 1933; RiNordrh-Westf, 360–3 Records Bucks 13, 1934–40, 398–426; 18, 1966–70, 261–76 24 Ber Saar, 1977, 29–67 Int Archiv Ethnog 24, 1918, 1–22; van Es 1972, 146–7 8 Ber Saar, 1961, 80–104 Tr Jahresber 12, 1923, 1–59; RiRhld-Pf, 338–40 Gaubatz-Sattler, 1994 Bull Rouennais 1907, 41–50 FdbaSchw NS 2, 1924, 28–9 Agache 1978, 287, 338 Fig. 31 Wilts Arch 33, 1904, 236–69; 81, 1987, 19–51 see Upchurch Anz Schw Altkde 27, 1925, 65–108 see Somerton CAG 49, Le-Maine-et-Loire, 77–8 Black 1987, 50, 226 Partridge 1981 Agache 1978, 287 Mitt Central-Comm Hist Denkmale NS 12, 1886, 72– 84 Hist Coll Staff 1938, 267–93 Britannia 7, 1976, 367–9 Ward-Perkins 1981, 196–7; Jahrb Altkde 2, 1908, 124–43 Barker 1901; Proc Somerset 116, 1972, 78–85 Trans Bristol 69, 1959, 5–58 J Northampton Mus 8, 1970, 3–102 BJb 177, 1977, 577–80 see Lörrach Röm-Germ Korrbl 5, 1912, 35–40 Sprater 1929, 122, 125 Burgenl Heimatbl 13, 1951, 49–65; 14, 1952, 97– 102; 41, 1979, 66–87; Wiss Arb Burgenld 35, 1966, 252–7; Thomas 1966, 177–92 Drack 1976 Oudheidk Med NS 9–11, 1928–30, 1 Thomas 1966, 214–15 Bud Rég 16, 1955, 393, 421–5; 18, 1958, 71–7 Thomas 1966, 216–24 Thomas 1966, 232–7 Schindler 1977 342

— List of villas — Burgweinting Byfield Cabo de Palos-El Castillet Cachy Canterbury Cardiff-Ely Carisbrooke Carnac Castle Dykes Cenero-Murias de Belono Cerbois-Les Coudraits Champagnole Champdivers La Chapelle-Vaupelteigne Chassemy Chastres Chastrès-lez-Walcourt Chatalka-Delimonyova Niva Chatalka-Lambata Châtillon-sur-Seiche Chedworth Cherington Chew Park Chichmanovtsi Chiddingfold Chilgrove Chiragan Cincis Citernes Ciumafaia Civray-Le Poirier Molet Clairy-Saulchoix Clanville Claydon Pike Cléry-sur-Somme Clinchamps Cobham Park Colchester-Camulodunum Colerne Collingham-Dalton Parlours Colsterworth Combe St NicholasWadeford

see Regensburg RCHME Northamptonshire III, 33–4 Gorges 1979, 309 Agache 1978, 339 Fig. 32 Cunliffe and Rowley 1978, 29 J Roman Stud 11, 1921, 67–85 VCH Hampshire I, 316–17 Miln 1877 see North Stainley Gorges 1979, 330 Leday 1980, Pl. XIX Gallia 44, 1986, 246 Jeannin and Chouquer 1978, 270 Rev Est 21, 1970, 261–330 World Arch 1, 1969, 106–35 Ann Namur 24, 1900, 27–32 Ann Namur 24, 1901, 121–8 Nikolov 1976, 26–35 Nikolov 1976, 40–7 Gallia-Inf 1990, 42–5; CAG Ille-et-Vilaine no. 347 RCHME Gloucestershire Cotswolds, 24–8 RCHME Gloucestershire, 29 see Stowey Sutton Nikolov 1976, 168 Surrey Arch Coll 75, 1984, 58–82 see West Dean see Martres-Tolosanes Acta Mus Napoc 2, 1965, 163–93 Agache 1978, 326 Acta Mus Napoc 10, 1973, 133–6 Leday 1980, Pl. XXI Agache 1978, 330 Fig. 23 see Weyhill see Lechlade Agache 1978, 336 Fig. 29 de Caumont 1830, Atlas (iii) Fig. 8 Arch Cant 76, 1961, 88–109 Hawkes and Hull 1947, 89–92; Cunliffe and Rowley 1978, 25–41 Arch J 13, 1856, 327–32 Wrathmell and Nicholson 1990 Frere 1960, 23–5 VCH Somerset I, 333


— List of villas — Comunion-Cabriana Compton Dando-Littleton Condé-Folie Cosa Courcelles-sur-NiedFrécourt Courcelles-sur-Nied-Urville Cox Green Crain Croft Ambrey Cromhall Crookhorn Csúcshegy Cuevas de Soria-Dehesa de Soria Cwmbrwyn Dalton Parlours Danebury Darenth Davenescourt Deanshanger Démuin Denton Deutschkreutz Deva Dicket Mead Dietfurt a.d. Altmühl Dietikon Dirlewang Ditchley Doische-Vodelée Downton Dracevica Dragonby Draughton Drax Dreieich-Götzenhain Druten Dura Europos – palace of Dux Ripae Dury East Dean-Holbury East Grimstead

Gorges 1979, 177 VCH Somerset I, 323–4 Agache and Bréart 1975, Fig. 41; Oxf J Arch 6, 1987 249 Potter 1987, 107 Jb Lothring 18, 1906, Table 15 Jb Lothring 18, 1906, 413; Grenier 1904, 807 see Maidenhead Rev Est 28, 1977, 117–32; Hist et Arch 58, 1981, 72 Stanford 1974 Proc Soc Ant Lond 2nd ser., 13, 1912, 20–3; Bristol and Avon Archaeology 6, 1987, 60–1 see Purbrook Thomas 1964, 216–24 Gorges 1979, 398–9 see Laugharne see Collingham Cunliffe 1991a; Cunliffe and Rowley 1978, 32 Arch Cant 22, 1897, 49–84; Philp 1991 Agache 1978, 287 Reports Northants 63, 1960–6, 22–6; Britannia 4, 1973, 294 Agache 1978, 287, 338 Fig. 31 Lincs Rep 10, 1964, 75–104 Thomas 1966, 128–3 Acta Mus Napoc 10, 1973, 136–40 see Welwyn Arch Jahr Bayern 1985, 75–7 Arch Schw 14, 1991, 250–6 Bayer Vgbl 11, 1933, 77–87 Oxon 1, 1936, 24–69 Arch Belg 3, 1987, 153–64 Wilts Arch 58, 1961–3, 303–41 WMBH 5, 1897, 163–7; Wilkes 1969, 397 see Roxby-cum-Risby Frere 1960, 21–3 Yorks Arch J 41, 1963–6, 620–86 RiHessen, 257–9 ROB no. 133 Rostovtzeff 1952 Agache 1978, 287, 338 Fig. 31 VCH Hampshire I, 1, 300 Sumner 1924 344

— List of villas — Eaton-by-Tarporley Eccles Eching-Autobahn Eching-Neufahrn Echternach Eckartsbrunn L’Ecluse-Leckbosch Ecoust-Saint-Mein Egregy Ehingen am Ries Eisenstadt Elchweiler Ellesborough-Terrick Elsted-Batten Hanger Ersigen Estrées-sur-Noye Etalle L’Etoile EuskirchenKreuzweingarten Evelette Ewhurst-Rapsley Exning-Landwade Ezinge Famechon Farmington-Clear Cupboard Farningham-Manor House Faversham Feddersen Wierde Feltwell Ferpicloz Finkley Fischamend-Katharinenhof Fishbourne-Palace Fishtoft Fitten Fliessem-Otrang Flocques Flumenthal Folkestone Fontaine-le-Sec Fontoy-Moderwiese Francolise-San Rocco

VCH Cheshire I, 210–11; Cheshire Arch Bull 9, 1983 67–73 see Aylesford Arch Jahr Bayern 1983, 65–7 Arch Jahr Bayern 1980, 84–5 Metzler, Zimmer and Bakker 1981 Röm-Germ Korrbl 4, 1911, 86–9 Wavriensia 29, 1980, 1–27 Agache 1978, 324 Thomas 1964, 33 Christlein and Braasch 1982, 61 Kubitscheck 1926; Thomas 1966, 137–51 see Birkenfeld Records Bucks 2, 1863, 53–6 Woodward 1992, 27–32 Jb Bern 45–6, 1965–6, 373–407 Agache 1978, 319 Arch Wallonne 2, 1994, 110–11 Agache 1978, 287 Kunstdenkmaler der Rheinprovinz 4/iv, 187–91; RiNordrh-Westf 426–7 Bull Hesbaye-Condroz 6, 1966, 15–17 Surrey Arch Coll 65, 1968, 1–70 Proc Cambridge 76, 1987, 41–66 Germania 20, 1936, 40–7 Cahiers Picardie 1981, 147–56; Bull Soc Ant Picardie 1976, 314–31 Trans Bristol 88, 1969, 34–67 Arch Cant 88, 1973, 1–21 Philp 1965 Germania 34, 1956, 125–41; 40, 1962, 280–317 East Anglian Arch 31, 1–48 RiSchw 394–5 see St Mary Bourne RLiO 5, 1904, cols 12–14; Thomas 1966, 244–5 Cunliffe 1971b; Sussex Arch Coll 124, 1986, 51–77 Lincs Hist Arch 12, 1977, 73 18 Ber Saar, 1971, 7–25 Germania 8, 1924; TrZft 4, 1928, 75–83; Cüppers 1979 Agache 1978, 287 Jb Solothurn 30, 1957, 228–33 Winbolt 1925; VCH Kent III, 114 Agache 1978, 287, 333 Fig. 26 Gallia-Inf 1989, ii, 107 Cotton and Métraux 1985 345

— List of villas — Frankfurt-Bergen-Enkheim Frankfurt-Bornheim Friedberg-Fladerlach Friedberg-Pfingstweide Friedrichdorf-Seulberg Frilford Frocester Court Fromentières Furschweiler Gadebridge Park Gambach Garden Hill Gargrave-Kirk Sink Gayton Thorpe Geislingen-Heidegger Hof Génimont Gentelles Gerpinnes Goodrich-Huntsham Gorhambury Gornea Graux Graz-Thalerhof Great Casterton Great Staughton Great Weldon Great Witcombe Grémecey Grenchen

Archiv Fft 3rd ser., 12, 1920, 313–26 Thomas 1903; RiHessen, 297–8 Arch Jahr Bayern 1990, 94–7; Zft Hist Ver Schwaben 72, 1978, 40ff. RiHessen, 307–9 Saalb-Jb 7, 1930, 92–109; RiHessen, 312 Arch J 54, 1897, 340–54 RCHME Gloucestershire Cotswolds, 56–8; Trans Bristol 89, 1970, 15–89 Gallia 39, 1981, 336–7 16 Ber Saar, 1969, 123–39 see Hemel Hempstead see Munzenberg see Hartfield Hartley and Fitts 1988, 75–8 Norfolk Arch 23, 1926–8, 166–209 Wagner 1908–11 (ii), 133–4 see Villers-sur-Lesse Agache 1978, 339 Doc Charleroi 7, 1875, 93–140 Trans Woolhope 37, 1962, 179–91; J Roman Stud 55, 1965, 208; 56, 1966, 205–6 see St Michael Miclea and Floresu 1980, 2, 187 Ann Namur 29, 1910, 137–44 Bl Heimatkunde 33, 1959, 9–19 Corder 1954 J Roman Stud 49, 1959, 118 Northants Arch 22, 1988–9, 23–67 RCHME Gloucestershire Cotswolds, 60–1 Gallia 24, 1966, 286–8 JbSGU 32, 1940, 133–4; Jb Solothurn 40, 1967, 444– 66 Agache 1978, 336 Oudheidk Med NS 15, 1934, 1–13 see Sachsenheim Gorges 1979, 358

Grivesnes Groesbeek-Plasmolen GrossSachsenheim Guadaira-Alcala/Casa de Pelay Correa La Guerche-Le Grand ChausseroiLeday 1980, Pl. XXII Guiry-Gadancourt Gallia 16, 1958, 266–80 Gunthersburg Park see Frankfurt-Bornheim Gunzenheim-StaatsforstSchwab Mus 1930, 159–60 Sulz Gyuláfiratót-Pogánytelek Thomas 1964, 34–49 Haamstede-Brabers van Es 1981, 168–9 Habay-Rulles see Rulles-Chaumont 346

— List of villas — Haccourt Hales Halstock Hambach Hamblain-les-Prés Hambleden-Yewden Manor Ham Hill Haps Harbonnières Harburg-GrossSorheim Harding-Mingies Ditch Hartfield-Garden Hill Hartlip Haunstetten Haut Clocher-St Ulrich Hechingen-Stein

Heerlen-Boventse Caumer Hemel HempsteadBoxmoor Hemel HempsteadGadebridge Park Heppenheim Hérouville-Lebisey Herschweiler-Pettersheim Herten-Warmbach Heuneburg High Wycombe Hirschberg-GrossSachsen Hobita Hoheneck Hohenfels-Im Keller Holbury Holcombe Holme House Hölstein Hombleux Hoogeloon Horath Hornchurch Hosszúhetény Houdeng-Goegnies Houthem-Vogelsang

Arch Belg no. 132, 1971 North Staffs J 9, 1969, 104–17 Lucas 1993 see Niederzier Rev Nord 66, 1984, 181–205 Archaeologia 71, 1921, 141–98 see Montacute Verwers 1872; Roymans and Theuws 1991, 137–57 Agache 1978, 287, 337 Fig. 30 Arch Jahr Bayern 1988, 105–10 Cunliffe and Miles 1984, 62 Britannia 8, 1977, 339–50 VCH Kent 3, 117–18 Arch Jahr Bayern 1990, 80–1 Jb Lothring 6, 1894, 313–17; 10, 1898, 171–94; Gallia 29, 1971, 17–44; 30, 1972, 41–82 RiBad-Württ, 306–9; Arch Ausgrab Bad-Württ 1980, 82–6; 1981, 137–40; 1992, 177–8; Dmpfl Bad-Württ 11, 1982, 171–3 Marres and van Agt 1962, 18 Neal 1974–6, 53–110 Neal 1974 RiHessen, 346–7 Grenier 1934, 800 Mitt Pfalz 80, 1982, 320–1 see Rheinfelden Kimmig and Sangmeister 1983 Records Bucks 16, 1953–60, 227–57 Arch Ausgr Bad-Württ 1986, 153–8 Acta Mus Napoc 10, 1973, 142–4 see Ludwigsburg TrZft 24–6, 1956–8, 543–7 see East Dean see Uplyme see Manfield Baselb Hmtb 5, 1950, 1–52 Agache 1978, 340 Fig. 33 Roymans and Theuws 1991, 137–52 TrZft 30, 1967, 114–43; RiRhld-Pf, 395–7 Cunliffe and Rowley 1978, 32 Thomas 1966, 272–8 de Maeyer 1937, 83–5 Oudheidk Med NS 6, 1925, 40–79 347

— List of villas — Hucclecote Hummetroth-Haselburg Hungen-Bellersheim Huntsham Hüssingen Hüttwilen Illogan-Magor Inzigkofen Irsch Iskar-Gara Isola d’Istria Izel Izernore Jemelle-Neufchâteau Jublains-la-Boissière Jumilla-Los Cipreses Kaalheide-Krichelberg Kadin Most Kaisersteinbruch-Königshof Katzenbach Kempten-Loja Kernen-Rommelshausen Keszthely-Fenékpuszta Kethel Keynsham-Cemetery Keynsham-Somerdale Kingsweston Kingweston-Littleton Kinheim Kirchberg Kirkel-Frost Kleinsteinhausen Koerich-Goeblingen Köln-Braunsfeld Köln-Mungersdorf Köln-Praetorium Köngen Königsbrunn Königshof Konken

Trans Bristol 55, 1933, 323–76; 79, 1961, 159–73; 80, 1962, 42–9 RiHessen, 360 Archiv Hess 11, 1865–7, 157–73 see Goodrich see Westheim see Stutheim J Brit Arch Ass NS 39, 1933, 117–75 FdbaBad-Württ 3, 1977, 402–42; RiBad-Württ, 347–8 see Vierherrenborn Bull Inst Arch Bulgare 11, 1937, 407–9; Dremsizova-Nelcinova 1969, 505 Not Scavi 6 ser. 6, 1928, 412–14 Arch Belg 223, 1980, 57–9 see Perignat Ann Namur 21, 1895, 403–49 CAG 53 Mayenne, 63–5 Gorges 1979, 311 van Es 1972, 151 Dremsizova-Delcinova 1969, 507 RLiO 6, 1925, cols 6–10, 34–50; 18, 1937, 154–8; Thomas 1966, 152–74 RiRhld-Pfalz 407–9 Allg Geschfrd 40, 1937, 51–71 FdbaBad-Württ 2, 1975, 193–204 Thomas 1966, 60–9 Ber ROB 23, 1973, 149–58 Archaeologia 75, 1926, 109–35 Archaeologia 75, 1926, 136–8; Bristol and Avon Archaeology 4, 1985, 6–12 see Bristol VCH Somerset I, 323–4 TrZft 40–1, 1977–8, 414–15; RiRhld-Pf, 414–15; Tr Grab Forsch 14, 1979, 263–9 see Küttingen 19 Ber Saar, 1972, 89–98 Mitt Pfalz 80, 348–51 Hémecht 25, 1973, 375–9 BJb 135, 1930, 109–45 Fremersdorf 1933 Precht 1973 FdbaSchw NS 19, 1971, 230–53; Luik and Reutti, 1989 Arch Jahr Bayern 1988, 78–9 see Kaisersteinbruch Mitt Pfalz 81, 1983, 74–7 348

— List of villas — Konska Konz-Lummelwies Konz-Pfalz Kösching Kralev Dol Kran Kulm Küttigen-Kirchberg Laboissière-en-Santerre Lalonquette Lamargelle-Versingue Landen-Betzveld Landshut-Sallmansberg Langenau-Osterstetten Langton Laperrière-sur-Saône Latimer-Chenies Laufen-Müschag Laufenburg Lauffen am Neckar Laugharne-Cwmbrwyn Lavans-les-Dole Lazenay-Les Sales Lechlade-Claydon Pike Leicester-Norfolk Street Leiwen-Bohnengarten Lemiers Lendin Leutersdorf-Maiweiler Leutersdorf Levet-Champ des Pois Levroux-Trégonce Liédena Liestal-Munzach Lisiici Littlecote Little Milton Littleton Ljubsko-Proboj Ljusina Llantwit Major

Bull Inst Arch Bulgare 14, 1940–2, 267–9; Dremsizova-Nelcinova 1969, 504 van Ossel 1992, 250 Germania 19, 1935, 40–53; 39, 1961, 204–6; RiRbld-Pf, 649–53 Christlein and Braasch 1982, 200 Acad Bulg Fouilles Rech 14, 1985, 9–26, 148–9 Arheologija (Sofia), 1977, 47–53 Keller 1964, 128–31 Anz Schw Altkde NS 10, 1908, 24–30 Agache and Bréart 1975, 81 Gallia 31, 1973, 123–55 Mem Côte d’Or 20, 1933–35, 99–106 Bull Liégeois 11, 1872, 117–21 Arch Jahr Bayern 1982, 26–32, esp. 31 Württ Vierteljahresh 11, 1888, 29–36 Corder and Kirk 1932 Chouquer 1979, 56 Fig. 4 Branigan 1971 Helv Arch 9, 1978–9, 2–66 Germania 24, 1940, 32–6; Gnomon 1939, 40–1; Rothkegel 1994 Spitzing 1988 Arch Cambrensis, 6 ser., 7, 1907, 175–209 Rev Est 19, 1968, 257 Leday 1980, Pl. IX Hingley and Miles 1984, 61 Britannia 12, 1981, 337–8 TrZft 24–6, 1956–8, 583–93 Oudheidk Med NS 15, 1934, 18–28 Mém Normandie 10, 1836, 387–91 TrZft 24–6, 1956–8, 546–9 Jahresber Ges Nutzl Forsch Trier, 1878–81, 52–8 Leday 1980, Pl. XXVI CAG Indre, no. 143 Gorges 1979, 343 Ur-Schweiz 17, 1953, 1–13; JbSGU 46, 1957, 113–15 RiSchw, 430–4 Wilkes 1969, 401 see Ramsbury J Roman Stud 40, 1950, 102; 43, 1953, 94 see Compton Dando WMBH 3, 1895, 280–3; Wilkes 1969, 397 Wilkes 1969, 404 Arch Cambrensis 102, 1952–3, 89–163; Britannia 5, 1974, 225 349

— List of villas — Lockington Lockleys Lodo Löffelbach Loja London-Cannon Street Lörrach-Bromach Ludwigsburg-Hoheneck Ludwigsburg-Pflugfelden Lullingstone Lussas-et-Nontronneau Madara Magor Farm Maidenhatch Maidenhead-Cox Green Maidstone-Loose Road Mailhac-Cayla Maillen-Al Sauvenière Maillen-Ronchinne Mainz-Kastel, ‘Aedicula’ Majdan Malguenac-Guilly Mali Mounj Mamer-Gaschtbierg Manching Manderscheid Manerau Manfield-Holme House Mansfield Woodhouse Marboué-Mienne Marchelepot Mareuil-Caubert Maria Ellend-Ellender Hof Maria Ellend-Ellender Weingärten Markt Pongau-Urreiting Marquivilliers Marshfield Martigues Martres-Tolosanes, Chiragan Matagne-la-Petite

Wilson 1974, Pl. 22 see Welwyn see Arradon Modrijan 1971 see Kempten Trans London and Middx 26, 1975, 1–102 Arch Ausgr Bad-Württ 1981, 160–2; RiBad-Württ, 428–9 FdbaSchw 19, 1911, 90–118; Arch Ausgr Bad-Württ 1992, 179–83 Arch Ausgr Bad-Württ 1988, 168–71 Meates 1972 Gallia 31, 1973, 463; 41, 1983, 445–6 Nikolov 1976, 169 see Illogan see Pangbourne Berks Arch 60, 1962, 62–91 VCH Kent III, 100 Dedet 1987, 186, Fig. 167 Ann Namur 19, 1891, 345–75 Ann Namur 19, 1891, 345–75; 21, 1895, 177–208 Oelmann 1938 Wilkes 1969, 404 Bull Morbihan 1899, 137–43 Wilkes 1969, 404 Hémecht 32, 1980, 465–79 Schubert 1994; Germania 40, 1962, Supplement 3, 4 BJb 39–40, 1866, 256–64 Acta Mus Napoc 10, 1973, 144–7 Harding 1984 Archaeologia 8, 1787, 363–76; Trans Thoroton 53, 1949, 1–14 Mém Soc Ant Fr NS 2, 2, 1836, 153–73; Gallia 39, 1981, 63–83 Agache 1978, 287, 335, Fig. 28 Agache 1978, 276 RLiÖ 4, 1903, cols 9–13; Thomas 1966, 248 RLiÖ 4, 1903, cols 13–14; Thomas 1966, 249 FdbÖ 3, 1941, 64–7 Agache 1978, 287 Blockley 1985 Dedet 1987, 186–7, Fig. 169 Joullin 1901 Arch Belg 253, 1983, 65–8 350

— List of villas — Maulévrier Mauren Mautern a. Donau Mayen-Allenz Mayen-Im Brasil Meckel Mehring Meonstoke-Shavards Farm Merdingen Merklingen Le Mesge Les Mesnuls Messkirch-Altstadt Mettet-Bauselenne Metz-Sablon Mezières-en-Santerre-La Croix-Saint-Jacques Mezières-en-Santerre-Le Ziep Michelstadt-Steinbach Mihajlovgrad Mileoak Milton Keynes-Bancroft Minerau Mogilets Monchy-Humières Moncrabeau-Bapteste Monréal Mons-en-Chaussée Montacute-Ham Hill Mont Beuvray Montmaurin Montrozier-Argentelle Morken Morthommiers-le-Crot Mühlacker-Lomersheim München-Denning Mundelsheim Munzenberg-Gambach Nadrin Nagold Namps-au-Mont Neckarrems NeckarzimmernStockbronner Hof

Mém Normandie 10, 1836, 369–87 Jb Schwaben 12, 1847, Supplement, 25–30 JÖAI(B) 29, 1935, cols 221–36 BJb 36, 1864, 55–71 BJb 133, 1928, 51–152 Steinhausen 1932, 180–2 Kurtr Jb 25, 1985, 33–9 J Roman Arch 1990, 195–204 RiBad-Württ, 441–2 Paret 1932, 117 Agache 1978, 325 Gallia 35, 1977, 333–4 BJb 74, 1882, 52–6; Arch Ausgr Bad-Württ 1977, 51–5 Ann Namur 33, 1919, 49–117 Drack 1990, 61; Koethe, 23 Ber RGK, 1933, 78–9 Agache 1978, 337 Fig. 30 Agache 1978, 290, 294, 332 Fig. 25 Jahresber Hessen 2, 1912, 56–8 Izv Muz Severozapadna Balgarija 4, 1979, 9–64 see Towcester Williams and Zeepfat 1994 Acta Mus Napoc 10, 1973, 144–7 Arheologija (Sofia) 11, 1969, 26–35 Agache 1978, 294, 327 Congr Arch 41, 1874, 40–7 BJb 143–4, 1938–9, 408–23 Agache 1978, 286 J Roman Stud 3, 1913, 128–33 Bulliot 1899 2, 166–72; Grenier 1934, 827 Fouet 1969 Congr Arch 1863 Hinz 1969 Leday 1980, Pl. 49 Fdba Bad-Würrt 16, 1991, 175–201 Czysz 1974 FdbaSchw NS 11, 1935–8, 105–11; Arch Ausgr BadWürtt 1988, 183–7 Germania 29, 1951, 150–3; RiHessen, 445 Arch Wallon 1980, 101 Aus dem Schwarzwald 33, 1925, 3–4 Agache 1978, 332 FdbaSchw 15, 1907, 42–5 Schumacher 1896, 9–12


— List of villas — Neerharen-Rekem Nemesvámos-Balácapuszta Nempont-Saint-Firmin (Somme) Nennig Neuburg a.d. Donau Neuhausen auf den Fildern NeumagenDhron-Papiermühle Neuss Newel Newport, I.o.W. Newton St Loe Niedereschach-Fischbach Niederzier-Hambach 59 69 77/264 403 512 North Cerney-The Ditches Northchurch North Leigh North Leigh-Shakenoak North Stainley-Castle Dykes North Warnborough North Wraxall Norton Disney Novi Šaher Noyers-sur-Serein Nünschweiler Nuth-Vaasrade Oberentfelden Oberesslingen Ober-Grombach Oberlunkhofen Ober-Ramstadt-Pfingstweide Oberweis Oberweningen Odell Ödheim Odiham-Lodge Farm

Bull Comm Roy Art 27, 1888, 325–74; Arch Belg 253, 1983, 56–60; 1, 1985, ii, 53–62 Thomas 1966, 73–107 Agache 1978, 131, 161 Wightman 1970, 145–7 Bayer Vgbl 26, 1961, 128–34 FdbaBad-Württ 3, 1977, 355–73 TrZft 40–1, 1977–8, 424–5 Neue Ausgr Deutld 1958, 294–5 TrZft 34, 1971, 143–225; RiRhld-Pf, 503–6 Arch J 9, 1929, 141–51 Proc Somerset 112, 1968, 104–5 FdbaBad-Württ 13, 1988, 351–93 Ausgr Rhld 1985/6, 40 Gaitzsch 1986, Abb. 6, 411–12; Ausgr Rhld 1981–2, 142–5 BJb 180, 1980, 467–91 Gaitzsch 1986. Abb. 6 Gaitzsch 1986, Abb. 7 Britannia 17, 1986, 412 Neal 1974–6, 1–52 VCH Oxon I, 316–18; Wilson and Sherlock 1980 Brodribb, Hands and Walker 1968–73 Arch J 32, 1875, 135–53 see Odiham-Lodge Farm Wilts Arch 7, 1862, 59–74 Ant J 17, 1937, 138–78 WMBH 6, 1899, 531–3 Gallia 20, 1962, 464–7; 22, 1964, 331–2 Mitt Pfalz 80, 365–7 Oudheidk Med NS 15, 1934, 28–32 Argovia 50, 1939, 153–9; Ur-Schweiz 22, 1958, 33–43 RiSchw, 457–9 see Esslingen see Bruchsal Anz Schw Altkde 2, 1900, 246–57 Schmidt 1971; RiHessen, 459–60 TrZft 9, 1934, 20–56 RiSchw, 460–1 Current Arch 66, 1979, 215–18 BJb 39–40, 1866, 213–16; Schr Württ Alt-Ver 7, 1866, 19–29 Proc Hants 10, 1931, 225–36 352

— List of villas — Odrang Old Durham Olfermont Olveston-Tokington Park Orazio Orlandovtsi Ormalingen Orton Longueville-Orton Hall Farm Öschelbronn Oss-Ussen Osterfingen Otford Overasselt Overstone Ovillers Pangbourne-Maidenhatch Panik Park Street Parndorf Peiting Perignat-Izernore Petersfield-Stroud Pforzheim-Hagenschiess Piddington Pilsdon Pen Pitney Plachy-Buyon-Les Trois Cornets Plasmolen Pleven-Kailuka Plomelin-Perennou Plouventer-Kerilien Port-le-Grand Proboj Primelles-Champ Chiron Puig de Cebolla-El Vilar Pulborough Pully-Prieuré Quarteira Quevauvillers Quintanilla de la Cueza Quinton Radley-Barton Court Farm Rainecourt

see Fliessem Arch Aeliana 4th Ser., 29, 1957, 203–12 de Caumont 1870, 385, 387 Trans Bristol 13, 1888, 159–69; 1889, 196–202 Lugli 1966 Dremsizova-Nelcinova 1969, 509; Nikolov 1976, 169 Basler Zft 9, 1910, 77–93 Todd 1978, 209–28 Bad Fdb 3, 1933–6, 321–9 van der Sanden and van der Broeke 1987 Schaff Beitr 14, 1937, 313–24; JbSGU 28, 1936, 73–4 Arch Cant 42, 1930, 157–71 Oudheidk Med NS 15, 1934, 13–18 Northants Arch 11, 1976, 100–33 Agache 1978, 287 Berks Arch J 65, 1970, 57; Britannia 2, 1971, 284 Glasnik Sarajevo 15–16, 1961, 132–8 see St Stephen see Bruckneudorf Bayer Vgbl 22, 1957, 223–5 Bull Arch Com Trav Hist 1909, 3–13 Arch J 65, 1908, 58–60 BJb 79, 1884, 28–104 Friendship-Taylor 1989 Proc Prehist Soc 43, 1977, 263–86 VCH Somerset I, 326–7 Agache 1978, 287, 338 Fig. 31 see Groesbeek Dremsizova-Delcinova 1969, 509; Bull Inst Arch Bulgare 14, 1940–2, 275 Bull Mon 3, 1837, 165–74 Ann Bret 77, 1970, 285–94 Agache 1978, 331 Fig. 24 WMBH 3, 1895, 280–3; Wilkes 1969, 397 Leday 1980, Pl. XL VIII Gorges 1979, 433 Proc Soc Ant Lond, 2 ser., 23, 121 Arch Schw 1, 1978, 87–92; RiSchw, 471–3 Gorges 1979, 482–3 Agache 1978, 287, 329 Fig. 22 Gorges 1979, 338 J Northampton Mus 11, 1972, 1–21 Miles 1986 Agache 1978, 287 353

— List of villas — Ramsbury-Littlecote Rankovici Rapsley Ravensbosch Raversbeuren Reach Redlands Farm Regelsbrunn – larger villa Regelsbrunn – smaller villa Regensburg-Burgweinting Reimlingen Remmingsheim Rheinbach-Flerzheim Rheinfelden-Herten/ Warmbach Rheinfelden-Salzbrünnele Ribemont-sur-Ancre Ridgewell Rielves Rijswijk-de Bult Ringstead Rivenhall La Roche-Maurice-Valy Cloistre Rockbourne Rodmarton Rognée Rohrbach (Aus.) Rome-Domus Augustana (Flavian Palace) Rome-Via Gabina Romegoux-La Vergnée Rommelshausen La Roquebrousanne Rothselberg Rottenburg am Neckar Le Roux-lez-Fosse Roxby-cum-Risby, Dragonby Rudston Rulles-Chaumont Saaraltdorf SachsenheimGrossSachsenheim

Britannia 20, 1989, 315–17; Walters and Phillips 1978 See Travnik see Ewhurst see Schimmert op den Bilik BJb 61, 1877, 128–34 RCHME Cambridgeshire North-East, 88–9 see Stanwick Thomas 1966, 257–9 Thomas 1966, 259–60 Röm-Germ Korrbl 7, 1914, 54–7; Verhand Hist Ver Oberpfalz 67, 1917, 24–30 Bayer Vgbl 11, 1933, 89–90 Paret 1932, 38–9, 363 Ausgr Rhld 1982–3, 154–8; van Ossel 1992, 219–22 Archaeol Ausgrab Bad-Württ 1984, 144–7 Bad Fdb 3, 1933–6, 210–19 Agache 1978, 287, 329 Fig. 22 VCH Essex III, 170–1 Gorges 1979, 422–3 Bloemers 1978 Northants Arch 15, 1980, 12–34 Rodwell and Rodwell 1986 Ann Bret 79, 1972, 215–51 Arch J 140, 1983, 129–50 Archaeologia 18, 1817, 116; RCHME Gloucestershire Cotswolds, 98–9 Doc Charleroi 21, 1897, 3–75 JÖAI(B) 1966, 23–5 Ward-Perkins 1981, 78–84 Widrig 1987 Rev Arch 6th ser., 16, 1940, 46–61 see Kernen Gallia 44, 1986, 477–8 Pfalz Mus 29, 1912, 31–3 Archaeol Ausgrab Bad-Württ 1981, 141–5 Ann Namur 29, 1910, 144–50 Ant J 50, 1970, 222–45 Stead 1980 Lambert 1987, 125–33 Jb Lothring 20, 1908, 157–76 Arch Ausgr Bad-Württ 1982, 127–34


— List of villas — Saint-Acheul Saint-Aubin-le-Mazaret Saint-Aubin-sur-Mer Saint-Frégant-Keradennec Saint-Germain-lès-Corbeil Saint-Herblain Saint-Julien St Lythans-Whitton St Mary Bourne-Finkley St Michael-Gorhambury Saint-Pierre-la-Garenne Sainte-Solange St Stephens-Park Street Saint-Ulrich Salzburg-Liefering Santa Colomba de Somoza-Maragatera Sarajevo-Stup Sarica Sarmentsdorf Sarmizegetusa Sarratt Saunderton Sauvenière Schaanwald Schambach Schimmert-op den Bilik Schleidweiler Schleitheim Schuld Schupfart-Betburg Selongey Serville Sette Finestre Settrington Shakenoak Shipham-Star Sigean-Pech-Maho Sigmaringen-Laiz Sigmaringen-Steinäcker Silchester Sinsheim Sivry

Bull Picardie 57, 1977–8, 293–306 Gallia 46, 1989, 285; Todd 1992 Gallia 6, 1948, 365–75 Ann Bret 79, 1972, 168–99 Gallia-Inf 1993, i–ii, 43–4 Gallia 41, 1983, 320–1 Gallia 27, 1969, 318–21 Jarrett and Wrathwell 1981 VCH Hampshire I, 302–3 Neal, Wardle and Hunn 1990 CAG 27 Eure, 186, no. 374; Coutil, 1898–1921, II, 272–7 Leday 1980, 2, Pl. 23 Arch J 102, 1945, 21–110; 118, 1961, 100–35 see Haut Clocher Mitt Salzburg 108, 1968, 341–60 Gorges 1979, 276–7 Glasnik Sarajevo 42, 1930, 212–25 Baumann 1983, 124 Anz Schw Altkde 32, 1930, 15–25 Miclea and Florescu 1980, 1, 62 VCH Hertfordshire IV, 1914, 163 see Bledlow Ann Namur 24, 1900, 11–32 Jb Liechtenstein 29, 1929, 149–55 Jahresber Bayer Bodendenkmalpfl 6–7, 1955–6, 14–34 Oudheidk Med 2, 1908, 25–44 see Zemmer Wanner 1899, 21–3; RiSchw, 505–6 Eiden 1982, Tafelband, 101–11 JbSGU 23, 1931, 77–9 Leday 1980, Pl. 23 Ann Namur 24, 1900, 21–6 Carandini and Settis 1979 Ramm 1978, 76–7 see North Leigh Proc Somerset 108, 1963, 45–93 Dedet 1987, 186, Fig. 167 FdbaBad-Württ 16, 1990, 441–508 Mitt Hohenzollern 16, 1882–3, 104–6; RiBad-Württ, 556–7 Proc Prehist Soc 53, 1987, 271–78 Wagner 2, 364–6 Lambert 1987, 143–4 355

— List of villas — Skeleton Green Smarje-Grobelce Sofia-Obelja Somerton-Bradley Hill Somerton-Catsgore Sontheim an der BrenzBeim kleinen See Sotzweiler Southwick Sparsholt Spoonley Wood Stadtbergen Stahl Stammheim Stanwick-Redlands Starzach-Bierlingen Stein Steinbach im Odenwald Stockbronner Hof Stolac 1 Stolac 2 Stolac 3 Stolac 5 Stolac 6 Stowey-Sutton-Chew Park Stratford-upon-AvonTiddington Stroud Strupnic Stutheien-Huttwilen Stuttgart-Stammheim Sudeley-Spoonley Wood Sudeley-Wadfield Suippes Sumeg Swindon-Okus SzentkirályszabadajaRomkút Szilágy-Arnyoldal Tác-Fövenypuszta

see Braughing Thomas 1966, 344–9 Arheologija (Sofia), 1981, 52–6 Britannia 12, 1981, 177–252 Leech 1982 FdbaBad-Württ 3, 1977, 334–54 see Tholey Sussex Arch Coll 73, 1931, 13–32; 123, 1985, 73–84 Britannia 4, 1973, 318–19; 22, 1991, 288 see Sudeley Bayer Vgbl 39, 1974, 114–26 BJb 62, 1878, 1–7 see Stuttgart Current Arch 122, 1993, 52–5; South Midlands Arch 21, 1991, 76–9 Württ Vierteljahresh 10, 1887, 77–80; FdbaBadWürtt 1, 1974, 501–26 Oudheidk Med 9, 1928, 4–9 see Michelstadt see Neckarzimmern WMBH 1, 1893, 287–90; Wilkes 1969, 397 WMBH 1, 1893, 290–1; Wilkes 1969, 397 WMBH 1, 1893, 291–5; Wilkes 1969, 397 WMBH 3, 1895, 272–80; Wilkes 1969, 397 WMBH 5, 1897, 169–72 Rahtz and Greenfield 1977 Palmer 1982 see Petersfield Wilkes 1969, 404–5 Roth-Rubi 1986 FdbaSchw 19, 1911, 84–9; FdbaBad-Württ 5, 1980, 140–2 Archaeologia 52, 1890, 651–8; RCHME Gloucestershire Cotswolds, 113–14 J Brit Arch Ass 2nd ser., 1, 1895, 242–50; RCHME Gloucestershire Cotswolds, 112–13 Villes 1985, 656, Figs 6–8 Thomas 1966, 111–16 Wilts Arch 30, 1898–99, 217–21 Thomas 1966, 118–22 Thomas 1966, 297–9 Thomas 1966, 299–326 356

— List of villas — Tailly-L’Arbre-à-Mouches Tarrant Hinton Tavaux Telita Tengen-Busslingen Terrick Téting-sur-Nied Thalerhof Tholey-Sotzweiler Thoraise Thuit Tiddington Tiefenbach Titelberg, ‘Aedicula’ Titsey Tockington Park Tollard Royal Towcester-Mileoak Travnik-Rankovici Treuchtlingen-Weinbergshof Trier-Basilica Trouey Upchurch-Boxted Uplyme-Holcombe Urreiting Uxheim-Ahutte Val Catena Valladolid-Prado (or Granja de Jose Antonio) Vallenay-Patureau Fourneau Vaux-sur-Somme-Bosquet Duval Verberie Verneuil-en-Halatte Le Bufosse Vesqueville Vicques Vierherrenborn-Irsch Vieux-Rouen-sur-Bresle Vila de Frades Ville-sur-Ancre

Agache 1978, 139, 156 Proc Dorset 104, 1982, 184–6; 105, 1983, 146–8; Britannia 13, 1982, 386–7 Jeannin and Chouquer 1978, 271 Pence 9, 1984, 51–65; Baumann 1983, 125–46 RiBad-Württ, 582–4 see Ellesborough Grenier 1906, 159–67 see Graz Germania 39, 1961, 474–8 Villes 1985, Fig. 5 de Caumont 1870, 388–9 see Stratford-upon-Avon Schumacher 1896, 6–9 Hémecht 22, 1970, 378–80 Surrey Arch Coll 4, 1869, 214–37; VCH Surrey IV, 1912, 367–9 see Olveston Proc Prehist Soc NS 34, 1968, 102–47 Northants Arch 13, 1978, 28–66 Glasnik-Sarajevo 10, 1955, 134–7 Arch Jahr Bayern 1984, 113–15; Bayer Vgbl 51, 1986 326–32 Zahn 1991 Leday 1980, Pl. 56 Arch Cant 15, 1883, 104–7; VCH Kent III, 1932, 106–8 Proc Devon 32, 1974, 59–162 see Markt Pongau Tr Zft 24–6, 1956–8, 552–60 see Brioni Grande Gorges 1979, 444 Leday 1980, Pl. VIII Agache 1978, 287; Agache and Bréart 1975, 126 Rev Arch Picardie 1983, 96–126 Gallia-Inf 1989, 244–6 Arch Belg 159, 1974 Gerster 1983; RiSchw, 531–3 TrZft 14, 1939, 248–53 Agache 1978, 297 Gorges 1979, 477 Agache 1978, 328 357

— List of villas — Villeneuve-sur-Cher, Les Augerets Villeneuve-Saint-Germain Villers-sous-Ailly Villers-Bretonneux Villers-sur-Lesse-Génimont Vix Voerendaal-Ten Hove Voerendaal-Ubachsberg Vouneuil-sous-Biard-Les Cassons Wachenheim Nordrh.-Westf. Wadfield Wahlen Walsbetz Walton-on-the-Hill, The Heath Wancourt Warfusée-Nord (-Abancourt) Warfusée-Sud (-Abancourt) Watergate Weeze-Baal Weitersbach Wellow Welschbillig Welwyn-Dicket Mead Welwyn-Lockleys West Blatchington West Dean-Chilgrove (Sussex) West Dean (Wilts.) Westerhofen Westheim-Hüssingen West Marden-Watergate West Meon-Lippen Wood West Tytherley-West Dean Weyeregg Weyhill-Clanville Whippingham-Combley Whittington Whitton Whitwell Wickford

Gallia 42, 1984, 290–1 Gallia 37, 1979, 307–8; Villes 1985, Fig. 9 Agache 1978, 294 Agache 1978, 294 Ann Namur 30, 1911, 190–1; de Maeyer 1937, 67–8 Villes 1985, Fig. 5 Oudheidk Med NS 34, 1953, 48–79; ROB 266, 278 , 286 Oudheidk Med NS 4, 1923, 65–77 Gallia 37, 1979, 405–6; Aucher and Aucher 1984

see Sudeley RiSchw, 535 de Maeyer 1937, 106–7 Surrey Arch Coll 2, 1864, 1–13 Agache 1978, 287 Agache 1978, 331 Fig. 24 Agache 1978, 321 see West Marden BJb 166, 1966, 379–432 TrZft 24–6, 1956–8, 511–26 VCH Somerset I, 312 TrZft 40–1, 1977–8, 443–4 Herts Arch 9, 1987, 79–165 Ant J 18, 1938, 339–76 Sussex Arch Coll 90, 1952, 221–40 Down 1979 see West Tytherley Sammelbl Ingolstadt 55, 1937, 3–35 Arch Jahr Bayern 1980, 134–5 Britannia 16, 1985, 314 Arch J 62, 1905, 62–4 Wilts Arch 22, 1885, 243–50 80 Jahresber Oberösterr Musver, 1922–3, 63–80 Archaeologia 56, 1898, 1–20 Britannia 7, 1976, 364–5 RCHME Gloucestershire Cotswolds, 126–8; Trans Bristol 71, 1953, 13–87 see St Lythans Britannia 8, 1977, 392, 395 Cunliffe and Rowley 1978, 29 358

— List of villas — Wiesbaden-Höfchen Wiesendangen-Steinegg Winden am See Winkel-Seeb Winterton Wittlich Wollersheim-Am Hostert Woodchester Woolaston-Chesters Worsham Wraxall The Wrekin Zaloje Zazenhausen Zemmer-Schleidweiler Zijderveld Zofingen

Nassau Ann 3, 1876, 22–7 JbSgU 57, 1972–3, 343–5 Burgenl Forsch 13, 1951, 3–47; Thomas 1966, 201–10 Drack 1990 Stead 1976; Todd 1978, 93–103 Westd Zft 25, 1906, 458–61; TrZft 16–17, 1941–2, 229–35 Germania 34, 1956, 99–125 RCHME Gloucestershire Cotswolds, 132–4; Britannia 13, 1982, 197–228 Arch Cambrensis 93, 1938, 93–125 see Asthall Proc Somerset 105, 1960–1, 37–51 Arch J 141, 1984, 61–90 see Biha BJb 39–40, 1866, 210–12 Jahresber Ges Nutzl Forsch Trier 1900–5, 31–9 Ber ROB 23, 1973, 103–7 Keller 1864, 150–2; Hartmann 1975



agriculture 10, 30, 33, 36, 45, 144, 151, 275 aisled farmhouses 30, 45 aisled house: cattle accomodation 45, 222–3; and decline of hall 263; definition 26; development 15; distribution of 220–2; hall 36–9; interpretation of 222–6; three aisles 225–7; row house characteristics 263; two aisles 228 alignment 186, 279, 300 altar 289 alternate development, principle of 15 anteroom 76, 79 apartment 48–9, 76, 87, 89–91, 105 apse 200, 207, 214 arches in porticuses 131 architects and diffusion of ideas 284, 288 articulation 77, 83, 91 assimilation to Roman rule 278 asymmetry 49, 102–3, 120–3 audience hall 176–8, 180, 182, 193 axial posts 36, 39, 93, 223–4, 226 axial symmetry 136 axiality 18–19 back-to-back: halls 115–16; houses 109–12; lobby 60, 290 bailiff 5, 74, 301

bank and ditch 152, 213, 230–1 barracks, slave 295–6, 300 basilica 163, 173, 176, 183, 185, 214, 257, 301; ‘basilican’ building, 40 baths 18, 24, 89, 129, 138, 163, 190, 197–8, 213–14, 238–9, 263, 301 bay length 223–5 block plans 204 boundaries: house built over 239–43, 249–50; land redistribution 277; permanent 285 breadth 26 building: techniques 257, 278; transition and change in 279 buildings: block grouping, 209–11; square 227–8, 232; trapezoidal 227–8 byre 26, 30, 36, 223–5, 229 cattle accomodation 30–1, 33, 35, 37, 45, 222–3 cellar 24, 33–4, 41–2, 81, 87, 101, 118 circulation 17, 63, 82, 89, 92 Clarendon, his house 5 classification, problems of 26, 46 closet 79, 124, 127 coin lists, interpretation 20 colonialism 278 colonnade 117–18, 131, 134 complex houses 231


— General Index — continuity of settlement 116, 228, 243–9, 254, 271, 273, 284 cooking 27, 29, 31, 32; and see kitchen 36 corridor 202, 204, 263; axial 112–15; inappropriate term, 142–3 cost 45, 110, 151, 156, 193, 219 cottage-house, cottages 5, 8, 26, 56, 74 country houses 5, 18 courtyard: distribution of types 283; divided 148–51; domestic 162–5; fanshaped 156–7; forms of 212–13; geometrical 151; hall 101–2; houses 18; irregular 144–9; for kin/slaves 295–300; long rectangular 159–62; open 24; rectangular 281; rectilinear 213; rhomboidal 158–9; as Romanisation 250–2; open or roofed 42; small house with? 94–100; social change and 285–8; sub-triangular 157–8; tapering 152–5; trapezoid 281; variety of shape 144; villa 106, 214–15, 281–2 crafts 30–1, 45 cultural values, Roman 233 dais, medieval, as place of privilege 29– 31 demolition of early villas 253 design, common approach to 204 dining room 120 display, architectural 5, 7, 13, 79, 80, 130, 254 ditches 149, 152, 213, 230–1, 240, 249–50, 277 domestic work 30, 45 doorways 14, 34, 37, 70, 72, 74, 77, 79, 83 dower houses 106 drainage 30, 42, 83 duality: architectural 57–9; of household 48, 50, 56, 88–9, 137–9; of occupation 6, 44, 45, 59, 87, 90, 102–3, 119, 208, 208–9, 282, 284; of use 75 elongation of buildings 211

emperors, travels 172–3, 182 enclosure 144, 158, 212, 213, 230, 239, 277; earthwork 219 end rooms: around lobbies 83–5; in broad halls 100–1; as byres 82–3; Inzigkofen and analogues 86–8; narrow 80–2, 88 entertainment 178, 268 entrance: axial 202–5; indirect, to dominant house 156; end 199–200, 289; hearth relationship 14, 29; lobby 56; one end 54–6; open ended porticus 130, 132–4; porticus and pavilion 119–20, 121, 125, 138; position 26; right angle 148; room function and 75; square house 228; two 17; two front 59–60; villa 220 equality: of household 74; and inheritance 276–7 estates, alleged; growth 285, 301 facade 117; detached 166–9 family 32, 45, 48, 76, 102, 105, 204, 208, 228, 231 farm: as food source 278; tasks 143 farmers 5, 57 farmhouse 5–6, 15, 30, 118 farmyard 99, 144, 158, 171, 212–13 feasting 178, 268–9 flooring 36, 80, 201 focal points 27, 32, 69, 131, 152, 156, 203 fragmentation of land holding 276–7, 285, 300 function of rooms: end 86; hall 36–7, 80–2, 89–91; hall or yard 41–2, 99– 100; mosaic and 4; public/private 178–9, 182; size difference and 65; social, of architecture 283; state/ citizen 173; typology and 13, 15 gallery, porticuses 138, 140, 143 gardens 42, 147, 189 gender division and occupancy 48 ghost pavilions 140–1 government, palace as seat of 172–3, 176–8, 181


— General Index — granary 227 grouping of buildings 208–11; linear 208–9 grouping of rooms, social significance of 91–3 “gwely”, clan 277, 279, 284, 299 Hale, Sir Matthew; mean house, 4 hall: activities in 102; aisled 36–7, 40–1; axial entrance 202–5; broad 26, 29, 31–2, 33, 35–6, 45, 288; classification and problems 26; decline 301; development 80–93; distribution of 283; divided 43–5; double ended 34, 35–6; early 253; elements and analysis 14; function and layout 26–8, 200; hearth position 29, 31–3, 136; lack of 65; medieval 15, 17, 23, 26, 29, 86, 225, 261; narrow 26, 32–3, 80–2; oblong 201; open 23–4, 100, 106, 257–62; pavilions 118; plan analysis 13; porticus unified with 43; as principal room 46; ridge posts 37–9; in row-house 62; S.E. Europe 199–200; single-ended 33–4; social change and 285–8; square 200–1; 25; transition to row house 285, 288; tripartite 114; unrectilinear 252; wide-nave 26, 39–40, 93; or yard 8, 41–2, 94–101, 202 halls: divided 43–5 hearth: central 23, 29, 30, 33; communal 33; entrance relationship 14, 29, 75; more than one 31–2; off centre 29; open 27; or oven 26, 32–3; pavilion 129; position 36; roofs and 41; secondary 142 Herrenhaus, -sitz 138 honour, place of 29, 31 hospitality: Elizabethan 172; justice and 179 house: courtyard 162–5; double depth 106–12; five room 50, 75; four room 60–2; L-shaped 211–12; multiple small room 65, 207–8; one room 23, 104–6; small, or yard 94–9; three

room 102, 105, 126; two room 204– 5, 208 household: complexity of units 67; division and status 86–7, 89, 91; duality of 48, 50, 56, 88–9, 137–9; hierarchy 68–9, 119; kin group and status 76; more than one 43–4; religion 290; row-house 46; size and rooms 102; status 79; two in row house 102–3; two and symmetrical entrances 136–7 hypocaust 101, 213, 301 hypocauston 59 industrial production 275 inheritance, partible 16, 276–7, 279, 281, 284–5, 299–300 intercommunication of rooms 6, 14, 42, 47–8, 55, 76, 82, 117–20, 129 Iron Age: pre-Roman houses, settlements 219–32, 277 justice: private 178–9 justice, administration from villa 106, 172, 193 kin group: cooperation 209; dominant household 132; entertainment 178; equality of household 74; head 119; and house type 278–9, 281–2, 283, 288, 290, 292, 295; lobby size 79; occupation 6; possession of land 275, 277, 284, 300, 302; loss of status 271; Romanisation 233, 238; round-house 231; social hierarchy 149, 151, 156–8, 163, 167, 238; status 254; in late villas 299, 301 kitchen 265–9, 273 land holding 275–7, 281, 284–5; fragmentation of 276–7, 285, 300 landed proprietorship 74, 87 lighting 78, 90, 110, 113–14, 131, 141 living rooms 46, 123 lobby: and circulation 90–2; distribution 75; L-shaped 70–4, 91; La Roche-


— General Index — Maurice 74–5; longitudinal 56–8, 70, 78; oblong 75; position of 51; or room 68–9, 73–5, 76; small 48, 50; square 72, 75, 284; transformation 78–9; transverse 48, 50, 72, 76, 78, 204, 284; types 70–9; widened 74 lodging, official 176 long-house 15, 83 lordship 57, 149, 200; hereditary 300; and joint proprietorship 195–8 lugs, on porticus 139–41 luxury: and formality 193–5; Romanisation and 238–9; social hierarchy and 172; villa 190–3, 301 manor house, Elizabethan and Jacobean 140; later 65, 67, 123 mansiones, see staging-posts matters ignored: chronology 19–20; surroundings 20 monospan buildings 36, 228–30, 232 mosaic 4, 129, 141–2, 213, 300–1 naves 36; and aisles of equal width 40–1; division 263 nobility and partible inheritence 276 nomenclature, room 11 North, Roger, and buildings 5 ornament 18, 65, 79, 80, 123, 131, 193, 278 palaces 172–7 palisade 230 parlour 86, 118, 124 partitions 36, 99, 101, 201, 221, 231 passage room 35, 75–6, 80, 92, 100–1, 204, 208 pavilion corner projection: as addition 13, 48; asymmetry 120–1; asymmetry and status 120–2; classic form 117– 19; detached wings or quasipavilions 128; difference and social significance 123; entrance 119–20; ghost 140–1; lack of 111; minimal 125–6; not rectangular 124–5; oblong 123–4; round 254–6; in row-house

126–8; in S.E. Europe 214; single 132; storeyed 128–9 peristyle villas 184–90, 193, 214–15, 283 pilasters 89, 121 plan analysis, origins of 13–14 pool, ornamental 89, 167 population stability of villa 271–3, 293–5 porticus: as architectural display 13; compact row-house 63; as entrance to pavilion 119; functions 139–43; as link 166; as living space 141–2; lobby and 56; as open screen 118; open–ended 130–4; splayed and tapering 134–5; two households and 136; wide or continuous 142–3 porticus with pavilions: as additions 33, 35–6; distribution 283; duality of occupation 45; as feature of villa 7; front and rear 137–9; hall houses 24, 80; Romanisation 41, 117; S.E.Europe 202, 209, 214; symmetry 19; yard and 97, 115 posts: axial 36, 39, 93, 223–4, 226; internal 97; king and queen 42, 95, 97–9; ridge 26, 36–9, 220–1, 223–9, 232 power: and planning 178; relations 116, 151, 160; and representation 278 privacy, use of lobby 48 privilege, place of 29, 31 proportion 26, 33, 35, 75–6; dis-, in end rooms 88 proto-villas 11 public/private space 86 quasi-pavilions 128 range: end rooms in broad halls 100–1; parallel 106–9, 114, 166–7; width 212 rebuilding 13, 15, 18, 171, 250, 253 recreation, porticus as space 138–41 regionalism 102, 171 religion, domestic 14, 91 removal to a new site 250 representational room 46, 60, 75, 91, 290


— General Index — reverse symmetry 57–9, 63, 70, 212, 283 Romanisation: definition 251; hall houses and 23; modes of 233–56; pavilion and 117, 126; plans and 18; porticus with pavilions 41, 134, 138; settlement 279, 288 roof: construction 10, 23; monospan 41–2; open halls 23–4; span 95, 97, 99, 110, 228, 257–8; support 26, 36; trusses 225 rooms: additional 26, 33, 80, 89–90; ancillary and halls 261–3; blocks of small 77–8, 207–8; development 15; division 268; enclosed 24; five room house 75; groupings 6, 14; groupings and social significance 91–3; Lshaped 201; minimum number for living 57; sequence and power 178; size 49; state 173, 176; subdivision 35; subsidiary 23; three and lobby units 65–7; use 6, 15–16, 46 round-houses 228, 230–2, 254–6, 277 row-house: articulation or separate rooms 76–7; Bierbach units 68–9; blocks of small rooms 77–8; compact 61–4, 78, 134, 195, 209, 211; development 65–79, 240; distribution of 283; Downton 49–51; early 290; elaboration of units, Laufen-Müschag 67–8; end entrances 54–6; equally divided 136; four cell 60–2; hall or 23, 84–7, 91; interconnecting rooms, Newport and Lamargelle 47–9; Kirchberg, three room and lobby units 65–7; lobby transformation 78–9; lobby types 70–5; longitudinal lobby 56–7; pavilions 126–8; porticus and pavilion front and rear 139; pre-Roman 232; reverse symmetry 57–9; room proportions 75–6; S.E.Europe 202–3; size 100, 102, 111; size and form of units 51–4; smallest 102–5; Sparsholt 49; two front entrances 59–60; unit significance 79; wide porticus 142; workhall 228 sampling 11–12

sequence of rooms 49–50 servants 30, 45, 76, 143, 183 service quarters 86 settlement: consolidation 279–82; dispersed 277; evidence of preRoman 219–20 shrines 48, 63, 106; freestanding 291–2; as unifying device 288–91 slaves 295–6, 299 social circumstances and villa design 219 social development: of hall 257; prosperity and 274; of row-house 270–1; of villa 264–70 social hierarchy: kin group 60 social hierarchy: buildings and 56–7, 121–3, 149, 156, 214–16; buildings and rooms 16; halls and yards and 285–8; justice and 178–9; lobby and 68; luxury and 172, 193; organisation 85, 87; rooms and 117 social history and architecture 4 social meaning: of axial corridor 114–15; of courtyard 151 social stratification 87, 115–16, 277, 289, 300 social structure: plans and 16–18; porticus and 130–2; set of hearth 29; siting of shrines 291; symmetry and 136; terminology 5; and villa trsansition 281–2 ‘squatter occupation’ 142 staging posts 172 staircases 65, 67, 128 standard of living 90, 271 status of house 124, 134, 208, 228, 278, 289 status, social: asymmetry as 121–3; of household 72, 79, 86; population and decline in 271; porticus with pavilions as 7; rooms and 48; size and 65, 67; space and 16; within kin group 275, 278–9, 299 stone footings 23, 37, 42, 132, 227, 232, 253–6, 279, 281 storage 143


— General Index — storeys 10, 24, 40, 128–9 superior/inferior ends: difference 37, 79; division 263; equality 86–8; family and servants 45; function and 26–8; lower end 30–1; porticus and 130; square rooms 75 surveying place 73 symmetry 14, 18–19, 50, 62, 65, 67, 87–8, 90–1, 102, 118, 128, 131–2, 136–7, 152, 252, 289–90 temples 106, 291–2 terminal rooms 111, 127, 129 terrace 87 tie-beams 95, 131 timber: buildings 10–11, 144, 227–8, 231, 250–2, 258, 279; staircase 128 title, legitimacy of 254 towers, fortification 215–16 transition, native to Roman 233–8, 250, 253–4, 279–81 travellers 172 type of house: hall 23–45; and range available 219; row-house 46–64 typology: objections to 4–5; of Roman villas 4, 6–8, 13 unification of buildings 240–1 unit: cells 55; conjugal 228; consumption 92; disparate 79; Downton 55, 59, 65, 102, 109, 136, 150, 185, 202, 261, 267, 279, 288; elaboration 67–8; four cell and two household 61; row type villa 68–9; size and form 51–4; symmetry 290; system 46, 48, 50, 52–4, 57 ventilation 110

veranda, timber 130, 132 vernacular architecture, relevance of 4, 14 viewing places 141 villa: architecture 9–10; classification, history of 6–9; constituents of 10–11; courtyard 163, 214–15; definition of 251–2; difference 292–3; diffusion of types 282–4; emergence of 278–81; fortification 215–16; hall type 199– 200; Iron Age preRoman 219–32; linear 167–71; luxury 173, 190–3; obsolesence 301; patterns of development 257–74; peristyle 184–90, 193; replacement of timber building 252–3; S.E. Europe 199–216; seat of lordship 178–83, 193–4; system development 284, see partible inheritence wage labour 114–15 waiting places 73, 79 walking indoors, Pepys and Clarendon 140–1 wall: construction 95, 97–8, 220; enclosure 149, 152, 158, 213; supporting 23, 128 warmth 14, 24, 27, 29, 31, 33 water cult 291 wings 80, 91, 106, 117, 126–7, 262, 282, 284; cross205–7, 288; detached 128; halls and 288 workhalls 57, 79, 87, 89, 91–2, 157–8, 163, 228



Achenheim (Alsace) 231 Aguilafuente-Santa Lucia (Segovia) 179, 182 Ahrweiler see Bad-Neuenahr Aiseau 119 Albesa-El Romeral (Lérida) 183 Allenz see Mayen Almenara de Adaja-Villa Romana (Valladolid) 179, 180 Alpnach-Dorf (Switz.) 96, 100, 105 Alresford (Essex) 55 Altstetten (Switz.) 252 Ambresin (Belg.) 108, 111 Anthée (Belg.) 66, 76–7, 78, 110, 159, 292–3, 294, 296, 299 Apahida (Rom.) 207, 208 Apethorpe (Northants.) 149 Arradon-Lodo (Morbihan) 166, 169 Ash (Kent) 40 Ashstead (Surrey) 111, 112 Ashwell-Ducklake Farmhouse (Herts.) 109–10 Asthall-Worsham 288 Aston Bury (Herts.) 65, 67, 123 Athies 159, 299 Atworth (Wilts.) 158 Aulnat (Puy-de-Dôme) 231

Ault (Picardy) 131 Aylesford-Eccles (Kent) 55, 59, 87, 110, 130–1, 167, 240, 242, 265, 289, 290, 293, 294, 300 Babworth-Dunston’s Clump (Notts.) 229 Bachenau 127 Bad Dürkheim-Ungstein (Bad.-Württ.) 112, 113–14 Bad Godesberg-Friesdorf (Nordrh.Westf.) 190, 301 Bad Homburg vor der Hohe (Hessen) 44, 149, 158, 213, 284 Bad Kreuznach (Rhld-Pf.) 301 Bad Nauheim 219 Bad Neuenahr-Ahrweiler 258, 261 Bad Rappenau-Zimmerhof (Bad.Württ.) 96, 100, 299 Badajoz-Dehesa de la Cocosa 181, 194, 214 Badbury (Wilts.) 163 Badgeworth (Glos.) 123 Bancroft see Milton Keynes Bargen im Hegau (Bad.-Württ.) 24, 25, 157 Barnsley Park (Glos.) 11, 18, 237, 238– 9 240, 248


— Site Index — Barton Court Farm see Radley Basel-Gasfabriek (Switz.) 220 Basse-Wavre (Brabant) 14, 107, 108, 110–11, 112, 116, 182, 194, 214, 292, 301 Beadlam (Yorks.) 57–8, 60, 72, 135, 241 Beckingen (Saarld) 127 Bedburg-Garsdorf (Nordrh.-Westf.) 39 221, 224, 247, 251, 252 Beegden (Neth.) 221, 228 Befort-Aalburg (Lux.) 222, 226 Bellersheim see Hungen Belleuse-les Mureaux (Somme) 162, 164, 300 Bellikon (Switz.) 248, 253 Bennwil (Switz.) 40 Bergen-Auf-dem-Keller 141 Betzingen (Bad.-Württ.) 82, 83 Biberach (Bad.-Württ.) 120, 124 Biberist-Spitalhof (Switz.) 104, 106 Bierbach (Saarld) 63, 64, 66, 68–9, 79, 90, 123, 134 Bierlingen see Starzach Bignor (Sussex) 155, 158, 163, 179, 181, 182, 240 Bihac-Zaloje (Yugosl.) 143, 201 Bilsdorf (Lux.) 97 Birkenfeld-Elchweiler (Rhld-Pf.) 278 Bistrica (Bulg.) 199, 200, 216 Blankenheim (Rhld-Pf.) 8–9, 19, 42, 54–5, 69, 73, 89–93, 95, 97, 131, 134, 140, 161, 182, 215, 261–2, 264–70, 273, 288–91 Blieskastell-Altheim (Saarld) 30, 33 Bocholtz-Vlengendaal (Neth.) 39, 85, 89–90, 91, 92, 93, 134, 150, 170, 221, 262, 263, 293 Böckweiler 30, 33 Bollendorf (Rhld-Pf.) 27, 29, 30, 32, 274 Bondorf (Bad.-Württ.) 95, 96, 97, 99, 100, 118, 124, 143, 286 Boos-Le Bois Flahaut (Seine-Mar.) 51, 53, 60, 62, 79 Börstingen 27, 29

Böttger 195 Bouchoir (Somme) 102, 103, 126 Box (Wilts.) 55, 163, 178–9, 180, 182, 301 Boxted see Upchurch Bözen (Switz.) 248, 253, 261 Bradley Hill see Somerton Brain-sur-Allones (Main-et-Loire) 31 Bramdean (Hants.) 51, 132 Braughing-Skeleton Green (Herts.) 227, 228 Bray-sur-Somme 139 Bregenz (Aus.) 114 Brewood-Engleton (Staffs.) 124–5, 128, 158 Brighstone-Rock 127, 128, 289 Brioni Grande-Val Catena (Yugosl.) 190, 191 Bristol-Brislington (Avon) 63 Bristol-Kingsweston (Avon) 43, 122, 124, 125, 141, 206, 241, 259, 261, 288 Brixworth (Northants.) 51, 60, 101, 185, 241, 255, 299 Broichweiden 127, 135 Brombach see Lörrach Bruchsal-Ober-Grombach(Bad.-Württ.) 27, 31, 145, 149, 150, 151, 159, 167, 185, 285 Brücken (Rhld-Pf.) 34, 35 Bruckneudorf-Parndorf (Aus.) 42, 97, 201, 207, 213, 259, 261–2, 301 Buchs (Switz.) 87 Buchten (Neth.) 36, 84, 87 Budakalász (Hung.) 200, 203 Budapest-Aquincum 175, 177, 182, 190 Budapest III-V Csúcshegy (Hung.) 132–4, 204, 265 Budapest III-Testvérhegy (Hung.) 171, 208 Bundenbach-Altburg (Bad.-Württ.) 221, 228 Burgweinting see Regensburg Byfield (Northants.) 28, 31, 121 Cachy (Somme) 152, 153 367

— Site Index — Caerwent 7 Canterbury (Kent) 229 Carisbrooke (I.o.W.) 37, 38 Cartagena-El Castillet (Murcia) 66, 69 Castle Dykes see North Stainley Cenero-Murias de Belono (Ovieido) 19, 52, 57 Champagnole (Jura) 72 Champdivers (Jura) 200 La Chapelle-Vaupelteigne (Yonne) 167 Chassemy (Aisne) 226–7 Chastres (Belg.) 27 Chastres-lez-Walcourt (Belg.) 135 Chatalka-Delimonyova Kale (Bulg.) 296 Chatalka-Delimonyova Niva (Bulg.) 143, 213 Chatalka-Lambata (Bulg.) 214 Châtillon-sur-Seiche/La Guyomerais 41, 165 Chedworth (Glos.) 17, 48, 136, 152, 157, 162, 163, 212, 214, 251, 289, 291, 300 Cherington (Wilts.) 83, 140 Chew Park see Stowey Sutton Chichmanovtsi (Bulg.) 208 Chiddingfold (Surrey) 107 Chilgrove see West Dean Chiragan 194 Cincis (Rom.) 205 Citernes (Somme) 152 Ciumafaia (Rom.) 207, 208 Civray-Le Poirier Molet (Cher) 61 Clairy-Saulchoix (Somme) 152 Clanville see Weyhill Claydon Pike see Lechlade Cléry-sur-Somme 51, 60 Clinchamps (Calvados) 31 Cobham Park (Kent) 50 Colchester-Camulodunum (Essex) 226 Colerne 83 Collingham-Dalton Parlours 101, 110, 114, 209, 239, 243, 245, 246, 248 Colsterworth (Lincs.) 231 Combe St Nicholas-Wadeford (Som.) 60 Compton Dando-Littleton (Som.) 55

Condé-Folie (Somme) 159, 240, 241, 242 Cosa (Italy) 215 Courcelles-sur-Nied-Urville 29, 81, 83 Cox Green see Maidenhead Crain (Yonne) 30 Croft Ambrey (Herefs.) 227 Cromhall (Glos.) 152 Cuevas de Soria-Dehesa de Soria 183–4, 215 Cwmbrwyn see Laugharne Dalton Parlours see Collingham Danebury (Hants.) 227, 228 Darenth (Kent) 37, 50, 153, 156, 251, 291 Davenescourt (Somme) 155, 162, 164 Deanshanger (Northants.) 50 Démuin 132, 133 Denton (Lincs.) 36, 37, 38, 263 Deutschkreutz (Aus.) 211, 213 Deva (Rom.) 201, 282, 283 Dicket Mead see Welwyn Dietfurt a.d. Altmuhl (Bay.) 225 Dietikon 299, 300 Dirlewang (Bay.) 137, 138–9 Ditchley (Oxon.) 47, 50, 79, 128, 185, 250, 253 Doische-Vodelée (Belg.) 34, 137, 138 Downton (Wilts.) 47, 49–50, 54, 55, 57, 59, 65, 67, 79, 102, 109, 127, 132, 150, 152, 202, 261, 267, 288, 289, 290 Drac?evica (Yugosl.) 202, 203, 204 Dragonby (Lincs.) 27, 241, 277 Draughton (Northants.) 230 Drax (Yorks.) 85, 166, 288 Dreieich-Götzenhain (Hessen) 34 Druten-Klepperhei (Neth.) 11, 251, 252 Dura-Europos-palace of Dux Ripae 177 Dury (Somme) 102, 103, 288 East Dean-Holbury (Hants.) 41 East Grimstead (Wilts.) 37, 38, 263 Eaton-by-Tarporley (Cheshire) 56, 134–5


— Site Index — Eccles see Aylesford Eching-Autobahn 221, 222, 225, 228–9 Eching-Neufahrn (Bay.) 221 Echternach (Lux.) 19, 183, 190, 278, 284 Eckartsbrunn (Bay.) 96, 100 L’Ecluse-Leckbosch (Belg.) 66, 79 Ecoust-Saint-Mien (Picardy) 152, 153, 299 Edwinsford 16 Egregy (Hung.) 204 Ehingen-am-Ries (Bav.) 253 Eisenstadt (Aus.) 183 Elchweiler see Birkenfeld Ellesborough-Terrick (Bucks.) 51, 60 Ersigen (Switz.) 124 Estrées-sur-Noye (Somme) 159, 162, 299, 300 L’Etoile (Somme) 103, 277 Euskirchen-Kreuzweingarten(Rhld-Pf.) 190, 278 Evelette (Belg.) 84 Ewhurst-Rapsley (Surrey) 55, 145, 149, 284, 289 Exning-Landwade (Suffolk) 36, 38, 263 Ezinge (Neth.) 30

Folkestone (Kent) 72, 163 Fontaine-le-Sec (Somme) 152, 288 Fontoy-Moderwiese (Lorraine) 71, 72 Francolise-San Rocco (Italy) 58, 257, 258, 261 Frankfurt-Bergen-Enkheim (Hessen) 35, 140 Frankfurt-Bornheim (Hessen) 111, 112, 140 Friedberg-Fladerlach 259, 262, 288 Friedberg-Pfingstweide 89, 147, 150–1, 159, 211 Friedrichsdorf-Seulberg (Hessen) 125 Frilford (Berks.) 107, 109, 111 Frocester Court (Glos.) 75, 76, 82, 92, 158, 185, 201, 258, 259, 261, 262, 288 Fromentières (Mayenne) 63 Furschweiler (Saarld) 119

Famechon (Somme) 291, 293 Farmington-Clear Cupboard (Glos.) 28, 31–2, 33, 34, 45, 119, 249, 261 Farningham-Manor House (Kent) 54–5, 77, 91–2, 131, 289, 290, 301 Faversham (Kent) 50, 51, 60, 70, 79, 241, 299 Feddersen Wierde (NiederSachsen) 30 Ferpicloz (Switz.) 65, 141 Fischamend-Katherinenhof (Aus.) 205 Fishbourne Palace (Sussex) 140, 174, 176, 177, 182, 194, 278 Fishtoft (Lincs.) 39, 221 Fitten (Saarld) 79 Fliessem-Odrang (Rhld-Pf.) 7, 125, 140, 161, 195, 196, 197, 198, 262, 284, 292, 293 Flocques (Somme) 142 Flumenthal (Switz.) 288

Gadebridge Park see Hemel Hempstead Gambach see Munzenberg Garden Hill see Hartfield Gargrave-Kirk Sink 48, 50, 52, 55, 56, 60 106, 109 Gayton Thorpe (Norfolk) 82, 119, 137, 139, 282, 290, 301 Geislingen-Heidegger Hof (Bad.-Württ.) 113, 114, 284 Gentelles (Somme) 149 Gerpinnes 83 Goodrich-Huntsham (Herefs.) 56, 261 Gorhambury see St Michael Gornea (Rom.) 215–16 Graux (Belg.) 115, 116, 259, 261, 288 Graz-Thalerhof (Aus.) 191, 192 Great Staughton (Cambs.) 35, 63, 118, 141–2 Great Weldon (Northants.) 140, 254, 267, 270–1 Great Witcombe 139, 166 Grémecey (Moselle) 31, 32, 35 Grenchen (Switz.) 81, 82, 88, 121 Grivesnes (Somme) 149 GrossSachsenheim see Sachsenheim


— Site Index — Guadaira-Alcala/Casa de Pelay Correa (Seville) 128 La Guerche-Le Grand Chausseroi (Cher) 62 Gunthersburg Park see FrankfurtBornheim Gunzenheim-Staatsforst-Sulz 107, 109 Gyulafirátót-Pogánytelek (Hung.) 199, 200 Habay-Rulles (Belg.) 138 Haccourt (Belg.) 70, 71, 75, 107, 173, 192, 194, 195, 196–7, 292 Hales (Staffs.) 72 Halstock (Dorset) 131, 167, 170 Ham Hill see Montacute Hambach see Niederzier Hamblain-les-Prés (Pan-de-Calais) 72, 241, 244 Hambleden-Yewdon Manor (Bucks.) 153, 156–7 Haps (Neth.) 221, 222, 223, 224, 225 Harbonnières (Somme) 103, 105 Harburg-GrossSorheim 100 Harding-Mingies Ditch (Oxon.) 230 Hartfield-Garden Hill (Sussex) 11, 18, 231, 237, 239, 251 Hartlip (Kent) 60, 89 Haunstetten (Bay.) 221, 222, 225 Haut Clocher-Saint Ulrich (Moselle) 7, 197-8, 284, 292 Hechingen-Stein (Bad.-Württ.) 98, 100 Heerlen-Boventse Caumer (Neth.) 43, 44 Heidelsheim (Bad.-Württ.) 39 Hemel Hempstead-Boxmoor (Herts.) 50, 248, 250, 282, 300 Hemel Hempstead-Gadebridge Park (Herts.) 53, 60, 82, 136, 241, 244 Heppenheim (Hessen) 34, 35, 81 Hérouville-Lébisey (Calvados) 51, 53, 54, 60, 62, 89 Herschweiler-Pettersheim 136, 137 Herten-Warmbach see Rheinfelden Heuneberg (Bad.-Württ.) 222, 226, 228 High Wycombe (Bucks.) 58-9, 72, 84

Hirschberg-GrossSachsen (Bad.-Württ.) 42 Hobita-Gradiste (Rom.) 204, 209, 213 Hoheneck see Ludwigsburg Hohenfels-Im Keller (Rhld-Pf.) 112, 113, 115, 142 Holbury see East Dean Holcombe see Uplyme Holme House see Manfield Hölstein (Switz.) 38, 40, 41, 93, 249 Hombleux (Somme) 70 Hoogeloon 221, 224, 225 Horath 139 Hornchurch 227, 228 Hosszuhetény (Hung.) 125, 215 Houdeng-Goegnies (Neth.) 128 Houthem-Vogelsang (Neth.) 44, 45, 291 Hucclecote (Glos.) 102 Hüfingen (Bad.-Württ.) 122, 125 Hummetroth-Haselburg (Hessen) 64, 75, 140, 204 Hungen-Bellersheim (Hessen) 161, 171, 213 Huntsham see Goodrich Hüssingen see Westheim Hüttwilen see Stutheim Inzigkofen (Bad.-Württ.) 41, 86, 87, 88, 94, 97, 98, 100 Irsch see Vierherrenborn Iskar-Gara (Bulg.) 210, 212 Isola d’Istria (Yugosl.) 204, 209, 210 Izel (Belg.) 104, 105 Izenore see Perignat Jemelle-Neufchâteau 64, 114, 170 Jublains-La Boissière (Mayenne) 241, 244 Jumilla-Los Cipreses (Murcia) 183, 184 Kaalheide-Krichelberg (Neth.) 229 Kadin Most (Bulg.) 214 Kaisersteinbruch-Königshof (Aus.) 93, 201, 204, 205, 208, 210, 245, 249, 21340 Katzenbach 156, 185, 255 370

— Site Index — Kempten-Loja Kapelle (Bay.) 100, 101, 103 Kernen-Rommelshausen (Bad.-Württ.) 107, 109 Keszthely-Fenékpuszta (Hung.) 202–3, 207, 208, 215 Kethel (Neth.) 229 Kinheim (Rhld-Pf.) 69, 85, 90–1, 92, 93, 112, 262, 263, 269 Kirchberg see Küttigen Kirkel-Forst (Saarld) 100 Kleinsteinhausen (Hessen) 139 Koerich-Goeblingen (Lux.) 24, 25, 85, 104, 106, 110, 148, 201, 285, 286 Köln-Braunsfeld 27 Köln-Mungersdorf (Nordrh.-Westf.) 8, 68, 73–4, 77, 132, 136, 142, 143, 253, 262, 271, 272, 273, 274, 279, 282, 293, 295 Köln-Praetorium 190 Köngen (Bad.-Württ.) 80, 116, 185, 285, 287 Königsbrunn (Bay.) 222, 226 Königshof see Kaisersteinbruch Konken (Rhld-Pf.) 27, 31 Konska (Bulg.) 208, 210, 212 Konz-Lummelweis (Rhld-Pf.) 24 Kralev Dol (Bulg.) 210, 212 Kran (Bulg.) 202 Kulm (Switz.) 77–8 Küttigen-Kirchberg (Switz.) 65, 66, 67, 68, 70, 78, 79, 84, 85, 87, 134, 141, 288 Lalonquette (Pyr.-Atl.) 31, 183, 184–6, 187, 188, 189, 191 Lamargelle-Versingue (Côte d’Or) 47, 48, 50, 62, 79, 102, 105, 159 Landen-Betzveld (Belg.) 67, 68 Landshut-Sallmansberg (Bay.) 227, 228, 253 Langenau-Osterstetten(Bad.-Württ.) 126 Langton Dwelling House (Yorks) 28, 33, 57, 85, 115, 262, 282 Laperrière-sur-Saône (Côte d’Or) 44, 282

Latimer-Chenies (Bucks.) 72 Laufen-Müschag (Switz.) 66, 67–8, 75, 124, 252 Laufenburg (Bad.-Württ.) 97, 134, 135 Lauffen am Neckar (Bad.-Württ.) 119–20, 125, 150, 171, 206 Laugharne-Cwmbrwyn (Dyfed formerly Pembs.) 28, 33, 242 Lavans-les-Dole (Jura) 29 Lazenay-Les Sales (Cher) 62 Lechlade-Claydon Pike (Oxon.) 36 Leiwen-Bohnengarten (Rhld-Pf.) 89, 253, 279, 280 Lemiers (Neth.) 138 Lendin (Seine-Mar.) 108, 109, 110 Leutersdorf (Rhld-Pf.) 86, 158, 167, 169, 262, 273 Levroux-Trégonce (Indre) 162, 298, 299 Liédena (Navarra) 41, 295–6, 297, 299 Liestal-Munzach (Switz.) 40, 71, 76, 77, 158, 159, 162, 297, 299 Limerlé (Belg.) 83 Lisiici (Yugosl.) 201, 202, 210, 213 Little Milton (Oxon.) 50, 241 Littleton see Compton Dando Ljubsko-Proboj (Yugosl.) 206, 207 Ljuina (Yugosl.) 183, 211, 214 Llanfrothen, Park 16 Lockleys see Welwyn Lodo see Arradon Loja see Kempten Lörrach-Brombach 42, 99 Ludwigsburg-Hoheneck (Bad.-Württ.) 97, 145, 148–9, 150, 251 Ludwigsburg-Pflugfelden (Bad.-Württ.) 24, 87, 120 Lussas-et-Nontronneau 61, 63, 64, 204 Madara (Bulg.) 215 Maidenhatch see Pangbourne Maidenhead-Cox Green (Berks.) 34, 259 261 Maidstone-Loose Road (Kent) 50 Mailhac-Cayla (Aude) 227 Maillen-Al Sauvenière (Belg.) 170, 208 371

— Site Index — Maillen-Ronchinne (Belg.) 114, 170–1, 211 Mainz-Kastel, ‘Aedicula’ 55, 131, 133 Majdan (Yugosl.) 200, 203 Malguenac-Guilly (Morbihan) 171 Mali Mounj (Yugosl.) 201 Mamer-Gaschtbierg (Lux.) 27, 31 Manching 220 Manderscheid 81, 88, 124, 136, 137, 138, 139, 152, 214 Manerau 204, 207, 210 Manfield-Holme House (Durham) 132, 247, 254, 255 Mansfield Woodhouse (Notts.) 37, 38, 50, 51, 152, 153, 159, 206 Marboué-Mienne (Eure-et-Loir) 163, 165 Marchelepot (Somme) 126, 160 Mareuil-Caubert (Somme) 131 Maria Ellend-Ellender Weingarten (Aus.) 204, 205 Marquivilliers (Somme) 139 Marshfield (Glos.) 74, 110, 149, 159, 203, 209, 284, 289 Marshfield-Ironmonger’s Piece (Avon) 143, 240, 241, 243, 245, 249 Martigues-L’Arquet (Bouches-du-Rhone) 227 Maulévrier 17, 41, 53, 59, 62, 72, 136, 152, 159, 163, 214, 282, 284 Mauren (Bay.) 84, 86–7, 200, 201 Mautern a. Donau (Aus.) 202, 208 Mayen-Allenz (Rhld-Pf.) 8–9, 26–7, 29–33, 41, 46, 69, 83, 93, 95, 99–100, 118, 152, 154, 228, 233, 250, 253, 262, 269, 274, 291–2, 301 Mayen-Im Brasil (Rhld-Pf.) 23, 24 Meckel (Rhld-Pf.) 159, 162, 293, 300 Mehring 35, 258, 262, 284 Merdingen (Bad.-Württ.) 139 Le Mesge (Picardy) 152, 154, 156 Les Mesnuls (Yvelines) 77 Messkirch-Alstadt (Bad.-Württ.) 96, 99, 146 Mettet-Bauselenne (Belg.) 209, 215 Metz-Sablon (Moselle) 291

Mézières-en-Santerre/La Croix SaintJacques (Somme) 112, 113, 152 Mézières-en-Santerre/Le Ziep (Somme) 143, 159, 161, 299 Michelstadt-Steinbach (Hessen) 35 Mikhailovgrad (Bulg.) 200 Mileoak see Towcester Milton Keynes-Bancroft (Bucks.) 206, 241, 244, 255, 293 Mogilets (Bulg.) 214 Monchy-Humières 131 Moncrabeau-Bapteste (Lot-et-Garonne) 182 Monreal (Rhld-Pf.) 104, 106 Mons-en-Chausée (Picardy) 202 Montacute-Ham Hill (Som.) 148, 171, 208 Montmaurin (Haute Garonne) 188–90, 191, 193 Montrozier-Argentelle (Aveyon) 183, 193, 194–5, 215 Morthommiers-Le Crot (Cher) 62 München-Denning (Bay.) 239 Mundelsheim (Bad.-Württ.) 81, 88, 119 Munzenberg-Gambach (Hessen) 109, 110, 111, 116, 139, 203, 284, 292 Neckarrems (Bad.-Württ.) 44, 299 Neckarzimmern-Stockbronner Hof (Bad.-Württ.) 83, 98, 119 Neerharen-Rekem (Belg.) 40 Nemesvámos-Balácpuszta (Hung.) 181–2, 183, 213 Nennig (Rhld-Pf.) 7, 18, 140, 173, 190, 194, 292 Neuburg a.d. Donau (Bay.) 112, 113, 115 Neuhausen auf den Fildern-Horb (Bad.-Württ.) 106 Neumagen-Dhron-Papiermühle (RhldPf.) 29, 119, 138, 248, 252–3 Neuss 29 Newel (Rhld-Pf.) 77, 84, 85, 92, 93, 142, 171, 263, 271, 273, 288 Newport (I.o.W.) 47–8, 49, 58, 61, 62,


— Site Index — 63, 67, 70, 85, 105, 132, 156, 195, 203, 214. 265, 268, 289 Newton St Loe (Wilts.) 51, 128 Niedereschach-Fischbach 96, 101, 102, 103, 201 Niederzier-Hambach (Nordrh.-West.) 83, 152, 153, 158, 159, 229, 284, 291 North Cerney-The Ditches (Glos.) 50, 53 North Leigh (Oxon.) 155, 163, 167, 180, 182, 214, 289, 299 North Leigh-Shakenoak (Oxon.) 70, 71, 72, 77, 301 North Stainley-Castle Dykes 28, 32 North Wraxall (Wilts.) 72, 162, 165 Northchurch 109, 132 Norton Disney (Lincs.) 37, 159, 209, 263, 293, 295 Novi Saher (Yugosl.) 209, 210 Noyers-sur-Serein (Yonne) 105, 106, 253, 254, 255, 279, 280, 282, 297, 299 Nünschweiler (Rhld-Pf.) 44, 299 Nuth-Vaasrade (Neth.) 36, 84, 88 Ober-Gambach see Bruchsal Ober-Ramstadt-Pfingstweide (Hessen) 106 Oberentfelden (Switz.) 162, 291, 292, 298, 299, 300 Oberweis (Rhld-Pf.) 7, 14, 18, 90, 161, 292, 293 Ödheim 288 Odiham-Lodge Farm (Hants.) 37 Odrang see Fliessem Olfermont (Haut-Rhin) 78, 100, 147, 151 Oplontis 10 Orazio (Italy) 74 Orbe (Switz.) 215 Orlandovtsi (Bulg.) 125, 183, 211, 214, 215 Ormalingen (Switz.) 54, 55–6, 59, 87, 131 Orton Longueville-Orton Hall Farm 167, 168

Öschelbronn (Bad.-Württ.) 252 Oss-Ussen (Neth.) 221, 222, 223, 224, 226 Osterfingen (Switz.) 121, 248, 252 Otford (Kent) 261 Overasselt (Neth.) 34, 35 Overstone (Northants.) 255 Ovillers (Somme) 105, 141, 142, 201, 202 Pangbourne-Maidenhatch (Berks.) 50 Panik 211, 214–15 Park Street see St Stephen Parndorf see Bruckneudorf Peiting (Bavaria) 140 Penshurst Place (Kent) 23 Perignat-Izernore (Ain.) 31 Petersfield-Stroud (Glos.) 37, 38, 263 Pforzheim-Hagenschiess 285, 286 Piddington (Northants.) 72, 301 Pilsdon Pen (Dorset) 221, 230, 231 Pirton, Walnut Tree Farm 67, 123 Pitney (Som.) 162 Plachy-Buyon-Les Trois Cornets (Somme) 132, 288 Pleven-Kailuka (Bulg.) 214 Plomelin-Perennou (Fin.) 166, 168 Plouventer-Kerilien (Fin.) 102, 103, 105 Port-le-Grand (Picardy) 247, 250 Primelles-Champ Chiron (Cher) 61, 62 Proboj see Ljubsko Puig de Cebolla-El Vilar (Valencia) 71, 75 Pulborough (Sussex) 140, 156, 168 Pully-Prieuré (Switz.) 191 Quarteira (Faro) 184 Quinton (Northants.) 34, 35, 79, 87, 126 Radley-Barton Court Farm 139, 231, 241, 246, 248–9, 261 Rainecourt (Somme) 126, 127 Rankovici see Travnik Rapsley see Ewhurst


— Site Index — Ravensbosch see Schimmert op den Bilik Raversbeuren (Rhld-Pf.) 90, 93, 123, 140 Reach (Cambs.) 55 Redlands Farm see Stanwick Regelsbrunn (Aus.) 206, 213 Regensburg-Burgweinting (Bay.) 125, 144, 145, 148, 149, 150, 251, 285 Reimlingen (Bay.) 101 Remmingsheim (Bad.-Württ.) 8, 96, 99 Rheinbach-Flerzheim (Nordrh-Westf) 84, 88 Rheinfelden-Herten/Warmbach (Bad.-Württ.) 103, 105 Rheinfelden-Salzbrünnele (Bad.-Württ.) 143 Ribemont-sur-Ancre 143, 159, 161 Ridgewell (Essex) 55, 70 Rielves 179, 180 Rijswijk-de Bult (Neth.) 36, 37, 95, 115, 220, 226–7, 229–30, 232, 233–4, 235, 238, 239, 240, 243, 249, 250, 251 Ringstead (Northants.) 247, 255 Rivenhall 19 La Roche-Maurice-Valy Cloistre (Fin.) 71, 74–5, 87 Rockbourne (Hants.) 51, 139, 154, 157, 158, 167, 299 Rodmarton 83 Rognée (Belg.) 163 Rohrbach (Aus.) 201 Rome-Domus Augustana (Flavian Palace) 173–4, 176, 177, 178, 179, 182 Rome-Via Gabina 41 Romegoux-La Vergnée (Charente-Mar.) 50, 53, 54, 59, 110, 136, 142, 299 Rommelshausen see Kernen La Roquebrousanne (Var.) 79 Rothselberg (Rhld-Pf.) 34, 35, 79, 120–1, 126, 140 Rudston (Yorks.) 39, 221, 244, 249–50 Saaraltdorf (Moselle) 24, 25, 39 Sachsenheim-GrossSachsenheim (Bad.-Württ.) 100, 138

St Lythans-Whitton (Glam.) 55, 56, 57, 140, 234, 236, 238, 240, 243, 248 St Michael-Gorhambury (Herts.) 55, 226, 230, 231, 277, 282 St Stephens-Park Street 51, 61–2, 69, 70, 263, 267, 270, 271 Saint-Acheul (Somme) 74 Saint-Aubin-le-Mazaret 40, 41 Saint-Aubin-sur-Mer (Pas-de-Calais) 44, 110, 124 Saint-Germain-lès-Corbeil 167, 253, 287 Saint-Herblain (Loire-Atl.) 253 Saint-Julien et Aubigny (Somme) 78 Saint-Pierre-la-Garenne (Eure) 31 Saint-Ulrich see Haut Clocher Sainte-Solange (Cher) 70 Salzburg-Liefering (Aus.) 214 Santa Colomba de Somoza-Maragatera (León) 183, 184 Sarajevo-Stup (Yugosl.) 204, 205 Sarica (Rom.) 204, 205 Sarmentsdorf (Switz.) 78, 79 Sarmizegetusa (Rom.) 19, 202, 203 Sauvenière (Belg.) 83 Schaanwald (Liechtenstein) 283 Schambach 87, 98 Schimmert-op den Billich (Neth.) 33 Schleitheim 260, 262 Schuld 91, 93 Schupfart-Betberg (Switz.) 65, 77, 84, 85, 92, 271, 272, 274, 288 Selongey (Côte d’Or) 72 Serville (Belg.) 8, 24, 25, 26, 29, 31, 32, 36, 37, 87, 118, 120, 121, 126, 127, 138 Sette Finestre (Italy) 215 Shakenoak see North Leigh Shipham-Star (Som.) 127, 132, 133, 171 Sigean-Pech Maho (Aude) 230, 231 Sigmaringen-Laiz (Bad.-Württ.) 96, 98– 9 Sigmaringen-Steinäcker (Bad.-Württ.) 43 Silchester 229 Sinsheim (Bad.-Württ.) 17, 43, 136, 141, 284


— Site Index — Skeleton Green see Braughing Smarje-Grobelce (Hung.) 107 Sofia-Obelja (Bulg.) 214 Somerton-Bradley Hill (Som.) 26–7 Somerton-Catsgore 28, 33, 34, 107, 116 Sontheim an der Brenz (Bad.-Württ.) 102, 105, 106, 112, 142, 285, 287 Sotzweiler see Tholey Southwick (Sussex) 278 Sparsholt (Hants.) 49, 59, 63, 127, 152, 157, 167, 183, 209, 246, 250 Spoonley Wood see Sudeley Stadtbergen (Bay.) 71, 72, 77, 125 Stahl (Rhld-Pf.) 8, 24, 25, 29, 32, 34, 41, 44, 94, 97, 99, 118, 122, 123–4, 125, 134, 138, 139, 199 Stammheim see Stuttgart Stanwick-Redlands Farm (Northants.) 128 Starzach-Bierlingen (Bad.-Württ.) 94 Stein (Neth.) 34 Steinbach im Odenwald see Michelstadt Stockbronner Hof see Neckarzimmern Stolac (Yugosl.) 201, 205–7, 213 Stowey Sutton-Chew Park (Avon, formerly Som.) 28, 32–3, 34 Stratford-upon-Avon-Tiddington (War.) 31, 126 Stroud see Petersfield Stutheim-Huttwilen (Switz.) 253 Stuttgart-Stammheim 86, 87, 88, 98, 100–1, 107 Sudeley-Spoonley Wood 11, 37, 114, 116, 162, 163, 165, 212, 262, 263, 266, 268–70, 301 Sudeley-Wadfield 11 Suippes (Marne) 229 Sümeg (Hung.) 202 Swindon-Okus (Wilts.) 31 Szentkiralyszabadja-Romkut (Hung.) 115, 199, 213 Szilágy-Arnyoldal (Hung.) 213 Tac-Fövenypuszta (Hung.) 215 Tarrant Hinton (Dorset) 251 Tavaux (Jura) 200

Telita (Rom.) 200 Tengen-Büsslingen (Bad.-Württ.) 99 Terrick see Ellesborough Téting-sur-Nied (Moselle) 191–2 Thalerhof see Graz Tholey-Sotzweiler (Saarld) 84, 282, 283 Thoraise (Doubs.) 221 Thuy (Eure-et-Loir) 190 Tiddington see Stratford-upon-Avon Tiefenbach 25, 29 Titelberg, ‘Aedicula’ 121–2, 126 Titsey 83 Tollard Royal (Dorset) 230 Towcester-Mileoak (Northants.) 111 Travnik Rankovicí (Yugosl.) 202, 203 Treuchtlingen-Weinbergshof 100, 291 Trier-‘Basilica’ 161, 171, 172, 177, 257, 273 Trouey (Cher) 79, 240 Upchurch-Boxted (Kent) 55, 72 Uplyme-Holcombe 34, 240, 242, 245 Val Catena see Brioni Grande Valladolid-Prado 184 Vallenay-Patureau Fourneau (Cher) 62–3 Vaux-sur-Somme-Bosquet Duval 103 Verberie (Oise) 222, 226 Verneuil-en-Halatte-Le Bufosse (Oise) 39, 167, 289 Vesqueville (Belg.) 43–4, 134 Vicques (Switz.) 67, 68, 84, 85, 139 Vierherrenborn-Irsch (Rhld-Pf.) 106, 253, 254, 255, 285, 287 Vieux-Rouen-sur-Bresle (Seine-Marne) 279, 300 Vila de Frades (Portugal) 128 Villeneuve-St Germain (Aisne) 230, 231, 232 Villeneuve-sur-Cher, Les Augerets (Cher) 37 Villers-sous-Ailly (Picardy) 141 Villers-Bretonneux (Somme) 107, 108, 110–11, 116, 263 Vix (Côte d’Or) 221


— Site Index — Voerendaal-Ten Hove (Neth.) 43, 44, 211, 289 Voerendaal-Ubachsberg (Neth.) 34, 118 Vouneuil-sous-Biard-Les Cassons 70, 103 Wachenheim (Nordrh.-Westf.) 288 Wadfield see Sudeley Wahlen (Switz.) 81, 82, 88, 142, 201 Walsbetz (Belg.) 109, 115, 234 Walton-on-the-Hill-The Heath 50, 255 Wancourt (Somme) 102, 143, 201, 261 Warfusée-Nord (-Abancourt) 11, 159, 161, 282, 299 Warfusée-Sud (-Abancourt) 11, 159, 163, 299 Watergate see West Marden Weitersbach 81, 82, 119, 185, 190, 260, 262, 263, 284, 288 Wellow (Som.) 56, 57, 70, 78 Welwyn-Dicket Mead (Herts.) 11, 150 Welwyn-Lockleys (Herts.) 11, 19, 51, 60, 109, 132, 226, 250 West Blatchington (Sussex) 37, 38, 263 West Dean-Chilgrove (Wilts.) 37, 60, 132 West Marden-Watergate (Sussex) 255

Westerhofen (Bavaria) 140 Westheim-Hüssingen (Bay.) 101 Weyeregg (Aus.) 208, 209 Weyhill-Clanville (Hants.) 37 Whittington (Glos.) 122, 125 Whitton see St Lythans Wickford (Essex) 229 Wiesbaden-Höfchen (Hessen) 152, 165 Wiesendangen-Steinegg (Switz.) 252 Winden am See (Aus.) 203, 207 Winkel-Seeb (Switz.) 19, 39, 40–1, 50, 59, 280, 282, 284, 291 Winterton (Lincs.) 36, 40, 166–7, 168, 178, 179, 263 Wittlich (Rhld-Pf.) 190–1 Wollersheim-Am Hostert (Rhld-Pf.) 40, 101 Woodchester (Glos.) 175, 182–3, 193, 195, 292, 301 Worsham see Asthall Wraxall (Som.) 28, 33, 140 The Wrekin (Salop) 227 Zaloje see Bihac Zijderveld (Neth.) 222, 223 Zofingen (Switz.) 68, 73



Agache, Roger 9, 11, 102, 161 Anthes, E. 7 Applebaum, S. 20, 82, 83, 288-9, 302 Baatz, D. 94-5 Baxter, Richard 4-5 Biro, M. 9 Black, E.W. 77, 279 de Boë, G. 224 Böttger, B. 292 Boon, George 241 Branigan, K. 8 Castro, Maria Cruz Fernando 9, 179, 188 de Caumont, Arcisse 6 Claudius 278, 279 Collingwood, R.G. 8, 251 Collis, J.R. 220 Cremonik, I. 202 Drack, W. 9, 46, 65, 68, 79 Drury, P.J. 14, 77 Fellmann, R. 93 Ferdière, A. 59 Fremersdorf, F. 8, 73, 273 Friendship-Taylor, R.M. and D.E. 35 Gaubatz-Sattler, A. 95, 143

Gorges, J.-G. 9 Grenier, Albert 8, 9, 59, 191, 198 Hale, Sir Matthew 4 Harding, D.W. 254 Harmand, J. 8 Haverfield, F.J. 7 Hawkes, C.F.C. 8 Hettner, Felix 7 Jones, Glanville R. 276 Klein, M. 148, 149 Koepp, F. 117, 292 Koethe, H. 14 Kolling, A. 69 Kropatscheck, G. 7 McKay, A.G. 3 de Maeyer, R. 8, 83, 118 Mansuelli, G.A. 9 Medawar, Peter 275 Mylius, H. 9, 19, 30, 194, 264, 291 Naidenova, V. 212 Nash-Williams, V.E. 8 Neuffer, Eduard 105 Nikolov, D. 9 North, Roger 4-5 377

— Index of Authors — Oelmann, Franz 8, 9, 10, 13, 14, 23, 24 27, 30, 41, 42, 73, 83, 84, 94, 95, 199 264, 301 Paret, O. 8, 81, 99, 148 Percival, John 3, 10 Reece, Richard 20 Reim, H. 41-2, 94, 97 Reutti, F. 77 Richmond, Sir I. 8, 9 Rostovtzeff, M.I. “et al” 10 Samesreuther, E. 8 Schumacher, K. 7 Sontheimer, L. 83 Stead, I.M. 249 Steiner, P. 30, 167

Stevens, C.E. 300, 301 Stone, Lawrence 57 Swoboda, Karl 7, 8, 9, 13, 14, 24, 117, 125, 127, 138, 143 Thomas, B. 208, 249 Thomas, E. 9 Truhelka, C. 205 Vasie, M. 9 Veblen, Thorstein 193 Wagner, H.O. 101, 103 Ward, J. 7, 8 Ward-Perkins, J.B. 19 Webster, Graham 3, 9 aus’m Weerth, E. 138 Wichmann, K. 198 Wightman, Edith 10-11, 190, 195