Universals, Concepts and Qualities - New Essays on the Meaning of Predicates

  • 60 263 9
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up

Universals, Concepts and Qualities - New Essays on the Meaning of Predicates

Universals, Concepts and Qualities New Essays on the Meaning of Predicates Edited by EF. STRAWSON and ARINDAM CHAKRAB

875 106 9MB

Pages 324 Page size 600 x 931 pts Year 2012

Report DMCA / Copyright


Recommend Papers

File loading please wait...
Citation preview

Universals, Concepts and Qualities New Essays on the Meaning of Predicates

Edited by EF.




© The contributors, 2006

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise without the prior permission of the publisher. P.P. Strawson and Arindam Chakrabarti have asserted their right under the Copyright, Designs and Patents Act, 1 9 8 8 , to be identified as Editors of this Work. Published by Ashgate Publishing Limited Gower House Croft Road Aldershot . Hants GU l l 3HR England

Ashgate Publishing Company Suite 420 1 0 1 Cherry Street Burlington VT 0540 1 -4405 USA

I Ashgate website : http : //www. ashgate.com I British Library Cataloguing in Publication Data Universals, concepts, and qualities : new essays on the meaning of predicates 1 .Universals (Philosophy) LStrawson, P.P' II. Chakrabarti, Arindam 1 1 1 .2 Library of Congress Cataloging-in-Publication Data Universals, concepts, and qualities : new essays on the meaning of predicates / edited by P.P. Strawson and Arindam Chakrabarti. p. cm. Includes bibliographical references and index. ISBN 0-7546-5032-4 (hardback : alk. piper) 1 . Universals (Philosophy) 2. Individuation (Philosophy) 1. Strawson, P.P' II . Chakrabarti, Arindam.

B I 05 . USU5 5 2005 1 1 1 ' . 2-dc22 200500255 0 ISBN 0 7546 5032 4

Typeset by Manton Typesetters, Louth, Lincolnshire, UK. Printed and bound i n Great Britain b y T J International Ltd, Padstow, Cornwall.


List of Contributors





Introduction Arindam Chakrabarti


Strawson on Universals Pranab Kumar Sen



Reply to Pranab Sen P.F. Strawson



Universals and Other Generalities lonardon Ganeri



Predicates and Properties : An Examination of P.K. Sen's Theory of Universals Fraser MacBride


Buddhist Nominalism and Desert Ornithology Mark Siderits


Universals Transformed: The First Thousand Years After Plato Richard Sorabji



Conceptualism Chris Swoyer

1 27


The Concept Horse Harold W Noonan



Universals and Particulars : Ramsey's Scepticism Bob Hale



How Not to Trivialize the Identity of Indiscernibles Gonzalo Rodriguez-Pereyra








Universals and the Defence o f Ante Rem Realism George Bealer

2 25


Particulars Have Their Properties o f Necessity David Annstrong



Properties i n Abundance Wo lfgang Kiinne


A Category o f Particulars



P.R Strawson


On Perceiving Properties Arindam Chakrabarti




List of Contributors

David Armstrong, Professor Emeritus of Philosophy at the University of Sydney, Australia, is best known for his work on a materialist theory of the mind and has published books and articles on analytic metaphysics and epistemology, especially on the problem of universals . George Bealer, Professor o f Philosophy at Yale University, USA, works i n meta­ physics, epistemology, philosophy of language, and philosophy of mind. Arindam Chakrabarti, Professor of Philosophy at the University of Hawaii at Manoa, USA, teaches and writes about philosophy of language, metaphysics, Indian analytical philos ophy, and philosophy and the emotions . Jonardon Ganeri, Reader in Philosophy at the University of Liverpool, UK, has' published books and papers on Indian philosophy of language, philosophy of mathematics, conceptions of rationality, Indian and B uddist logic, and Indian ethics . Bob Hale, Professor o f Philosophy at the University o f Sheffield, UK, works mainly on topics in the philosophies of mathematics, logic and language, with occasional forays into metaphysics and meta-ethics . Wolfgang Kunne, Professor of Philosophy at the University of Hamburg, Germany, has recent publications that include a book on conceptions of truth and papers OR philosophy of language and metaphysics . Fraser MacBride, Reader in Philosophy at Birkbeck College London, UK, has published numerous articles on metaphysics, the philosophy of mathematics and the history of analytic philosophy. Harold W. Noonan, Professor of Philosophy at the University of Nottingham, UK, teaches and writes on metaphysics, philosophy of language, philosophical logic, and history of philosophy. Gonzalo Rodriguez-Pereyra, Professor of Philosophy at Universidad Torcuato Di Tella, B uenos Aires, Argentina, and the University of Nottingham, UK, teaches and writes on metaphysics. t Pranab Kumar Sen, formerly Emeritus Professor of Philosophy at Jadavpur University, Calcutta, India, taught and wrote on logic, epistemology, philosophy of language, and ethics. He is acknowledged, for his books and published papers, in



India, Europe and America as one of the most meticulous teachers of the philosophies of Kant, Russell, Quine, Strawson, Davidson and Dummett. He was a close friend of the last three and G.H. Von Wright. Mark Siderits, Professor of Philosophy at Illinois State University, USA, teaches and writes in the areas of dissical Indian and B uddhist philosophy, philosophy of language, and analytic metaphysics . Richard Sorabji, Emeritus Professor of Philosophy at King' s College London and Fellow of Wolfson College Oxford, UK, works in the field of ancient philosophy. t P.F. Strawson, formerly Emeritus Professor of Philosophy at Magdalen College and University College, Oxford, UK, taught and wrote on the problems of language and metaphysics. He is best known for his classic debates on referring with Russell and on truth with Austin, for his book Individuals: An Essay in Descrip tive Metaphysics, as well as for his commentary on Kant's ' Critique of Pure Reason ' . He is recognized as one of the most influential philosophers of the analytical tradition in the twentieth century. Chris S woyer, Professor of Philosophy at the University of Oklahoma, Texas , USA, teaches and writes about metaphysics, philosophy of psychology, philosophy of science, history of modern philosophy, and epistemology.


We, the editors, are grateful to Mrs Rama Sen for giving us the permission to publish her late husband Professor Pranab Kumar Sen's paper in this volume. Manidipa S en and Madhucchhanda S en helped us at different stages of this publication, as did Mrs Ann Strawson. Without the hard work of Ellen Fridland, graduate researcher at the Philosophy Department of the University of Hawaii at Manoa, standardizing the format and preparing an integrated electronic version of the entire volume would have been nearly impossible.

Chapter 1

Introduction Arindam


What sort of a thing is a philosophical problem? By what criterion of identity do we decide that this is the same problem as that? How do we tell whether Plato's problem of innate ideas , Descartes's problem of innate ideas, and Chomsky's problem of innate ideas are all one and the same or three distinct problems ? When an issue, such as the problem of informative identity-statements , is raised and clearly formulated for the first time by some philosopher, does that philosopher discover the problem or bring it into being? Is a philosophical problem ever solved? If and when it is solved, does it go out of existence or is it shown to have been non­ existent to begin with? These kinds of puzzles, of course, will be looked upon or looked down upon at best with a therapeutic attitude by a certain influential line of thinkers who compare philosophical problems to a fly's sense of being stuck in a fly-bottle. B ut, instead of pretending to be outside the bottle already, suppose that we allow ourselves to think ontologically in an unashamedly 'objectually quantifying' fashion about these peculiar stifling circumstances themselves, circumstances from which we need a release called 'the/a solution' . I think we would then find that for ages philosophers have fallen repeatedly into the same types of threats of self-contradiction, the same kinds of clashes of opposite intuitions concerning fundamental concepts . Identical problems have been formulated across cultures and times sometimes in similar and sometimes in radically different terms. Even if thes e are a special s ort of psychological disorder, there seems to be repeatable patterns to them. Suppose, after much critically fortified reflection, we come to the conclusion that such a recurrent philosophical problem is a kind of abstract entity, then the problem of universals would itself be an abstract entity though not quite a universal. Take the following traditional questions : 'Is wisdom something out there besides wise individuals such as Socrates and Buddha? Would there exist one and the same cowness over and above the set of numerous individual cows even if the word "cow" or its equivalents in any other language did not exist? Is there an essence called "numberhood" besides numbers such as one, two , and fifty-six million, inhering in all those numbers including those that no one yet has thought of? Do properties such as being acid or being oxygen have any causal role in bringing about changes in the physical world just as individual drops of acid or puffs of oxygen doT Although they are distinguishable queries , together they all appear to 'express' and 'constitute' the one abstract entity called 'the problem of universals'. One can immediately think of a couple of considerations against regarding a philosophical problem as a full-fledged objective universal. First, a philosophical


Universals, Concepts and Qualities

problem has a history and seems to change from age to age, whereas (timeless or eternal) universals are not supposed to undergo change. S econd, while properties are expected to have property-bearers or exemplifiers and universals are expected to have instances, usually we cannot make sense of a universal or property having many parts or members . Yet a philosophical problem often has parts, a number of sub-issues grouped under it. It is often a set of problems. And the problem of universals is no exception. In its contemporary form, the problem of universals can be subdivided into the following problems , each of which could be expressed, paradigmatic ally, in terms of a theoretical interrogative, as a problem ought to be. What is a universal? Is it a mistake to quantify over the predicate-position or to take our reifying talk about common characteristics of obj ects seriously? Do universals exist independently of our thought and language, even if we cannot offer clear-cut criteria of identity for them? Are there universals which do not ever, yet, or any longer have any particular instance existing in nature? Can we have a screening device which would reduce the number of objectively real universals that we admit into our ontology by selecting from the countless generalities somehow 'meant' by all sorts of predicates (including negative, conjunctive and disjunctive ones) in a language? Of the logico-linguistic distinction between subjects/singular terms and predicates/functional expressions and the ontological distinction between particulars and universals , which one is prior? Should propositions and sets be counted as universals? Can we reduce universals to general concepts that we, thinking beings, possess or construct to represent or cope with the external world rather than items in that world? Are all properties universals or are there particular · properties specific to their bearers? Can an individual be reduced to just the set of its particular and general properties or does it stand outside the set of properties which are attributes of it? Given the contingent fact that a particular exists at all, is it necessary or contingent that it has the properties (universal or particular) that it does? Do universals have any causal role to play in the spatio-temporal world? Are universals, or at least some of them, perceptible or are all universals merely intelligible objects of thought? This book collects fourteen contemporary philosophers' carefully reasoned attempts at answering some of these old questions. B ecause there have been maj or shifts with regard to which of these questions are taken to be more important and which less , and later philosophers have added newer questions couched in terms not available to an earlier generation of philosophers , the problem of universals can be said to have changed as well. Let us look closely at one such alleged change. Ancient Indian VaiSe�ika realists defined a universal as that which is eternal, one, and inherent in many entities . In response, ancient Buddhist nominalist flux-theorists questioned the very idea of anything eternal and found the idea of one-in-many unintelligible. In most traditional Western discussions as well, the problem of universals was classically understood as the problem of the one-over-many. Even realists start by recognizing that there is initially a shared sense of conceptual difficulty in explaining how numerically different particulars can all come across as being of the same ilk or manifest a 'manner or respect of being the same ' , especially when one is trying to stick to the tangible naturalistic particulars alone as the only entities around. How can there be such undeniable singleness running through such palpable plurality? Yet it has been argued at the beginning of this new millennium



that the problem of universals is now better formulated as the problem of many­ over-one rather than as the problem of one-over-many. How could this switch happen? In a paper)l1}!Jind (April 2000), one of our contributors (Gonzalo Rodriguez­ Pereyra) has shown how. Though facts and states of affairs (as Wittgenstein's Tractatus would have them) suffered much damaging criticism, a version of them is back through the now-popular metaphysics of truth-makers. In this new - not altogether happy - idiom of truth-makers, the traditional problem of universals is the problem of explaining how the same property F can be present as a common factor in different truth-makers for different sentences. How, in other words, can the same property at the same time be a constituent of the fact that A is F, the fact that B is F, and the fact that C is F, given that A, B and C are all distinct existents ? Now, why is this a problem, when hardly anyone mistakes the 'is' of predication to be the sign of identity? This is a problem only if we commit ourselves to F being a separate but common constituent in each of these facts which are distinct in virtue of being distinguishable truth-makers, respectively, of the sentences 'A is F' , 'B is F' and 'C is F' . If the sentence 'A is F' were made true simply by the existence of A, then the problem of universals would vanish because A, B and C would be respectively the truth-makers of those three sentences and there would be no problem-causing additional common constituent of all three truth-makers. But we c annot let the truth-maker of 'A is F' shrink to A alone, much as the Quinean Nominalist may wish to do so. If 'A is F' were made true by A alone, then, by the same token, 'A is G' and 'A is H' would also be made true by A. But the fact that A is F is plainly not the fact that A is G, and neither of them is reducible to the fact that A is H, which is to say that it is the multiplicity of F and G and H, in spite of the singleness of their common bearer A, which forces the problem of universals on us. Therefore the problem of universals (what is there inside the fact that A is F besides A?) boils down to the problem: why do we need many truth-makers for 'A is F' , 'A is G' and 'A is H' , over and above that one single particular constituent A? This is clearly the problem of many-over-one. Whether it is best viewed as the problem of one-over-many or as the problem of many-over-one, by looking at the average life of a philosophical problem in general and from the multi-millennial and multi-cultural career of the problem of universals in particular, one can safely proj ect that this problem may go out of fashion but is not likely to go out of existence. During the second to fifth decades of the twentieth century, Western analytic philosophers would have sneered at most of the issues that constitute the problem of universals as pseudo-problems or mere muddles of misunderstood words taken out of the contexts of their proper scientific or everyday use. But not any more. Metaphysics - even a priori metaphysics - is very much in fashion within analytic philosophy now. As this collection of new essays by fourteen contemporary analytical philosophers amply demonstrates, the problem of universals, threatened from time to time by oblivion but never by death, once again has come near the centre of the field. Pranab Kumar Sen begins our lead essay by giving Strawson a maj or share of the credit for bringing b ack metaphysics to the centre stage of analytic philosophy (after, first a Logical Positivist, and then a later Wittgensteinian, lull). Of course, Strawson's ontology was initially presented as Descriptive Metaphysics, aiming


Universals, Concepts and Qualities

only ' to lay bare the most general features of our conceptual structure' (Individuals, p. 1 ) . Strawson himself evinced a certain Kantian caution against making straight metaphysical claims about the way the world is . But reminding us of Strawson' s rejection of Kant' s transcendental idealism, S e n takes Strawson ' s work on the theory of universals as a contribution straightforwardly to the theory of what there is rather than as merely a description of how we are generally constrained to think of and talk about the world. (In his 'Reply' Strawson does not obj ect to this bolder take on his work.) Unceremoniously continuing the Indian philosophical tradition of making the inost original philo sophical contributions under the garb of 'commentary' , Sen gives us a series of original insights and arguments about universals, and their relation to language, logic, thinking and experience, through his rational reconstruction and gentle but firm criticism of Strawson's most mature views on universals. Sen ' s long essay falls into seven main parts . In the first part, he gathers together a comprehensive list of kinds of generalities that Strawson recognizes as universals, especially during the early stages of his evolving thinking on this matter. In the second part, he unearths two distinct necess ary conditions used by Strawson for counting something as a general or as a particular thing. B oth of these have to do with whether something can or cannot be referred to by a singular expression the unique reference of which is determined solely by the meaning of the words making up that expression. The traditional criterion that universals are, whereas particulars are not, instantiable and predicable of other things, while both of them can figure as subjects of predication, is also mentioned by Sen as something that Strawson uses at places. In the third part, Sen suggests some drastic modifications of Strawson' s list of universals by bringing charges of over-coverage and under­ coverage. He recommends clipping off feature-universals (introduced by mass­ terms) and propositions, and adding relations to the stock of universals. In the fourth part Sen raises some worries about Strawson's criteria for particularhood and Strawson's somewhat complicated doctrine about instantiation, exemplification, attributive tie and characterization as different relations or noncr.elational ties connecting universals to particular substances or particular qualities falling under them. Sen is inclined to believe that the account could be made more elegant if we eliminate the metaphysically redundant relationship of 'characterization ' , and simply accept that S ocrates has wisdom, though he is not an instance of it in the s ame sense in which his wisdom is. In the fifth part, Sen takes a critical look at the logical foundations of Strawson' s theory of properties b y analysing the concepts o f predicate and quantification subj ects on which Sen has written extensively and illuminatingly earlier on in his career. It is here that Sen comes back to the question of the status of a fact or proposition. Since the fact that S ocrates is wise is not a universal itself, what sort of an abstract entity is it? S en suggests a novel answer. Perhaps we could regard that fact, which Sen identifies with Socrates' being wise , as a particular instance of the general predicate: someone's being wise. Propositions (at least true propositions) would then be completed instances of a certain kind of incomplete or gappy generalities expressed by such predicate expressions such as 'that ... is wise '. Sen carefully desists from attributing this view to Strawson. In the sixth part of the essay Sen looks at Strawson's endorsement of the standard argument for the existence



of universals and his refutation of the Quinean counter-arguments against the existence of universals. The. final and most audacious seventh part of this essay is devoted to two substantial opJe.frt.ion$ against Strawson's theory of universals . S en makes it very clear in this part that he disagrees with Strawson with regard to the alleged lack of causal efficacy of universals and with regard to the nature of our direct knowledge ' of universals . It is this last theme from S en that Chakrabarti picks up in the last essay of this collection. In his 'Reply to Pranab Sen' (Chapter 3), Strawson welcomes the prominence that Sen gives to the particular items which he, S trawson, calls 'property-instances ' and which are now generally called 'tropes '. He applauds Sen's elegant demonstration of the theoretical simplification that this makes possible. Regarding the question of the extension of the term 'universal ' as of relatively minor importance, Strawson says that the crucial distinction is that between, on the one hand, abstract intensional objects in general, 'including propositions and numbers as well as the undoubted universals of property, relation and kind' , and, on the other hand, particular entities such as substances, events and tropes . An apparently deeper difference from Sen concerns the direct sense-perceptibility and causal efficacy of universals which Strawson questions but Sen seems ready to concede. But even this difference is less profound than it appears, as becomes clear in Strawson 's own contribution to this volume. The two essays following Strawson's Reply are detailed engagements with Pranab Kumar Sen's work on the theory of universals, though each of them goes overtly beyond Sen's ideas . One feature of Sen's lucid writing is that it makes any attempt at further elucidation or exposition redundant Yet Jonardon Ganeri (Chapter 4) manages to make s ome of Sen's metaphysical agenda clearer. He brings out quite sharply that Sen has done a pretty drastic j ob of pruning as well as adding to the list of types of universals that Strawson had eventually drawn up . After Sen's maj or revisions (Ganeri calls it ' de-mobbing' ) , Strawson' s world of substances, particular qualities, events, universals, and the non-relational tie(s) between them, looks surprisingly similar to the Vaise$ika ontology of the 'seven types of entities-meant­ by-terms ' . Of course, Ganeri knows very well that this resemblance is still partial, since the Vaise$ika categories of ultimate individuators of impartite particles , and objective absences, are alien even to Strawson's generous metaphysical imagination. That is why he claims that the similarity would obtain between a pruned Strawsonian and a pruned Vaise$ika world. The most original and daring arguments in Ganeri's chapter appear in its last three sections. The traditional criterion of universals endorsed by both Sen and Strawson is this: 'Universals can appear either as subj ects or as predicates of propositions, whereas particulars can only appear as subjects ' . But in sentences such as 'Gold is precious ' or 'Figs are sold in dried form' or ' Swimming is good for health' , a mass-term, a bare plural, and an abstract or gerundial noun are used as grammatical subj ects . What is really being talked about in those sentences ? Being precious is, of course, predicated distributively of actual pieces or nuggets of gold, which, as Sen and Ganeri rightly point out in criticism of Strawson, are not really related to gold as cats are related to the universal catness . Each particular fi g i s said t o b e s old i n dried form. But the grammatical subj ect­ term 'Figs ' introduces something else into the propositional content besides those


Universals. Concepts and Qualities

particular fruits, by what Ganeri calls deflected predication. This is' not done directly through referring by means of a subject expression, nor by means of a predicate expression, S omething is 'invoked' as (what in New Nyaya would be called) 'the delimitor of subjecthood', in virtue of which the particulars figure as subjects of predication: the property of being gold, the property of being an arbitrarily chosen fig, the general property of being any particular act of swimming. With this new idea, gleaned from the New Nyaya theory of meaning, Ganeri now wants to mediate between Strawson and Sen. Strawson rightly holds that feature-introducing mass-terms, which pass the generality test because they can occur both in subject and predicate positions, must be introducing some kind of generality into discourse. Sen, on the other hand, rightly refuses to call gold or water universals, because their specimens are their parts , not their instances . What is common between the uses of 'gold'in ' Gold is rare' and 'What I have on the nib of my pen is gold ' , according to Ganeri, is neither a universal (so Sen is right) nor a particular (so Strawson is right) ; it is a property that serves as a delimitor of subjecthood or predicatehood. Ganeri calls it, provocatively, a non-universal generality. It is actually a strikingly original way of presenting the fairly entrenched Nyaya notion of a ' titular property' (upadhi) which is required as a property for explaining the content-structure of a cognition but cannot or need not be postulated as a real essence or natural kind universal in the world. Fraser MacBride has a rather different understanding of Sen's project. Sen had asked: 'How many properties are there? ' According to the prevailing empiricism that has come to dominate contemporary metaphysics, this question can only be answered a posteriori by the findings of total science. Sen, MacBride thinks , developed an opposing position inspired by a rationalist tradition, a tradition according to which the existence of properties may be established a p riori by reflection upon the meanings of predicates. In 'Predicates and Properties ' (Chapter 5) Fraser MacBride seeks to articulate, develop and evaluate the arguments underwriting Sen's theory of properties. Two of the distinctive claims of this theory - claims upon which Sen insisted. over a number of years are these: ( 1 ) predicates do not refer to properties ; and (2) logically equivalent simple predicates nevertheless express the same property. The first of these claims is argued for upon the grounds that unless we endorse it we will be unable to distinguish a genuinely truth­ evaluable sentence from a mere list of expressions or concatenations of referring terms . The second of these claims is motivated by Sen's adherence to the Quinean maxim 'no entity without identity' . According to MacBride, the first of these claims is motivated by another Quinean presupposition that Sen makes, namely that quantification is only admissable into the name-position. However, MacBride argues, contra Sen and Quine, for the eligibility of quantification into predicate­ position and the admissibility of entities without identity. In doing so, he seeks to repair Sen's theory in such a way as to allow for properties that corresp ond to both simple and complex predicates, The result would be a more generous theory of universals, but one which would perhaps make a 'gentle naturalist' like Sen or Strawson uncomfortable. From such a defence of the rationalist realist position, we then go to the opposite extreme - a radical empiricist reductionist position which treats universals as mere phoney shadows of general words . Such strong anti-realism about universals is -



represented in this volume by Mark Siderits's contemporary analytical defence of Buddhist Nominalism (Chapter

6). Buddhist ontology is austere like Quine's 'Desert

Landscape', containing only essenceless unique particulars arising and then vanishing, formiJ'lginnumerable interdependent series of causal-temporal succession. The Buddhists did not fail to appreciate the force of the Indian realists' argument that we need to make sense of general words as well as of our conceptually enriched experience of many things as belonging to the same kind. Being pioneers in the field of Indian logic, they were also keenly aware that we need inductively projectible connections - necessary co-locations of the form wherever there is F, there is G


between general characteristics of things for our inferences to work.

But they explained our ability to apply the same predicate to a variety of distinct particulars through their semantic theory of exclusion

(apoha) which Siderits lays

out in crisp detail in his contribution. According to this theory, a descriptive term '1' applies to a particular just in case it is not in the class of things said to be 'noH'. The seeming circularity of this account - doesn't the use of 'not-t' require us to get the meaning of 't' first? - is avoided by distinguishing between two kinds of negation, and by showing how perceived similarity can reflect the interests of the perceiver. Siderits discusses several difficulties faced by this theory, the most serious of which is that it may in the end prove unable to avoid positing a subjective disposition to find certain presentations (positively) similar, through the back door of which similarity-grounding properties would tend to sneak back into the account. Finally, the chapter examines the question whether the Buddhist Nominalist approach might yield a viable alternative to Peacocke's theory of concepts. There is a major difference between the history of the problem of universals in the West and the corresponding history of the same problem in India. No philosopher or philosophical school in India ever set up a world of Platonic Forms as a 'real-er' world of which the world of substantial particulars was taken to be a less real copy. Although sometimes the word 'form' or 'structure'

(ilkrti) was used in ancient

Sanskrit to mean a universal property, never did even the most committed realist think of humanity as a perfect, ideal, paradigmatic human being standing separately from individual human beings. The tie between the real universal potness and particular pots was hotly debated in Sanskrit philosophy. Sometimes it was taken to be an identity-in-difference and sometimes it was said to be the tie of 'inherence', a unique sort of relation which gives rise to the sense that the property is 'in here' inseparably in each of its exemplifiers. This is closer to the Aristotelian conception

(ta katholou) but totally unlike the Platonic concept of Form (eidos or eidea) . A s i s well known now, while Plato's fascination with Forms had something

of universals

to do with his reverence for geometry as the perfect discipline giving us insight into reality, the Indian philosophers were, from around the same period or even earlier, obsessed with grammar and semantics. The theory of universals arose in philosophy of grammar and in one school of Vedic exegesis as part of the theory of meaning of multiply applicable words. In later Vai§e�ika thought, the metaphysical need for universals was normally argued for from three different angles; as connotations or 'reasons for application' of common nouns (such as 'cow') and adjectives (such as 'real'); as best explanations of our experience of qualitative sameness and resemblance across diverse particulars; and as those in virtue of which general law­ like connections could be established between particular cause-events and effect-


Universals, Concepts and Qualities

events. But the ancient context of the realist-nominalist debate was primarily the theory of meaning of general words or descriptive names. As we see in Siderits's reconstruction of the Buddhist position - and there were other nominalists too, even within the orthodox Vedic camp (e.g. the Maddva Vedanta philosophers who were staunch external world-realists and pluralists, yet jettisoned the category of universals as well as inherence as a relation) - nominalism also took a negative­ imagist turn in India which it did not take in Ockham, Berkeley, or Goodman. Richard Sorabji (Chapter 7) offers a magisterial survey of the many transformations of and disagreements within the metaphysics and epistemology of generalities in ancient Greek thought during the first thousand years after Plato. This survey is not merely exposition and history of ideas. It also reconstructs the arguments by which the bloated ontology of Platonic universals was being 'deflated' first by Aristotle and his commentators and then more drastically by the Stoics. Somewhat like the post-Fregean Western scene, the problem-space in this post­ Platonic period got predictably convoluted by its entanglement of questions of concept-acquisition and concept-deployment with questions of the existence and graspability of universal properties. One of the extraordinarily interesting pieces of information that Sorabji gives us, in this context, is that Chrysippus the Stoic identified a universal with a predicate which is an 'incomplete sayable (lekta) ' . This unusual view sounds like a vague anticipation of Frege's idea of reference of predicates as unsaturated entities (except that Frege's functions belong to the realm of reference whereas the Stoic lekta gravitate more towards Frege's realm of thoughts and senses). In the last part of this relevant review of the ancients, Sorabji states succinctly his own stand in the several controversies that the Greeks have bequeathed to us. Somewhat like Strawson, he is unwilling to concede to Plato that universals themselves can be causes, although, somewhat like the Indian realists, he grants that they have a role to play in causal explanations, since partiCUlar heating-events cause boiling-events only in virtue of exemplifying the general property of being a case of heating. He repudiates the.existence of uninstantiated universals and, again· like Strawson, hesitates to admit that the universal itself can be perceived. His gloss on Aristotle's paradoxical claim 'that although one perceives a particular, perception is of the universal', gives us a glimpse of 'ancient cognitive science' about the infant's coming to recognize red things as red on the basis of initial non-recognizing perceptual encounters with samples of red colour. Kant had called the acquisition, possession and empirical application of concepts 'a secret art residing in the depths of the human soul' (Critique of Pure Reason, Kemp Smith translation, B180-81). Aristotle and his commentators seem to have worked nearly as hard as recent philosophers at unravelling the nature of this mysterious mastery over generality. Neither committing themselves to Platonic forms, nor committing universals altogether to the realm of linguistic fiction, some Stoic deflationists, Sorabji shows us, had reduced generalities to concepts with which we think of the world. This middle position of 'conceptualism', having been long dormant thanks to 'the linguistic turn' , is ripe for a revival, according to Chris Swoyer (Chapter 8). Although with the recent rise of cognitive science, 'concept' -talk has become as much legitimized as 'meaning' -talk was before Quine taught us to fear meanings as



creatures of darkness, the fIrst difficulty that Swoyer faces, of course, is that the concept of a concept is still far from clear. Swoyer recognizes that the classical view of concepts as what is captured by tight definitlQJJ§")S dead. It has been replaced by other more psychologically robust and vagueness-tolerant competing theories of concepts such as the resemblance-based prototype theory, theory theory, or atomistic informational or mental representation theories of different shades. Yet Fodor's concepts and Peacocke's concepts do not seem to be the same entities, and the concepts that fuel Swoyer's conceptualism seem to be yet another type of items, explicitly said to be 'mental' but 'not purely mental'. While Peacocke has stated in no uncertain terms, 'Concepts are abstract objects' (A Study of Concepts, p. 9 9 ) , Swoyer's major motivation for adopting conceptualism is that 'it would allow us to avoid abstract entities' . Giving us a quick overview of the connections between theories of generality, theories of meaning and cognitive content, and theories of perception and inference now available in the market, Swoyer puts forward a reductive solution to the problem of universals by a psychological account of how we come to think of similarities and form classes out of diverse perceived particulars. But somewhat unexpectedly, he supplements that account by an ontology of spatio-temporally available ' in re properties'. Even at a pre-linguistic level, humans and other developed animals, Swoyer agrees, live successfully in the world by classifying, recognizing, projecting and generalizing. He shifts 'some of the burden' of explaining this fact from the external reality to an internal realm of concepts. Some of it but not all, since there exist, he thinks, real properties in nature too. It would be interesting to compare Swoyer's cautious conceptualism with Fodor's equally qualifIed mixture of anti-realism and realism about natural kinds, flaunted in this typical Femark: 'It turns out, in lots of cases, that we make things be of a kind by being disposed to take them to be of a kind . But not in every case; not, in particular in the case of kinds of things that are alike in respect of the hidden sources .of their causal powers, regardless oftheir likeness in respect of their effects on us' (Concepts , p. 162, italics in original). We have already seen that to accept the traditional divide between universals and particulars is to accept that properties such as humanness or wisdom can fIgure both as what is predicated of individuals and as discussable individuals themselves, unlike particular wise humans such as Gandhi, who cannot be predicated of anyone and defInitely not attributed to multiple particulars. If one regards the semantic distinction between the meanings of singular terms and the meanings of predicate­ expressions to be sacrosanct and non-overlapping, then the above role-reversal of universals from predicate-positions to subject-positions would be an intolerable consequence. Harold Noonan (Chapter 9) shows us how, in order to avoid this repugnant consequence, Frege banished universals from his mature ontology by redrawing the map (of the realm of reference) along the clear and inviolable division between objects introduced by singular terms and functions introduced by predicate­ expressions. Thus Noonan is certain that Frege's functions (though he permits some of them to be called 'properties') are not universals, for universals themselves are objects, albeit abstract ones, but no function can be an object. . .


Universals, Concepts and Qualities

But, even as Frege declares that what is referred to by a predicate can never be the object picked out by a name, he seems to refute himself by talking about some object which the italicized noun-clause itself refers to. Noonan's chapter is a subtle defence of Frege's apparently paradox-inviting position, in the face of Crispin Wright's admittedly powerful objection that the categorial divide between concepts and objects finally breaks down. Some philosophers, for example P.K. Sen, have rejected Frege's extension of the idea of having a reference to predicates. But even if one forgets about searching for the reference of ' . . . is a horse', and takes the concept horse as the semantic value of the predicate or what the predicate introduces or invokes or connotes, the same paradox would seem to arise. The co-referentiality of the two expressions 'whatever is introduced/connoted by ' ... is a horse' and 'the concept horse' would tend to make an object out of an unsaturated functional entity. While honouring all the disquotational platitudes such as: "'the concept horse" refers to the concept horse', Noonan argues that concepts would never be objects, because then it would have to be possible to give the same uniform reading to 'X' in the following sentence-schema: '''... is a pot" applies to something if and only if it X and "the concept pot" refers to X' . But that is not possible. Whatever reading of 'X' would make sense out of the bit 'if it X' will make nonsense out of the bit 'refers to X ' . X cannot retain its type­ identity as an expression and be both an unsaturated gappy predicate and a saturated singular term, both verbal and nominal in grammatical terms. Noonan says that the Fregean paradox is unavoidable but it reveals the fact that the logical distinction between concepts and objects, unlike the traditional distinction between universals and particulars, is rooted in the linguistic distinction between names and predicates which language itself tends to smudge with its tendency to 'call' even unnameable entities by name-like locutions. Thus, Dne major question concerning allegedly non-particulilI entities,... pmper):ies, abstract general qualities, features, universals, shared characteristics, predicable incomplete sayables, functions, whatever we call them - is this: which comes first, the logico-grarnmatical distinction between subjects and predicates, or the ontological distinction between particulars and non-particulars? Is the linguistic distinction a clue to discovering a real distinction out there in the world? Or, do we commit ourselves, for extra-linguistic empirical or apriori reasons, first to the universal­ particular divide and then break up an intrinsically indivisible sentential unity into the parts of singular terms and predicates? Showing his commitment to the thesis that the logical distinction between singular terms and predicates is independent of and prior to the ontological distinction between objects and properties, Bob Hale (Chapter 10) presents, first, a careful and sympathetic formulation of Ramsey's vigorous attack against any version of the traditional distinction between particulars and universals. He then analyses Dumrnett's defence of Frege's distinction between objects and concepts (which he, unlike Noonan, takes to be a logico-linguistically sophisticated version of the traditional distinction) in the face of Ramsey's attack, and raises some technical issues regarding the effectiveness of Dumrnett's rebuttal of Ramsey. While Ramsey



rejected the logical distinction between subjects and predicates, but would have had no qualms in assigning both types of expressions distinct sorts of non-linguistic referents had the distinction been real, Hale defends the logical distinction between types of eXR.l'�c��ons but questions the legitimacy of drawing any metaphysical conclusions, traditional or Fregean, from that distinction. Since the justifiability of Frege's ascription of reference to predicates (and generally to incomplete expressions) has been powerfully challenged - indeed Pranab Kumar Sen, though a staunch realist about universals, has always rejected this Fregean move - acceptance of the logical distinction does not by itself carry any commitment to universals as referents of predicates, and is thus, so far, consistent with nominalism. Properties are needed not only to explain our sense of (loosely speaking) sameness or (strictly speaking) similarity across things which we count as distinct, but also to explain why we count two things as distinct in the first place. Between the Leibnizian principle of Identity of Indiscernibles (and its contraposed equivalent: Necessary Discernibility of Non-identicals) and the ontology of universal properties, at least two clear conceptual links can be immediately established. First, we need to quantify over properties in order to state that principle in a logically perspicuous language. Indeed, if we remember the (Vaise�ika) stricture against calling any attribute a universal property unless it has more than one instance, we are forced to derive the following multiply quantified result from the principle of Identity of Indiscernibles: for every property, if it is to be had by any individual which requires that it has to be shared at least by two distinct individuals, then there must exist at least one other property that one of them has and the other lacks. Second, if we apply the principle to the identity and discernibility of properties themselves, we are forced to think of property of properties in the following way: if two properties are distinct, then there must be a property that belongs to the first property but does not belong to the second property. In his technical chapter on how not to trivialize the principle of Identity of Indiscernibles (PIT), Gonzalo Rodriguez-Pereyra (Chapter 11) tries to define one such property of properties: 'the property of being trivializing' which.some properties have but others do not, although he is not concerned with the question about the ontological status of such second-level properties. The central insight behind the principle of Discernibility of Non-identicals is far from trivial. It goes against the possibility that the only distinction between two particulars would be purely numerical, or that the statement that A is distinct from B would be, to use Dummett's phrase, 'barely true'. Now, even this non-trivial principle which, one imagines, would be contested by radically nominalistic proponents of strictly bare propertyless particulars, would become utterly trivial and lose all possible opposition if we include, within those distinguishing properties, properties involving identity, for example, the property of being identical to A. In an ontologically weightless way, even the radical nominalist can talk about such a property and agree that if A and B are distinct, then A has the property of being identical to A while B lacks it. Nothing more is then said by the principle than that B cannot be the same as A, unless B is the same as A. So there must be some limits to what sort of properties can be included within the range of discernment-grounding properties, on pain of making the principle of Identity of Indiscernibles tautologous. He disagrees with Strawson, who insisted that in order not to be trivialized, the domain of properties


Universals, Concepts and Qualities

that the principle quantifies over can only include 'pure' (fully general) properties, and has to exclude relational properties which depend upon the particular identity of a relatum, such as: being two miles from the Eiffel Tower (his example), or being married to Elizabeth Taylor (my example). Such properties, though to some extent general, involve reference to a particular. For Rodriguez-Pereyra, even such impure properties can be non-trivializing, that is to say, that even they can make a qualitative difference. He goes through five alternative definitions of a trivializing property, a property, that is, whose inclusion in the domain over which the PIT quantifies would render the principle trivial. His intricate reasoning behind the rejection of the first four definitions and for accepting the fifth one, incidentally, throws a lot of light on the notion of qualitative as against barely numerical difference. How serious is our commitment to existence when we concede that there 'are' properties which belong to particulars and kinds to which particulars belong? In an interesting paper called 'Language-Created Language-Independent Entities' (Philosophical Topics, Spring 1996), Stephen Schiffer has remarked that we should be just as tolerant of properties as we are of fictional entities. 'Properties', he proposes, 'are hypostatizations of our predicate-nominalizing practice.' George Bealer (Chapter 12) warns us against such pretend-realism, which is nominalist at heart (notice the phrase 'language-created'). Most arguments for universals are based on appeals to common sense or to intuition - for example, the intuition that there are some moral virtues and honesty is one of them. Such arguments are unconvincing to sophisticated nominalists because they believe that these arguments can be deflated using one or another anti-realist technique well known from the philosophy of mathematics literature. For example, according to various forms of fictionalism, we have the indicated intuition only against the background of a tacit verbal acceptance of a realist theory of properties; consequently, the intuition supports only the very weak conclusion that, given the fiction of properties, the property honesty exists. Alternatively, various advocates of substitutional quantification would accept the inference from the intuitive premiss that honesty is a virtue to the conclusion tbatthere exists a propeTty (i.e. honesty), and they would accept the indicated premiss. But these substitutionalists would also hold that this argument does not further the realist agenda because the quantifier 'there exists' in the conclusion is a mere substitutional quantifier, not an objectual quantifier. Bealer constructs a certain style of modal argument designed to ward off these and other such deflationary moves. The resulting 'transmodal' argument, if correct, supports not only the conclusion that properties, relations and propositions exist, but that they necessarily exist. According to this conclusion, for example, the property of being red necessarily exists and so would exist even if there were no red things, just as the traditional doctrine of ante rem realism requires. Very much unlike Bealer, David Armstrong is neither an apriori-metaphysician nor a believer in instance-independent Platonic properties. Somewhat like the medieval Nyaya-Vai§e�ika philosophers, who would appeal to empiriCal intuitions but not to apriori knowledge of universals, Armstrong believes that there should be some severe tests that the meaning of a predicate has to pass before it can be assigned the status of a 'real universal. He rejects disjunctive, negative and unexemplfied properties. So, in a sense, Armstrong's theory of universals is sparser than Bealer's and Strawson's (and definitely, as we shall see, than Kiinne's). In his

Introduction new chapter for this volume (Chapter


13) he addresses a troublesome issue about

the modal status of the link called instantiation or possession. Of course, in his world particulars and states of affairs constituted by particulars and their properties exist or obtaifr cootingently. But given that a certain particular exists, and instantiates or possesses a certain property, is it contingent or necessary that that particular has that property? Since, unlike Bealer, he is no rationalist, Armstrong used to believe that the truth that a contingerit particular has a property is also contingent except in rare cases of essential properties that he grudgingly admitted. But, without giving up his general commitment to empiricism, he now wishes to defend the view that even a contingent particular, given its existence, bears the properties it does necessarily. He proves this in four stages by considering four alternative accounts of the matter. There are two kinds of views about the nature of properties: that they are universals, or that they are tropes. Then there are two kinds of views about the link between properties and the particulars that bear them: that the particular is simply a bundle of its properties, or that the particular stands apart and bears the properties, tropes, or universals as its own attributes. Armstrong shows how under all four combinations of these views, the truths about property-possession come out as necessary. If, for instance, the bearer is just the collection or set of its tropes or features, then its identity as a set makes every actual member indispensable. If the property-set loses any single member, it is no longer that same set. Hence the necessity of its property-bearing. If, on the contrary, the particular exists apart and bears its properties as its attributes, rather than parts or members, then one may think that it could afford to lose a property or two and still retain its sameness. But any man in another possible world, who has all the properties of Bertrand Russell but does not have the property of being a pacifist, would not be Bertrand Russell, but would at best be a closely resembling counterpart. Hence, in all possible worlds where Russell exists he must bear the properties he actually bore in the actual world. This position regarding, as it were, the contingent necessity of property-bearing by particulars has very peculiar modal 'and metaphysical consequences, and Armstrong ponders many of them in his chapter. He extends this view about properties carefully and with caveats to relations as well. Even causal laws, which he takes to be links between universals, he now takes to hold necessarily. This rankles doubly with the Humean legacy of regarding causal connections to be obtaining not between event-types but between particular events, and of denying any necessary connection in nature. But Armstrong is not afraid to go against Hume on this. He presents his arguments for this surprising position about contingent particulars necessarily having universal or particular properties, anticipating and answering some objections. Although Wolfgang Kunne (Chapter

14) accepts the standard Fregean under­

standing of 'predicate' , he distinguishes, within the predicate-expression, between

general term (e.g. adjective or verb-stem) and copula (e. g . 'is' or verb-ending). The expresses a concept, but connotes a property (or relation) which is denoted by the corresponding abstract singular term (noun or phrase) and is itself

general term

an abstract object. 'Particularist' attempts at rejecting such objects by reductive paraphrase are ultimately unsuccessful, according to Kunne, and, in any case, unavailing. The singular terms referring to such general objects are ineliminable in


Universals, Concepts and Qualities

higher-order predications about them. Just as one can introduce a property into a discourse either by using a general term that connotes it or a singular term that denotes it, so we can quantify over properties by quantifying into the position of either a general or a singular term. As far as his ontology goes, Ktinne frankly advocates a generous conception of properties; for a general term to connote a property it is not required that the property be exemplified. But he acknowledges a limit to generosity, in the case of such paradox-generating general terms as 'non­ self-exemplifying' as applied to properties. His argument throughout is illuminatingly illustrated with examples. It is with the force of actual examples once again that, in his own fresh contribution to this volume, Peter Strawson (Chapter 15) makes an intuitively compelling case for a category of partiCUlars such as a smile, a blush, an angry frown, a gesture, the particular blue of someone's eyes, which are specific to the face and time, so to speak. Such non-universal unrepeatable qualities, called 'gwJa' (attributes) are extremely well entrenched in classical Indian (Nyaya-Vaise�ika) metaphysics. Possibly recognized by Aristotle, and certainly recognized by some medievals and by Bolzano, such property-instances have become quite prominent in Western analytical metaphysics relatively recently, particularly in so-called 'trope-theory' (see, as a sample, the anthology called Language, Truth, and Ontology, edited by Kevin Mulligan (1992) where such stalwarts as Chisholm, Armstrong, Lehrer and Hochberg talk about tropes and truth-makers). Strawson disapproves of the popular name 'tropes' because it has an established literary use which tends to import distracting associations into this metaphysical context where unrepeatability is crucial. But he offers at least two clear reasons for accepting these as constituting the meanings of some predicates such as: ' . . . blushed crimson' where both the blush and the crimson are attributed to the face but are not universal properties because they could be seen and they could change, for example deepen or fade. The smile on a particular face cannot be shared by another face, and even on that face it fades. Strawson's recognition of these parasitic particulars (he refuses to use the oxymoron ' abstraet .particulars " for them) goes along with the acknowledgement that universals of which property-instances are instances are themselves sortal universals just as universals of substance-kinds are. Strawson ends the chapter with his anti-conceptualist refusal to reduce universals to concepts. Such a reduction can never work straightforwardly, since everybody who possesses the concept of wisdom is not wise. It is because we have the concept of divinity that we can seriously doubt if anybody actually has that property. In spite of his catholic realism, however, Strawson sticks to his old view that the natural world of causal interactions does not contain universals in it. Universals remain, even in his most Nyaya-friendly moods, objects of thought for Strawson, never objects of sense-experience. In a conciliatory move, he recognizes that there is little difference between saying that universals are perceived in their particular instances and saying that the particulars are perceived as instances of the universals of which they are instances, and that he would say the latter with unqualified alacrity. In his chapter, Arindam Chakrabarti (Chapter 16) picks up this old dispute between Strawson and himself about the perceivability of universals. Strawson has always been certain that they cannot be objects of sensory encounter, for they do



not haye any spatio-temporal location. Even in his chapter for this volume, where he has made room for particular qualities in the world and admits that these quality�tokens could be perceived externally or internally along with and in the same sense a"� J�.particular substances to which they belong, Strawson reiterates his conviction that the universals themselves cannot be objects of sense-experience, because they are 'intelligibilia' . What are the arguments against the perceptibility of universals ? Chakrabarti locates three such maj or arguments given, respectively, by classical Indian Nominalists, by Strawson himself, and by philosophers of broadly Fregean persuasion. These are: the argument from inaccessibility at the first encounter (if properties were perceived, we would perceive them even at the time of seeing the first single exemplifier, but we don ' t) ; the argument from lack of causal efficacy (if properties were perceived they would be causes of perception but they are not) ; and the argument from obj ect-function distinction (if properties were perceived they would be objects but they are not) . He casts serious doubt on the soundness of each of these arguments and makes a case for direct perceptual knowledge of some universal properties, marking a deep epistemological disagreement between the two realisms about universals upheld by the two editors of this anthology. Of course, Chakrabarti reminds his readers, to see a universal immanent in a particular is not necessarily to see that it is a universal. To know that the property one has seen is a universal, one needs to think and have philosophical - not sensory - knowledge. This collection of essays has been assembled and published in honour of the late Professor Pranab Kumar Sen, to whose memory it is dedicated. It is fitting, therefore, that the first chapter in the book should be an unpublished work of his own, indeed the very last complete essay that he wrote before his death; and since it was intended to be his contribution to the volume of the Library of Living Philosophers concerned with the work of P.E Strawson, it is no less fitting that what Strawson would have written as his reply should immediately follow it.

References Fodor, Jerry ( 1 998), Concepts ( 1 996 John Locke Lectures). Oxford: Oxford University Press. Kant, Immanuel ( 1 933/1973), Critique of Pure Reason , trans . Norman Kemp Smith. London: Macmillan. Mulligan, Kevin (ed.) ( 1 992), Language, Truth and Ontology. Dordrecht: Kluwer. Peacocke, Christopher ( 1 995), A Study of Concepts. Cambridge, MA: MIT Press . Pereyra, R.G. (2000), 'What i s the Problem o f Universals ? ' , Mind, 109 (April), 255-7 3 . Strawson, P.F. ( 1 959/1 979), Individuals: A n Essay i n Descriptive Metaphysics. London: Methuen.

Chapter 2

Strawson on Universals Pranab Kumar Sen

One of the most distinctive and, at the same time, encouraging features of the philosophy of our time is the rehabilitation of metaphysics as a legitimate branch of philosophic enquiry. This revival has been made possible through the works of a number of leading philosophers of our time, philosophers who could shake off the strangle-hold of the anti-metaphysical rhetoric that succeeded in nearly completely inhibiting all metaphysical impulses in lesser philosophers. One of these leading . philosophers of our time is Sir Peter Frederick Strawson. Strawson's contribution to metaphysics is often taken to be restricted to an enquiry into the most general conceptual scheme or schemes in terms of which we seek to understand our experience and the world. Thus when we think of 'Strawson 's

descriptive metaphysics he elaborated Individuals . ! This is unfortunate, for it gives a truncated view of what Strawson has done for metaphysics as a whole. Even in Individuals, Strawson metaphysics' we almost inevitably think of the

in his

contributes not only to the investigation of conceptual schemes but also to the enquiry into 'what there is' , or ontology. Strawson himself shows, intermittently, a tendency to underplay the importance of this aspect of the metaphysical inquiry to which he has made so significant a contribution. Strawson 's contribution to ontology lies mainly in his work on the theory of universals. In fact, as it may be pointed out, the question regarding what things are . there (in the world) becomes urgent only when the existence is chumed of entitl es · other than the day-to-day medium-size particulars of sense, especially of the so­ called ' abstract entities' which include universals. Although the reality of the particulars of sense has also been called into question, and sometimes even denied altogether, by philosophers belonging to different times and traditions, these objects seem to have a kind of

prima facie claim to reality which nothing else has.

However, the. context in which the subject of universals is given so impressive a treatment by Strawson in his

magnum opus - Individuals - is provided by his

discussion of the conceptual scheme in terms of which, he argues in the work, we think whenever we think about the things of the world. This might have suggested that the

only claim that Strawson is making in his most definitive treatment of the

subject is that we do need the concept of a universal for our thinking, that this concept is indispensable for our thought; and that he is not making the further claim that corresponding to these concepts there also are in the world, along with the sensible particulars, general

things, independently of how we actually, and

perhaps unavoidably, think. This apparent reluctance to make ontological commitment with respect to the concepts in terms of which we think and perceive


Universals, Concepts and Qualities

our world may also be thought to be in conformity with Strawson's well-known Kantian inclinations. To recall the Kantian view here, we cannot think except in terms, for example, of the concept of a thing or a substance having properties or attributes, but that does not mean, for Kant, that there really are substances with attributes in the world of things as they are in themselves, that is, iridependently of how we view our world. But it would be a mistake, I think, to identify Strawson with this particular Kantian stand. For the first thing, as we know, Strawson has [mnly rejected Kant's 'transcendental idealism' ,2 and, for the second, Strawson has never aligned himself with the conceptualist position, according to which the only general things that we need to admit are general concepts. In fact, h�as never seriously considered the conceptualist option, but has always restricted himself to a consideration of the opposition between realism and nominalism. With these preliminary remarks, we may begin our discussion of Strawson's view of universals. In our discussion we shall depend mainly on the following of his writings: (a) 'Particular and General';3 (b) Individuals;4 (c) 'Entity and Identity';5 (d) 'Universals';6 (e) Skepticism and Naturalism: Some Varieties;? (f) 'Two Conceptions of Philosophy';8 and (g) 'Arindam Chakrabarti on Non-Particular Individuals' . 9 Our main objective in this chapter will be to bring out, as clearly as we can, Strawson's own theory of universals, which is subtle, rich and comprehensive in a way in which very few of the wide variety of theories of universals offered by philosophers over the ages can be said to be. We shall not have much time to spend on the polemics Strawson has joined again and again with other philosophers on the subject of universals. But there will be some discussions of these polemics too. The last section in particular will be devoted entirely to them. As we are more interested in being clear about Strawson's positive contributions to the subject rather than in anything else, we shall depend mainly on the material provided by 'Particular and General' and Individuals.


In my exposition of Strawson's theory I shall primarily be concerned with two questions: (a) what things are universals according to Strawson? and (b) what kind or kinds of thing are they on his theory? As answers to these two questions cannot be given in isolation, they will be considered together. As far as I can see, there are mainly three things for us to look at to find an answer to our questions: first, there is the wide assortment of examples given at different places by Strawson of what he says are 'general things', or 'universals', these two terms being used by him more or less interchangeably (sometimes rather controversially, as we shall see later on). There is, in particular, the rather elaborate list he gives in 'Particular and General', where a kind of threefold division of the diverse generalities in the list is also attempted. Second, there are the 'quasi­ definitions' Strawson proposes, in the same essay, of the particular and the general in terms of two mutually exclusive necessary conditions - one for the general and the other for the particular - at the end of a detailed examination of the variety of examples in the above list. Third, there is Strawson's elaborate treatment of the

Strawson on Universals subj ect. o f universals i n


Individuals, which i s especially marked b y a twofold

division of universals, supplemented later on by a consideration of universals of a

third kind as well, bringing the division of universals close again to the threefold division in ' Pq.rtiQu.lar and General' . Also noteworthy in the s ame work is Strawson' s

treatment o f "the" relations in which universals of different kinds stand to their corresponding particulars. Each one of these should tell us something about how

the universal is conceived by Strawson. However, as we find, although e ach of

these gives us some clue to it, none of them by itself tells us exactly how universals

are viewed by the philosopher. It is only by putting all thes e clues together, and by

weighing them against one another, that we get a clearer idea of Strawson's p osition.

As we shall see, these clues do not point to exactly the s ame view or views, for

Strawson ' s thoughts on universals seem to have changed in some respects over

time, though not radically, I hasten to add. Our clues tell us about these changes as


Let us first look at the classified list of generalities in 'Particular and General ' . 1 0

The list is as follows :


Materials, which are designated by material-names like 'gold' , ' snow ' , 'water' ,


Substances, which are designated by substance-names like ' ( a) man ' , ' ( an)


'j am' and ' music ' .

apple' and ' (a) cat' .

Qualities or properties , which are designated by quality- or property-names

like 'redness ' , 'roundness ' , ' anger' and ' wisdom' .

It is on the basis of this list that S trawson formulated, towards the end of his

essay, 1 1 the two mutually exclusive necessary conditions for the p articular and the general. But the wide diversity of examples in S trawson's list makes it extremely

difficult for him to offer us a quick generalization about what a universal as such i s .

S o l e t us also take note of how he looks a t the diversity.

Immediately after giving his threefold division of universals, or generalities,

Strawson points out the following differences between them. I2 (a)

The generalities of the third kind differ from those of the other two in two


First, of all noun-phrases supposed to stand for universals , abstract noun-phrases

are the most

dispensable. It seems that we c an always replace an abstract noun

standing for a quality or property by the noun or adj ectival phrase from which it is

derived, without any loss either of truth "alue or of meaning. Apparently, Strawson

has in mind such equivalences as the following: instead of saying ' S ocrates had courage ' , we can always say ' S o crates was courageous ' ; instead of saying ' S o crates

had wisdom' we can say ' S ocrates was wis e ' ; and instead of saying ' Wisdom is a virtue' we can s ay 'All who are wise are virtuous ' . This might suggest that Strawson

is also saying that the alleged universals too, and not only the noun-phrases which

are taken to stand for them, are the most dispensable of all the different kinds of

universal that have been admitted by philosophers . For a general heuristic principle which has often been invoked by philosophers to decide whether or not we have to

Universals, Concepts and Qualities


admit a putative entity into our ontology is that we have to admit it only if an

expression which is designed to

refer to that entity is indispensablg") B ut, as we

shall see later, Strawson does not subscribe to this principle. Even if w'e succeed in

eliminating all abstract noun-phrases like ' wisdom' and ' redness ' , we c an still be

committed to the existence of universals, for there are more ways of making

ontological commitment thari just one, Strawson maintains .

Second, b y contrast with the generalities o f the two other kinds, they ar e c ap able

of having instances belonging to different c ategories ; their instance s are not

neces s arily confined to any single category, as the instances of the generalities of

the other kinds are. Thus a man, a remark and an action are things of three different categories, and yet all of them could be wise, and thus be instances of wisdom; a

pillar box, a sunset and a mental image are things of three different c ategories, and

yet all of them could be red, and thus be instances of red (redness) ; and a word, a gesture, an expression and a man are again things of diverse categories, and yet all

of them could be angry, and thus be instances of anger. But there is no such

possibility in the case of universals of the two other kinds . All possible instances of

(a) man would be things of the s ame category, and s o would be all possible (an) apple.

instances of (b)

Strawson calls our attention next to the difference which obtains b etween the

universals designated by noun-phrases of the second kind, that is, substance-names ,

on the one hand, and the rest of the universals , o n the other. The difference is this : if

F is a universal of the second kind, then if X is an instance of (an) F then X (itself) is (an) F . But this is not the case with the universals of the two other kinds . Thu s : if something is an instance of a horse then it is a hors e ; but we c annot s ay that an

instance of gold (designated by the material-name ' gold ' ) is itself gold; what is an instance of gold is a piece, a bit or a lump of gold, or it is something which is made of gold.


Apart from the distinctive feature noted in


above to characterize universals

of the third kind, which are qualities or properties, there is another. It is that,

corresponding to a universal of this kind, there is a

special class of instance s . For

example, corresponding to the universal wisdom, we have as an instance not only

but also the wisdom of Socrates; and, corresponding to the universal but also the redness of Smith 's face. Such ' dual instantiation ' , if we may c all it s o , is not S o crates ,

redness , we have as an instance not only Smith' s face (which is red) , possible in the case of universals of the two other kinds .

Strawson ' s idea here, we c an say, is roughly this : property-universals have as

instances not only the partiCUlar things which all share them as common properties,

but als o the

particular properties (or qualities) which these things individually have as distinctively their own. S o , to take one more example, not only does a man

instantiate happines s by being happy ; s o does his happiness, which is distinct from the happiness of any other man .

We c an now turn to our second clue, which is provided by the two mutually

exclusive necessary conditions for the particular and the general. 1 3 The conditions are stated by Strawson as follows :

Strawson on Universals (1)



It i s a necessary condition for a thing's being a general thing that i t can b e referred t o b y a singular substantival expression, a unique reference for which is determined s olely by the meaning of the words making up that expression. IUs a nE!,9 •



Universals, Concepts and Qualities


ethnographers report on the Trobriand concept of causation or the Samoan concept of shame, they are often providing a sketch of a cultural average or central tendency. 2. 6

Centrality and Degree

The next point is more controversial. We can employ some concepts without knowing much about them, but in other cases we, or specialists, need a good deal of information to correctly classify instances of a concept or to reason with them ( ,Is the dark spot on the X-ray a malignant tumour? ' ) . Two key questions for any theorist of concepts are how much of this information should be built into the concept, and how much of it two people must share to have the same concept. At one end of the spectrum we find holistic theories that make the content of a concept depend on all of our other concepts and all of our beliefs . Views of this sort occur in some accounts of the incommensurability of concepts (e.g. mass, gene) before and after a scientific revolution and in comparisons of different cultures (e.g. the Ancient Greeks ' concept of rights) . At the other extreme we have atomistic views that treat concepts as directly referential mental representations with no descriptive content. These pictures have an all-or-(almost)-nothing flavour, but in a moment we ' ll see that there are a number of options in between. It is important that on some accounts of concepts (e.g. prototype theories and conceptual role theories) concepts can have descriptive content without any sharp analytic or synthetic distinction and without any claim that some of a concept' s content is essential t o i t while other content is not. As with many other things, for example, ships (like Theseus ' ) , there is often no precise point in our concepts where one more change in belief would rub out the old concept and give rise to a new one . On this view there is no useful distinction between change in concept and change in belief. Indeed, whether we regard a given case as one or the other can sometimes even depend on social and political factors that have little to do with meaning or metaphysics (Stich, 1 996, ch. 1 ) . These views require defence, but my-reasons for holding them are similar to .those of Sellars (e.g. 1 954; 1 974, p. 3 84) and Harman (e.g. 1 987; 1 999), so I won't defend them here. The key point, though, is that some bits of content and some inferential roles can be more central than others. It is more central to my concept of a dog that it is an animal and barks than that all dogs have fewer than six million spots . Furthermore, two people (or one person in different contexts) can agree and disagree about many things even if their concepts are not completely identical. We cannot profitably discuss fruits if you insist, in the face of all the things anyone can say, that they are a species of algebraic numbers and I think of them in a normal way. But we can discuss many things about fruit without talking past each other even if we disagree about whether tomatoes are fruits or vegetables . Stability requires similarity in concepts, but not total similarity. Furthermore, what beliefs or inferential roles are central to a concept can depend somewhat on context. 2. 7


Almost every psychological phenomenon is highly sensItIve to context. Our perceptions in a given situation are influenced by our perceptual set. What we


1 37

retrieve from memory is influenced by the context in which we retrieve it, including expectations , priming, retrieval cues . Our preferences between two options are influenced, sometimes even reversed, by the way the options are framed or described (e.g. Kahneman and Tversky, 2000) . There is also a great deal of evidence that context can affect the way we apply a concept, where we draw its borders of application (e.g. B arsalou, 1 987) . 2.8

Indefiniteness of Content

The contents of our memories, beliefs, desires, values and attitudes are often indefinite, protean, and partially constructed in a given context in ways that depend heavily on that context, for example, expectations , priming, framing effects, the way questions are worded, emotional state and mood (e.g. Kahneman and Tversky, 2000; Slovic, 1 9 9 5 ) . It seems possible that the concepts and beliefs stored in memory are more abstract than we might suppose. If so, there may sometimes be no simple answer to questions about the content of our concepts and beliefs, though there may be in particular contexts . 2.9

Kinds of Concepts

Empirical work has largely focused on concepts of various sorts of obj ects (e.g. birds, fruit, furniture) and (especially in cross-cultural research� colours. But there are many other sorts of concepts, and the differences among them may be of philosophical or psychological or linguistic interest. We can classify concepts by their content: moral concepts (duty, justice, individual rights) ; social concepts (love triangle, bureaucracy) ; logical concepts (some, not, and, only ij); concepts involving the self (e.g. the self), concepts of abstract things (Euclidean triangle, vector space over the reals) ; aesthetic concepts (beautiful, kitsch) , and so on. We also classify concepts by their structHre- (or the stmcture ·of their instances ) ; sortal concepts (concepts of sorts o f things , e.g. dog, chair) ; quality concepts (associated with adj ectives, e . g . , white, square); relational concepts (loves, lies in between) ; compound concepts (axis of evil, pet fish) , and so on. Different families of concepts may also be structured in different ways. For example, there are hierarchies of concepts connected by relations like is a kind of; dogs are a kind of mammal and mammals are in turn a kind of animal. The organization here is a descendant of Aristotle 's hierarchy of genera and species. The nature of a concept may even depend on how it was learned, say by learning a definition versus being exposed to examples and receiving feedback about the accuracy of one ' s classifications . A popular current distinction is between concepts based on similarity, for example, red, and those based on rules , for example, prime number (Smith and Sloman, 1 994; Nosofsky et aI. , 1 994; Erickson and Kruschke, 1 9 9 8 ; Patalano et aI. , 200 1 ; cf. Tenenbaum, 2000) . As w e will see, this i s part of a more general tendency to see cognition as involving both rule-based constructs and similarity-based constructs .

138 2.10

UniversaLs. Concepts and Qualities

Cognitive Functions of Concepts .

Concepts play essential roles in all our higher mental processes, including perception, learning, memory, inference, problem-solving, decision-making and action. I will touch on some of these here.

Classification We emp loy concepts constantly to sort the world into types of things (people, lakes), events (explosions, bank robberies), stuff (water, mud), qualities (redness, having a mass of 3kg), and more. Proper classification is often vital; the psychiatrist needs the correct psychodiagnosis to decide what drugs to prescribe for her suicidal patient, and she may use devices (like the Minnesota Multiphasic Personality Inventory) designed to help with classification. Nosology and taxonomy are important areas of study in their own right (e.g. Waller and Meehl, 1 997) , and classification systems like Mendeleev' s periodic chart play an essential role in organizing knowledge . There can even be deep divisions about how things in a given field should be classified, as illustrated by the debates among pheneticists , cladists and others in evolutionary biology. Perception Perception, along with action, is one of the two interfaces between the mind and the rest of the world, and it depends on concepts in various ways. For example, most vision is to some degree 'theory-laden' ; we see things as things of a given sort. I don ' t see a medium-size brown shape in the comer. I see a dog. This involves 'top-down processing ' , which depends on our concepts (e.g. dog, in, corner) and our perceptual set. Or consider speech perception: it is very difficult for a person to hear mere sounds when she hears her native language spoken (as she may when listening to a language she doesn't know) . She virtually has to hear it as meaningful words and sentences. There is disagreement about how much top-down processing (if any) there is, but most vision scientists endorse some sophisticated descendant of Helmholtz 's view of perception as 'unconscious inference' . Inference Concepts mediate inference. When I see that the creature under the rock is a rattlesnake, I immediately infer that it is dangerous and best avoided. Indeed, many concepts contain mini- (usually defeasible) predictions about what will happen under various circumstances . If that liquid is water, it will quench my thirst, dissolve sugar, extinguish fire, clean the dishes . And many concepts als o contain mini-postdictions . To be legally drunk you have to have consumed alcohol, and many diseases (influenza, measles) require a certain aetiology (as well as pathology) . A large part of problem-solving involves subsuming things under concepts that make the solution easier to spot (Simon, 1 997 is a classic here) . It is easier for most of us to solve problems in analytic geometry if we represent them using concepts involving Cartesian coordinates rather than concepts involving polar coordinates. Experts often differ from novices at solving a problem in their field (e.g. physics , handicapping the horses) precisely because they specify i t i n terms of theoretical concepts (e.g. acceleration, force, energy) that make the problem easier to solve. There are formal models of various sorts of inference, for example, deduction, non-monotonic reasoning, B ayesian c onditionalization, inference to the best



explanation. But things often become murky when w e apply such models t o real life. For example, once we get away from mathematics or computer programming, and logic classes, the 4istinction between deductive inference and inductive inference is often blurry. We can treat Ann 's argument as deductively valid (and hence deductive) if we supply a missing premiss that she might well endorse . Or we can treat it as inductive reasoning based solely on the considerations she mentioned. This fits naturally with the view that there is no useful analytic-synthetic di stinction. Still, some inference is clearly ampliative, leading to conclusions containing information that was not in the premisses or evidence. Here we g o beyond the information given (Bruner), or learn to go on in the same way (Wittgenstein) . This is often called inductive learning, and a great deal of learning, including language acquisition and concept formation, is learning of just this sort. Furthermore, inference is bound up with many other mental processes . For example, many of our explanations of things ( , what could motivate suicide b ombers? ' ) cite causes, and so causal inference plays a central role in explanation (Heit, 2000 surveys the empirical literature on some of these topics). C.I. Lewis remarked that knowing is for doing, and a great deal of thinking involves practical reasoning: deciding what actions are likely to satisfy our desires and how to carry them out most efficiently. Such reasoning ranges from forming intentions to computing traj ectories for the movement of our limbs. Different ways of framing a situation, characterizing it in terms of some concepts (e.g. gains) rather than others (e.g. losses) often influence practical reasoning and decision­ making, - leading, for example, to preference reversals , different content in the retrieval of memories, and so on (Swoyer, 2002 contains an overview) . 2. 1 1

Using Concepts

There are a number features of our use of concepts. I will note several of the more important ones here. ' . : . :...

:, . ;

+�' .

Projectibility We can only use some concepts in our inductive behaviour, for example, in prediction. Adapting Goodman's ( 1 946) predicate ' grue' to our ends, we stipulate that an object is grue just in case it is examined before 20 1 0 and found to be green and is otherwise blue. If all the many emeralds examined to date are green, it is reasonable (subject to the inevitable inductive risk) to conclude that all emeralds are. But now, in 2004, all the examined emeralds are equally grue, but it is not reasonable to conclude that all emeralds are grue. Some or our concepts are projectible, and some aren't. Green is; grue isn't. And the distinction makes a big difference to our inductive practices (try projecting grue and see what people think). Typicality effects In a series of classic experiments Eleanor Rosch and her co­ workers found that if experimental subj ects are asked whether a pictured animal is a bird, they respond more quickly to pictures of some birds (e.g. robins) than to pictures of others (e.g. chickens). In this sense robins are more typical birds than chickens, and such results are now called typicality effects . Typicality has a strong positive correlation with a number of other phenomena of psychological interest. Subj ects make fewer errors classifying typical specimens

1 40

Universals. Concepts and Qualities

than when classifying atypical ones. When asked to recall the different kinds of birds in the pictures, they tend to Llllnk of the typical examples fIrst. When asked to list the typical features of birds they tend to list salient features of robins and crows, rather than of chickens and penguins. Finally, children tend to learn typical examples before atypical ones . Many of these effects occur with concepts like bird, fruit, furniture and colours . S omewhat surprisingly, many of them also turn up with concepts that seem to have precise defInitions (e.g. odd number) . Many philosophers hold that some sentences o r statements, for example, 'bachelors are unmarried adult males' , are analytic (true solely in virtue of the words they contain) . Moreover, many non-philosophers have fairly robust intuitions that some sentences are true by defInition or that some groups of sentences are inconsistent ( 'I don 't know where the witness lied, but he definitely contradicted himself ' ) . These intuitions need t o b e explained, but various facts suggest that i t is worth trying to explain them without appealing to notions like analyticity or synonymy. Among other things, people often believe that words and concepts have precise defInitions even when they can't begin to provide them. Moreover, many of our concepts - for example red, game, chair, sloppy drunk, person - simply seem to lack definitions .

Lexical intuitions

We can combine a fInite stock of concepts to form indefInitely many more complex concepts . For example, we can combine the concepts mediocrity, old, witch and negation to form the concepts mediocre w itch, old mediocre witch, o ld witch who is not mediocre, and so on. This phenomenon is known as productivity. Productivity is important because it allows us to think about indefInitely many things without having indefinitely many concepts . Furthermore, understanding is to some extent a package deal. I cannot really understand some concepts (e.g. old and mediocre) unless I can understand related ones (e.g. mediocre and old) . This phenomenon is known as systematicity. The most natural explanations of conceptual productivity and conceptual systematicity turn on the claim that concepts obey a principle of compositionality. On a compositional account the meaning or content of a complex concept is completely determined by the meanings of its constituents and the way they are combined. Productivity is then explained by the claim that we have a fInite stock of simple concepts and recursive rules for combining them, which allow us to generate an unlimited number of additional concepts. A related story explains systematicity. The notion of compositionality has fragmented into various subtypes , and one can certainly worry about the notion of simple concepts (as opposed to concepts that are treated as simple within a particular theory of content) . Still, it is often held that any theory of concepts must be compositional in some respectable sense, and theories that aren't - and there are a number of them - are often written off on this ground alone (Rips, 1 995 contains a good overview of the empirical literature) .

Productivity and systematicity

Conceptualism 3.


What Concepts Are: Theories of Concepts

I t would b e premature t o w e d conceptualism to any particular theory of concepts , but I will quickly note the maj or theories with an eye to their implications for conceptualism and ability to explain the items on the above list of phenomena. 3. 1

Representations and Operations Are a Package Deal

Before getting down to details, we should note that theories of mental representations and theories of the mental operations that work on them are a package deal. We rarely have observational access to either (even in introspection), and so we postulate representations of a certain s ort (e.g. semantic networks) in tandem with mental operations of a certain sort (e.g. spreading activation) to explain phenomena of a given sort (e.g. retrieval from long-term memory) . It is the entire package that is tested and used to predict and explain. This can lead to real-life underdetermination of theory by data, since we can sometimes save a picture of concepts by tinkering with our account of the operations that process them, or vice versa (e.g. Anderson, 1 97 8 ) . Theories o f concepts are evaluated by the usual standards for empirical theories : simplicity, comprehensiveness and, above all, their ability t o explain the phenomena in their domain. An explanation requires at least two components : first, something to be explained, the explanation target; second, something to explain it. Theorists sometimes disagree about whether something constitutes a genuine target, but most of them view the phenomena noted in the previous section as genuine explanation targets . 3.2

The Classical View

The classical account of concepts dominated philosophy for centuries until, in the mid- 1 950s, it came under fIre from Wittgenstein, Quine, Sellars, Goodman and others. This account was als o popular in psychology until the work of Rosch and her co-workers in the early 1 970s. The leading idea of the classical view is that the content of a compound concept is given by a conjunction of other concepts, and that these provide necessary and sufficient conditions for something's being in the extension of the concept. Aristotle thought of defInitions of species (e.g. human) in this way, as the conjunction of a genus (e.g. animal) and a differentia (e.g. rational) . DefInitions in mathematics often fIt the pattern: a prime natural number is a natural number that is divisible only by itself without remainder. We could generalize the classical account to allow logical operations (like negation and existential quantifIcation) in addition to conjunction by adapting the operations in recent formal accounts of properties (e.g. B ealer, 1982; Menzel, 1 9 9 3 ; Swoyer, 1 997; 1 9 9 8 ; Zalta, 1 98 8 ) to concepts . The classical account is sometimes thought to provide precisely bounded necessary and sufficient conditions for something to fall under the concept, but if one or more of the defIning conditions is vague, the concept may be vague too . For example, our ordinary concept of being an adult is vague, so the concept being an adult unmarried male is vague, and hence the concept bachelor is vague too.

1 42

Universals, Concepts and Qualities

Pros I will call concepts (if any) that conform to the classical model classical concepts and say that they have a classical structure , S ome concepts do seem to be classical, among them bachelor, prime number and deductively sound argument. The classical view can explain many roles of concepts in deduction, many cases of conceptual combination, and many lexical intuitions, for example, the intuition that bachelors are unmarried adult males. ' The account is compatible with typicality effects, since many classical concepts (even prime number) exhibit such effects , and with vagueness, providing the defining concepts are vague. Classical concepts can also be proj ectible, though of course many are not. Cons Despite its simplicity and intuitive appeal, the classical theory, construed imperialistically as a general theory of concepts, is dead. In many cases the anticipated definitions never turned up . There were warnings in Wittgenstein and other philosophers, and empirical work only reinforces the point. We cannot give necessary and sufficient conditions for something's being red, a game, a chair, or beautiful, and in many cases there is little reason to suppose that there are any definitions waiting to be found. Even when s omething resembling a definition is available, there are often problems . First there is the grounding problem: some of the concepts used to define others must at some point hook up with the world outside the mind, and it is not easy to find plausible examples of such concepts . It is als o p ossible to use a classical concept without knowing its full definition and, indeed, even when having false beliefs about it. I used to think that something is a first-degree murder just in case it is a premeditated killing, but I am told that in my home state this is incorrect. The classical theory is a pure case of the view that general terms have descriptive contents that determine their extensions . Hence it isn ' t surprising that the classical view makes too much analytic (since the defining conditions give the meaning of the defined concept) , too much necessarily true (on the c ommon assumption that the equivalence between the two parts of a definition is necessary), and too much a priori (on the common assumption that our knowledge of meanings or definitions is a priori) . 3.3

Similarity and Similarity-Based Accounts

Similarity has been a central notion in philosophy and psychology for a very long time. It is often natural to view the acquisition of a concept, especially one keyed fairly directly to experience (like red), as learning to go on to classify newly encountered items that are similar to the original red things as red too. Similarity is appealingly flexible: one thing may be similar to a second with respect to colour but not shape, and similar to a third with respect to shape but not colour. But flexibility is also similarity 's Achilles' heel. The relation is just too promiscuous ; any two things are similar in s ome ways and not in others (e.g. Goodman, 1 972). The point that any two things are similar in many ways and dissimilar in many others is compatible with the existence and importance of psychological similarities, those that influence how we categorize and think about things. But a good similarity­ based explanation of a cognitive process like classification must specify the relevant



1 43

sorts of similarity. This is often where the real work is required, and the resulting account of what makes for the 'relevant' or the 'right' sort of similarity is what does . the real explanatory work. Study after study s h9WS that our assessments of similarity are highly sensitive to context (cf. Medin et al . , 1 99 3 ; Goldstone et al . , 1 997; Goldstone and Rogosky, 2002) . The features of things that we consider and the weight we give them in determining similarity are strongly influenced by our interests, the s ort of concept involved, and other features of the context. For example, beef is more similar to soybeans than to avocado if I am concerned with protein, but more similar to avocado than to s oybeans if I am concerned with fat. Judgements about relevant similarity in a domain also change as one becomes more skilled in thinking about things in the domain. Part of learning about French impressionism or classical mechanics is becoming attuned to the existence and role of resemblances that escape the novice. There are two theories of concepts that are based on similarity. Prototype theories As with the other theories we' ll examine, prototype theory is a family of accounts. We might motivate the general approach with a story. A child encounters creatures and is told which ones are birds . Eventually she gets the hang of things and can go on, with just the occasional error, to sort newly encountered things into birds and non-birds. How? She counts things as birds just in case they are sufficiently similar, ' in psychologically relevant ways ' , to the birds in her original sample. In the case of concepts like red, which lack obvious internal structure, we learn that certain things are red, and we have a cognitive mechanism that determines whether colours are similar enough to it to count as red too . With more complex concepts, for example, fruit, furniture or kinds of-animals, say birds, we form a mental representation that encodes salient information about typical birds ; this makes up a sort of prototype profile that might include features such as having a beak, feathers , and the ability to fly and build nests . We also have cognitive operations that determine the relevant sorts of similarity for this concept and .compute .how similar various things are; in these respects., to the infDrmation in our bird profile. A key innovation of prototype theories is that prototype profiles are not definitions. Prototype concepts have descriptive content (their profiles), but they do not have defInitions . Prototype theories can explain typicality effects (not surprisingly, since they were designed to) , though this would be more impressive if classical concepts didn' t exhibit the s ame effects . Prototypes theories can also account for vagueness and the fact that we often cannot fInd defInitions. Many prototype concepts are also projectible and natural. However, prototype theories have difficultly explaining many of our lexical intuitions, and they stumble over the fact that we may be able to use a concept like elm even when we know very little about it or its profIle, or even when we are mistaken about it. We can also master concepts that lack prototypes . These include concepts , like that of a Hobbit, which have no instances, though perhaps we could acquire profiles for these by reading Tolkien or seeing the movies . The features that loom large in our assessments of typicality sometimes depend on context (B arsalou, 1 987), though since assessments of similarity depend on context this isn ' t a problem in principle, though it does make things more complicated. And then there is the

Universals, Concepts and Qualities

1 44

common problem about what features · are and how they latch onto the world outside the mind. One of the most persistent criticisms of prototypes is that they are not compositional (the discussion begins with Osherson and Smith, 1 9 8 1 ). To take the standard example, something is a pet fish just in case it is both a pet and a fish. But the prototype of pet fish is not a function (here some sort of conjunction operation) of the contents (here the prototypes) of pet and fish. In our culture most people's prototypes of pets include things like being a dog or a cat, but not fish. And our prototypes of a fish probably involve middle-sized fish like tunas and trout. But the prototype of a pet fish involves something more like goldfish and guppies, which is not in the intersection of the prototypes of pet and fish .

Exemplar theories The fundamental idea behind exemplar theories of concepts is that instead of storing profiles of features (modem analogues of Locke ' s abstract ideas) we store memories of specific instances we have encountered (rather like Berkeley's replacements for Locke's abstract ideas). For example, we store memories of the various birds we have encountered. We then compute similarities between newly encountered items and our exemplars in memory. From a philosophical perspective exemplar theories are akin to philosophical theories based on a general relation of similarity (basically what we find in Camap' s Aufbau) . A problem with such brute resemblance i s that i t is very complicated, context-dependent, and often isn ' t discriminating enough to divide things up into the concepts that we actually have . This can prompt a move toward resemblances of aspects, for example, colour-resemblance, shape-resemblance and the like. Such relations are easier to work with, but from a philosophical perspective they look uncomfortably like properties and from a psychological perspective uncomfortably like prototype profiles. In both cases they presuppose things, the relevant similarities or aspects, that many theorists think need to be explained. In some experiments exemplar models accommodate the data better than prototype theories do, but I won:t follow this up,�becaru;e .ex�mplar; .theories.have many of the same. pr.os ·and · o cons as prototype theories . 3.4

The Theory Theory

Similarity-based views have some plausibility for concepts keyed fairly directly to observation, but they fare worse with concepts like nation-building, global warming and (to cite a standard example) being drunk. There are no simple, easily observable similarities among people who are drunk: there are sloppy drunks, moody drunks , belligerent drunks . And we often need t o mobilize a good deal of knowledge to know when the concept drunk applies. Once we give up on definitions, it can be tempting to build a lot of this knowledge into the concept. On this view a concept is a sort of mini-theory, and we classify things under the concept that best explains them. Sam was drunk; that explains why he slurred his speech and insulted his wife. This approach is often called the theory theory of concepts. The theory theory reminds us that we often need to rely on more theoretical similarities than prototype and exemplar theories usually do. It also accounts for the role of inference and explanation in the application of many concepts . But


1 45

combination of concepts now requires the combination of theories, and no one has a good account of this. There is also the now-familiar point that we can use many concepts despite being ignorant, or even wrong, about them and their extensions, and since the theory theory builds so much information into concepts it is especially vulnerable here. The approach is also too holistic to explain lexical intuitions . Indeed, all sorts of information may be relevant in applying a concept, but how much of it should be built into the concept? Once you start, is there any pi:incipled place to stop? The theory theory makes the important point that classificatory tasks can involve a wide variety of things. But to be a theory (not to mention a computational theory) of concepts it needs to involve more than a sprawling Quinean holism. It needs to tell us what information belongs in a concept, and why. 3.5


In Section 2 we encountered arguments that some concepts , for example, e lm and red, lack descriptive content. On this view concepts are simple unstructured mental representations, cognitive atoms , and the only thing essential to a concept is that it denotes the property that it does (e.g. redness, being an elm, justice) .

Pros Conceptual atomism explains the lack o f definitions for many concepts and accounts for our ability to use a concept like elm without knowing much about elms . It nicely explains conceptual stability, since a concept may remain anchored steadfastly to a property throughout many changes in our beliefs about its instances . It also explains conceptual combination; something is a pet fish just in case it has the properties of being a pet and being a fish. Cons Conceptual atomism allows us to have the concept dog without having the concepts animal, eating, barking and having a head, which strikes many of us as counter-intuitive. It cannot explain lexical intuitions, typicality effects, or anything else that smells like des.criptiYe .c.ontenr, though it might be argued that such things need to be explained by something other than a theory of concepts. A maj or problem for conceptual atomists is to explain how a conceptual representation (e.g. red, elm, justice) hooks up with its relevant property. There are various proposals about this, most involving some sort of causal chain between the property and the concept, but none is without problems . The problem is especially acute for uninstantiated properties, particularly those not definable in terms of instantiated properties, because there are no causal chains running between the natural order (where we and our concepts exist) and abstract properties . 3.6

Conceptual Role Semantics

Conceptual role semantics (sometimes known as procedural semantics) was originally developed as a very general account of linguistic meaning (e.g. Sellars, 1 954; Harman, 1 987), but it is quite natural to reinterpret such an account as theories about the content of concepts . On such accounts the content of a concept is determined by its role in inference and other mental activities like perception and action, by its role in what Sellars calls the evidence-inference-action language

Universals, Concepts and Qualities

1 46

game. As we noted in Section 2. 1 , the application of many concepts commits us to predictions (sugar: it will dissolve if I put it in water) and postdictions (forgery: it was produced by someone other than the person we would normally suppose). Indeed, this is the rule, not the exception. S ome concepts, for example, colour and taste, are geared fairly directly to the environment, and part of the mastery of such concepts includes acquiring the ability to recognize and classify things according to their colour or taste. ' B ut simply being able to respond differentially to red things is not enough to have the concept red. Having the concept involves having at least some of the following dispositions : knowing roughly the sorts of conditions under which red things won ' t look red, being able t o draw various inferences , for example, 'this i s red so it' s coloured' , 'this i s red so i t ' s extended' , 'this i s currently red all over so i t can ' t b e green all over too ' , 'this is scarlet so it's red' . More importantly, though more vaguely, it requires us to be able to think in a variety of ways about red things and about colours . Other concepts are more theoretical. Our concept of causation is b ound up with our ability to determine what causes what in many everyday circumstances, how dispositions of things lead them to cause certain things under certain conditions , for example, if the piece of metal is magnetized it will attract the iron filings, and knowing what sorts of evidence will support causal attributions, having a sense about how causation and counterfactual conditions are connected, and so on. Despite occasional claims to the contrary, there is nothing that requires conceptual roles to be entirely in the mind (as 'narrow content' ) , and they certainly weren' t thought o f i n that way by either S ellars o r Harman. M y concept red involves the ability to recognize red things out in the world. My concept water links up with water because of the way I apply the term 'water' to it. S ome philosophers supplement the view that the content of concepts is determined by their conceptual or psychological role with the view that concepts also have a directly referential component. Th is is always an..option, but it' s not how early theorists like S ellars or Harman­ thought about things ; indeed, Sellars urged that there is no single relation of reference, though there are many relations between concepts and things in the world that anchor our framework of concepts to the world outside the mind . There i s also nothing that commits the conceptual role theorists to either analyticity, on the one hand, or to an amorphous holism, on the other. S ome inferences and beliefs involving a concept are more central to the concept than others. Philosophers have offered various accounts of centrality, though I suspect it has a variety of sources . At all events , I ' ll simply note the intuitive point that it is more central to my concept dog that dogs are animals than it is that no dogs have over a trillion hairs. 3. 7


It has become increasingly obvious that concepts have been asked to fill too many roles that are often incompatible, and that no existing accounts account for all of them. We can divide these roles, provisionally, into three general kinds.


1 47

Role one: anchoring One role clusters around stability, conceptual combination, truth conditions , and lack of definitions . We might call this the ancho ring role. It also fits with many aspects, translated into the idiom of concepts, of direct theories of reference and externalism in semantics . The idea is that concepts directly refer to things in the world, and for a general term it is difficult to see what this could be other than a property (we might hope, though, that in re properties would be enough) . Role two: conceptual roles We just discussed cognitive or inferential role. It involves the cognitive functions of concepts discussed in S ection 2 . 1 . Role three: other psychological phenomena These include some o f the phenomena and effects discussed in S ection 2.2, including typicality effects and ease of classification. It' s an oversimplification, but the conceptual role of a concept is reminiscent of Marr's ( 1 9 82) picture of the function that a given bit of cognitive processing computes, while things like typicality effects may tell us something about how we go about computing it, what algorithm the mind uses. This might involve the ways we identify instances of a concept. Again, we might compute similarity in different ways, for example, by deploying rules that count and weigh features, or by more associationistic connectionist mechanisms . Again, it is part of the conceptual role of concepts like dog and all that if all dogs are animals and if Max is a dog, then Max is an animal. But we might work this inference out using some mental analogue of formal logic or we might use inheritance in a hierarchy of c oncepts . For example, it might be part of our network of concepts that dogs are a kind of animal and that Max inherits the feature of being an animal from the fact that he is a dog. What are our options when a single entity seems unable to meet all of the demands placed on it? First, we can introduce different kinds of entities, one to play each type of role. Bealer's ( 1 9 82) distinction between qualities an.d concepts is an imp()rtant example of this in philo sophy. In psychology we can posit various types of mental representations in addition to concepts, for example, semantic networks, B ayesian networks , or scripts. We can also introduce different kinds of concepts, for example, classical concepts , prototype concepts and so on (cf. Medin et aI. , 2000; Medin and Coley, 1 998). But these distinctions don't really solve the present problem, which is trying to find things that play the various roles in one and the same case; we want to explain the anchoring aspect of the concept water and its functional role in inference. Moreover, if we go down this route we will face difficult questions about how the various entities connect up with each other. Accounts that posit multiple aspects of a single sort of entity, just concepts, have been more popular. There are several theories of this sort going back for over twenty years (e.g. Osherson and Smith, 1 9 8 1 ; 1 9 82; Miller and Johnson-Laird, 1 976). The approach has become more sophisticated over the years . In his recent discussion Rips ( 1 995) argues that concepts have two aspects : the representation of a category of things, which is an atomistic directly referential aspect (to provide stability and related things); and, a representation about the category, in his case a

Universals, Concepts and Qualities

1 48

version of the theory theory (to include the descriptive content that is often relevant in applying a concept and that helps explain some of the other phenomena mentioned above). There are several more degrees of freedom here. First, a number of recent theorists have proposed dual-process accounts of c ognition. Different theorists develop this idea in different ways, but the rough idea is that human beings have two quite different cognitive subsystems (or two types of subsystems) . There is an ' explicit' subsystem that is largely conscious, symbolic, verbal, rule-governed, serial, flexible and capable of reflection. B ut there is also an 'implicit' subsystem that is largely non-conscious , associative, impulsive, affective and that reacts automatically to stimuli (e.g. Sloman, 1 996; Chaiken and Trope, 1 999). We further increase the possibilities for multiplication if we hold that the brain houses various domain-specific modules that involve different sorts of processing. Various combinations are possible . The basic trick is to put enough in the head that concepts can play their causal, psychological roles, while having them anchored tightly enough to the world outside our heads that they can be stable over time and among people. My own preference is for an account of concepts with two aspects . The first aspect is conceptual role; the second psychological phenomena like those discussed above. But however this may be, there are enough degrees of freedom that we can play a number of things off against one another in a way that leaves empirical theories of concepts more than a little underdetermined by the evidence.


Concepts and the Problems of Universals

Universals have been invoked to explain a wide range of phenomena. I will note some of these, say how properties have been used in attempts to explain them, then ask how well concepts might do instead. The list is by no means complete and not all of the targets are equally compelling, but many are of current interest. I have argued that argl.JT]1ents . in ontology are best construed as ampliative (l 98J.; .1999a),. but most of what follows is straightforwardly adapted to the view that philosophical arguments should aim to be deductively sound. I will divide the explanatory targets into four group s . 4. 1


Traditional ontology S ome things are alike in certain respects : they are human beings, red, loud. And various groups of things can stand in similar relations , for example, loving one another. Possession of a shared property (humanity, redness, loudness) or standing in a shared relation (loving) have been said to explain such resemblances , while possession of different properties explains the differences. Furthermore, many things change over time. The table was brown before I painted it, but is white now. The table itself is present throughout the change in colour, so the mere ontological blob of a table can't explain the alteration. For that we must add that it exemplified b rownness yesterday and exemplifies whiteness today. Such accounts have struck some philosophers as pseudo explanations, and it is fortunate for realists that properties can be defended on the grounds that they explain


1 49

other phenomena that more clearly need explaining. Still, the ability to explain qualitative recurrence and change is a perennial motivation for realism (e.g. Plato 's Republic, 507b ; 596a-b ; Armstrong, 1 9 84, p . 250). What about concepts ? Assuming some constraints imposed by the world outside the mind, the conceptualist can argue that the resemblances that seem natural to us only seem so because we are disposed, through our biological or cultural or linguistic heritage, to find s ome groupings more natural than others . Moreover, everything changes constantly in all sorts of ways. The ways that matter to us, the ones we tend to think of as 'real change ' , are likewise determined by the concepts and similitudes we find natural.

Recent ontology Recently some philosophers have argued that universal properties provide the best explanations of various features of measurement, causation, natural laws, supervenience, and the like (Swoyer, 1 999b contains an overview). In many of these cases properties play a central role in making things happen. The liquid in the glass caused the litmus paper to turn blue because the liquid is alkaline. But the fact that the liquid falls under the concept alkaline doesn' t explain why it turned blue. Concepts are not well suited to explain such things (though with luck in re properties might be enough for the j ob). 4. 2

Cognition and Epistemology

We have already noted the numerous roles concepts play in cognition, culture and the like.

Recognition and classification Most animals, including humans, are constantly classifying, sorting, categorizing. Many philosophers have argued that an organism's ability to recognize and classify newly encountered things such as cats , red obj ects and unwarranted insults is best explained by the view that the old instances and the new share a property . (e g . , redness), and,thatthe organism is S9:r:neJ1,Ow.attunes.l W recognize or respond to it. B ut if there is a plausible conceptualist metaphysical story about an objective basis in the world outside the mind that provides a foothold for classification, concepts may be able to explain the phenomena as well as properties. The idea here is to shift some of the burden from 'reality itself' to the realm of concepts . ..

Inductive learning and inductive bias Language learning, concept formation, mastering social norms and moral principles, calculating sums , projecting predicates, giving causal explanation and so on involve inductive learning. We are exposed to instances (e.g. grammatical sentences, typical members of the extension of a concept) and learn to classify new and novel things correctly (the p oint is a generalization of so-called poverty of stimulus arguments in linguistics; cf. Morgan and Demuth, 1 996; Chomsky, 1 975). The problem is that there are too many different, incompatible, ways one might go on. I see an object that is, among other things, green. Should I classify it as green, or grue, or being grue or else a prime number? The list is endless . If we are to succeed we cannot be completely open-minded. We need a strong predisposition

1 50

Universals, Concepts and Qualities

( ,inductive bias ' ) to classify things one way (e.g. as green) rather than another (e.g. as grue) . A cognitive system without inductive biases cannot learn. S o although we need some basis in the world for our application of concepts, we also need a strong psychological disposition to classify in some ways rather than others. Since classification often involves thought without languages (as in learning our first language), this is very naturally explained by the view that we come into the world with certain concepts (e.g. the obj ect concept, various concepts needed to learn language) or at least with strong dispositions to form these concepts (given the dispositional nature of our use of concepts, thes e may not be as different as they at first seem) . In short, concepts and conceptual operations, ones involving concept acquisition and use, play a central role in inductive learning. A priori knowledge and de dicto necessity S ome philosophers , including several British empiricists , have argued that a priori knowledge is knowledge of the necessary connections among concepts (the concept bachelor includes or entails the concept male) . One might add that de dicto necessity, the necessity of sentence­ like creatures such as the thought that bachelors are unmarried, is also based on necessary relations among concepts . Perhaps, indeed, ' the only necessity is conceptual necessity ' (C.I. Lewis thought something like this) . It is not clear that there really is much a priori knowledge or necessary truth, but those who think there is might find concepts useful here. 4.3

Language and Semantics

Natural language and cognition involve some of the s ame things, for example, intensionality. The overlap is especially obvious if we think cognition takes place in a 'language of thought' .

Semantic values of abstract singular terms Abstract singular terms like ' avarice' , 'triangul