Semantics and Conceptual Change

Citation preview

Semantics and Conceptual Change Jerrold J. Katz The Philosophical Review, Vol. 88, No. 3. (Jul., 1979), pp. 327-365. Stable URL: http://links.jstor.org/sici?sici=0031-8108%28197907%2988%3A3%3C327%3ASACC%3E2.0.CO%3B2-9 The Philosophical Review is currently published by Cornell University.

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/journals/sageschool.html. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

JSTOR is an independent not-for-profit organization dedicated to and preserving a digital archive of scholarly journals. For more information regarding JSTOR, please contact [email protected].

http://www.jstor.org Fri May 18 08:52:34 2007

SEMANTICS AND CONCEPTUAL CHANGE* Jerrold J. Katz

T

he discussion of conceptual change that has taken place over the last two decades is peculiar in a n important respect: meaning and change of meaning loom large in these discussions, yet no consideration seems to have been given to what linguistics says on these topics. It is almost perverse that in discussions of science carried out by philosophers of science, in which highly controversial claims concerning science turn on assumptions about meaning, there is no attempt to use ideas from the science that studies meaning.' In ignoring linguistics, these discussions restrict themselves to an overly narrow range of positions on conceptual change in science. In the present paper, I will show that linguistics has a significant contribution to make in extending this range of positions. I will show that there are ideas in linguistics that lead to a new position on con-

*This paper has been presented at the Conference on Conceptual Change at Ripon College, and to the philosophy departments at the University of Toronto, Columbia University, University of Massachusetts in Amherst, and the State University of New York at Albany. I wish to thank Joe Bevando, Tamara Horowitz, Paul Horwich, Arnold Koslow, Keith Lehrer, Sid Morgenbesser, Tom Nagel, Charles Parsons, and Peter Unger for their comments on earlier drafts. I wish to specially thank Virginia Valian for her helpful comments on the final draft. .-- ..~ -~. -.I Ironically, the closest thing to such an attempt in the literature is a neglected, early piece by Putnam (Putnam, H., "How Not to Talk about Meaning," in Boston Studies zn The Philosophy of Science, Vol. 11, R. S. Cohen & M. W. Wartofsky, eds., Humanities Press, New York, 1965, pp. 205-222) in which he defends the ordinary distinction between meaning and belief against Feyerabend's conception of theory-relative meaning. What is ironic, as we shall see below. is that it is now necessarv to defend the same distinction against Putnam. It is also worth mentioning for the record that ~eyerabend'sreaction was to make the pronouncemen; that "As far as I am concerned, even the most detailed conversations about meanings belong in the gossip columns and have no place in the theory of knowledge." (p. 230) ' Shapere, D. "Meaning and Scientific Change," in Mind and Cosmos: Essays in Contemporary Science and Philosophy, Vol. 111, University of Pittsburgh Series in the Philosophy of Science, University of Pittsburgh Press, 1966, pp. 41-85. -

-

J E R R O L D J. KATZ

ceptual change in science. I will exhibit its advantages over the positions taken thus far, and I will show how it might clarify disputes in the philosophy and history of science. The discussion of conceptual change has been carried out largely as a debate between the "new philosophers of science," as Shapere refers to Kuhn, Feyerabend, and Toulmin, and the older, logical empiricist philosophers of science.' Philosophy of science, on the logical empiricist conception, is "metascience," on analogy to metalogic. It concerns, as Shapere characterizes it, the "logical form" . . . of scientific statements rather than their "content," with, for example, the logical structure of all possible statements claiming to be scientific laws, . . . with the logical skeleton of any possible scientific theory, rather than with particular actual scientific theories.'

This aim is to be achieved using the techniques of modern mathematical logic in approaching . . . problems . . difficulties were to be overcome . . . by giving a more satisfactory reformulation in terms of that logic."

The particular choice of applied predicate calculi for such reformulation is guided by an empiricist metaphysics which claimed that all scientific theory must, in some precise and formally specifiable sense, be grounded in experience, both as to the meanings of terms and the acceptability of assertions.

Scientific theories were conceived of as axiomatic systems connected to experience through rules that interpret their theoretical vocabulary exclusively in terms of an observation vocabulary. Scientific development is a process of stockpiling truths and weeding out falsehoods.

' Ibid., p. 42. Ibid., p. 42. Ibid., p. 44. Ibid., pp. 44-47. See, for example, Hempel, C. G., "Problems and Changes in the Empiricist Criterion of Meaning," Revue Internationale de Philosophie, Vol. IV, No. 11, 1950, pp. 41-63, and Hempel, C. G., "Implications of Carnap's Work for the Philosophy of Science," in The Philosophy of RudolfCarnap, A. Schilpp, ed., Open Court Press, 1963, pp. 685-710.

SEMANTICS AND CONCEPTUAL CHANGE

A number of things prepared the way for the rebellion against this "old philosophy of science.'' The four most significant seem to be these.6 First, the old philosophy of science lost credibility because it failed to make good its promise to eliminate metaphysics through the logical analysis of language, especially to explain how the meaning of theoretical terms can be given in .~ the emphasis on terms of an observation v ~ c a b u l a r y Second, "the logical skeleton of any possible theory" made the work of these philosophers increasingly less relevant to "actual scientific theory." Third, Wittgenstein's criticism of the ideal of rational reconstruction undermined the belief that it actually removed problems. Lastly, history of science emerged as a "new professionalized discipline," with results challenging the logical empiricist's "upward and onward" picture of scientific development.' The new philosophers of science earned this title by first formulating an alternative conception of scientific theory and scientific development. Its principal feature was a picture of scientific development in which periodically sciences experience revolutions that so radically redefine them that the normal scientific tradition that emerges . . . is not only incompatible but often actually incommensurable with that which has gone b e f ~ r e . ~

The emerging theory does not refute the old theory, it replaces it. Since the replacement can so fundamentally transform the science that no common methodological basis exists on which to compare the claims in the old and new theories, scientific development is not cumulative. The significance of meaning and change of meaning for scientific development can be appreciated from the fact that both the cumulative and the noncumulative pictures of scientific development are grounded in accounts of meaning. Logical empiricists ground their picture in an empiricist account of "Meaning and Scientific Change," pp. 46-48. T. S., The Structure of Scientfic Revolutions, University of Chicago Press, Chicago, 1962, p. 102. There is a difficulty with Kuhn's formulation here. Since he intends "incommensurability" to apply when the theoretical terms in the old and the emerging traditions are- interpreted so that the basic principles of these traditions are logically independent, the two traditions cannot be "incompatible" when they are "incommensurable". I will assume that Kuhn's reference to incompatibility is just a slip.

' Kuhn,

JERROLD J. KA TZ

meaning. The account makes it reasonable to think that there is a theory-neutral vocabulary of observation terms that provides theory-independent, experiential meanings for the theoretical terms of all scientific theories. Thus, there is a basis for claiming that the same meaning can be assigned to a term in different theories and that translatability between theories is always possible. The new philosophers of science, too, ground their picture in an account of meaning. The failure of the logical empiricists to vindicate the empiricist account of meaning enabled the new philosophers of science to replace the empiricist account with their account of meaning on which there is no theory-neutral basis for interpreting scientific terms. Such terms are idiosyncratically interpreted in the theories employing them and their semantics is relative to that theory. Feyerabend claims: the meaning of every term we use depends upon the theoretical context in which it occurs. Words do not 'mean' something in isolation; they obtain their meaning by being part of a theoretical system."

Hence, if the new theory emerging from a revolution is sufficiently different from the old in respect of the interpretation of terms, the theories, not surprisingly, turn out incommensurable. The controversial claims of the new philosophers of science about scientific theory and scientific development thus depend on replacing the logical empiricist's absolutist account of meaning with an account on which meaning is theory-relative. Given that there is no a priori reason to suppose that these two accounts of meaning are the only possible accounts, the question arises of why philosophers haven't tried to formulate an absolute semantics that does not rest on the logical empiricists' empiricism and its dubious theoretical term/observation term distinction." An answer is not hard to find. The new philosolo Feyerabend, P. "Problems of Empiricism," in Beyond the Edge of Certainty, R. Colodny, ed., Prentice-Hall, Inc., Englewood Cliffs, 1965, p. 180.

" Even as resourceful a critic of the new philosophy of science as Shapere tries instead to deemphasize the impoftance of meaning. (See "Meaning and Scientific Change," pp. 68-69). Shapere claims that there is a "danger" in making use of the notion of meaning. He says: "-we expose ourselves to the danger of relegating some features of the use of a term to the 'less important' status of not being 'part of the meaning.' Yet those very features, for some purposes, may prove to be the very ones that are of central importance in

SEMANTICS AND CONCEPTUAL CHANGE

phers of science's position on meaning has broad support in the history and philosophy of science. The position that meaning is theory relative has the authority of one of the most widely respected figures in the history and philosophy of science, Pierre Duhem. Duhem wrote: According to whether we adopt one theory or another, the very words which figure in a physical law change their meaning, so that the law may be accepted by one physicist who admits a certain theory and rejected by another physicist who admits some other theory."

The position is encouraged by other influential figures like N. R. Campbell. Also, putting the issue of the theoretical/observational vocabulary distinction to one side, there is not all that much different about the old and new philosophies of science. In Carnap's case, this is clear just from the unrestricted ontological freedom we are given in answering so-called "external questions." The similarity between the old and new philosophy comparing two uses, for relative importance of features of usage must not be enshrined in a n absolute and a priori distinction between essential and inessential features. It thus seems wiser to allow all features of the use of a term to be equally potentially relevant in comparing the usage of the terms in different contexts. But this step relieves the notion of meaning of any importance whatever as a tool for analyzing the relations between different scientific 'theories.' " First, the distinction between "part of the meaning" and "not part of the meaning" is not the distinction between "essential" and "inessential." But even if it were, it is surely just an equivocation on these terms to take them, as Shapere does, to carry the force of, respectively, '"important" and "unimportant." There is no reason to suppose that, because some feature is not part of the meaning of a term or not essential, it is less important than some feature that is part of the meaning of the term. There seems to be widespread confusion among first-rate philosophers of science over the difference between definitional and essential properties. For instance, N. R. Campbell's criticism of the distinction between defining and nondefining properties of scientific terms rests on such a confusion. Campbell poses the question as follows: "Are there properties of silver which simply define what we mean by silver and such that, if they were altered, the substance would not be silver; and are there on the other hand nondefining properties, such that they might be changed without affecting the fact that the substance in question is silver?" Campbell, N. R., Foundations of Science, Dover Publications, Inc., New York, 1957, p. 47. But properties such that if something did not have them, it would not be the same thing (p. 47) are essential properties but not necessarily definitional properties. Being the single even prime is a property such that nothing without it can be the number two, but it is not a definitional property of two. I' Duhem, P. The Aim and Structure of Physical Theory, Princeton University Press, Princeton, 1954, p. 167.

J E R R O L D J. KA TZ

also follows from other, more widely held, doctrines of Carnap's.13 Moreover, the position of the new philosophers of science is significantly strengthened by the hostility of present philosophical opinion toward the traditional notion of meaning. The effect of this hostility is to prevent us from restricting principles for interpreting terms from entire systems of beliefs to analytic beliefs. As Quine put it: Once I reject the distinction between analytic sentences and other communitywide beliefs, however, my nearest approximation to a null theory is the class of all community-wide beliefs. l 4

Although Quine's attack on the analytic-synthetic distinction initiated this hostility, it has been kept up, on the one hand, by followers of Quine's like Davidson and Harman,15 and on the other, by philosophers coming to the issue of the separability of language and theory from the study of reference like Putnam and Kripke. l6 With widespread opposition to distinguishing the linguistic component of a theory from the rest of the theory as that part which interprets the terms of the theory, it is easy to see why the account of meaning of the new philosophers of science ran into so little trouble. Since their relativization of meaning to theory was the real force behind the radical doctrines about the no;cumulativeness of scientific development, their conformity to orthodox opinion on the language/theory distinction shielded l3 See Carnap, R., "Empiricism, Semantics, and Ontology," in Semantics and the Philosophy of Language, L. Linsky, ed., University of Illinois Press, Urbana, 1952, pp. 208-230. See English, J., "Partial Interpretation and Meaning Change," The Journal of Philosophy, Vol. LXXV, No. 2, 1978, pp. 57-76 for an argument that widely accepted logical empiricist doctrines are close to the doctrines of the new philosophers of science. lJQuine, W. V. "Reply to Chomsky," Synthese, Vol. 19, No. 1/2, 1968, p. 282. j5 Davidson, D. "Truth and Meaning," Synthese, Vol. 17, No. 3, 1967, pp. 304-323; Harman, G. Thought, Princeton University Press, Princeton, 1973, pp. 84-111. l6 Putnam, H. "It Ain't Necessarily So," T h e Journal of Philosophy, Vol. LIX, No. 22, 1962, pp. 658-671; "Is Semantics Possible?", Metaphilosophy, Vol. 1, 1970, pp. 189-201; "The Meaning of 'Meaning,"' Language, Mind, and Knowledge, Vol. VII, Minnesota Studies in the Philosophy of Science, K. Gunderson, ed., University of Minnesota Press, 1975, pp. 131-193. Kripke, S. "Naming and Necessity," Semantics of Natural Language, D. Davidson & G. Harman, eds., D. Reidel Publishing Co., Dordrecht, Holland, 1972, pp. 253355.

S E M A N T I C S A IVD CONCEPTUAL C H A N G E

these doctrines from the most basic form of criticism to which they are open. l7

With so widespread a skeptical attitude toward the language/ theory distinction, there are surprisingly few serious arguments to show that the distinction cannot be drawn. To my knowledge, there are only three. In this and the next two sections, I shall show that all three are inadequate. Then I will develop an account of meaning and of the language/theory distinction that is absolutist like the logical empiricist account but makes none of the dubious assumptions that got the latter account into trouble. Finally, I sketch the implications of my account for conceptual change in science. Perhaps the best known attempt to argue that the language/ theory distinction cannot be drawn is Quine's argument that meaning and related notions cannot be explained well enough for them to qualify for use in scientific theories about language. Feyerabend's less well known argument that in a conflict between two different high level background theories there is in principle no possibility of translation comes to much the same thing as Quine's indeterminacy of translation.'' Furthermore, Feyerabend's account of how different high level background theories are comparable in terms of his "pragmatic theory of observation" is quite similar to Quine's account, in his theory of stimulus meaning, of how different languages confront the same experience. lg Thus, I will take Quine's argument as representative of Feyerabend's and others' attempts to establish that there is no truth of the matter about questions of meaning and synonymy. As will be recalled, the argument in "Two Dogmas of Empiricism" begins by considering three definitional ways of explain" For my response to the Putnam and Kripke papers, see Katz, J. J. "A Proper Theory of Names," Philosophical Studies, Vol. 31, No. 1, 1977, pp. 1-80. The present paper is the discussion promised in footnote 60 (p. 78). '"ee Shapere's discussion of Feyerabend'i claim in "Meaning and Scientific Change," pp. 57-58; Quine: W. V. Word and Object, M . I . T . Press, Cambridge, 1960, pp. 26-79. l9 "Problems of Empiricism," pp. 217f; Word and Object, pp. 80f.

ing semantic notions: lexicographical definition, rational reconstruction, and notational abbreviation." Quine rejects all three as incapable, in principle, of clarifying such notions. Quine then takes up two theoretical ways of explaining these notions: interchangeability and semantical rules. He tries to show that these methods of clarification also can be dismissed because they lead to circularity. Quine concludes that there is no analytic-synthetic distinction to be drawn. The argument to this conclusion has a logic that has eluded both Quine's critics and his sympathizers. Grice and Strawson commit the original sin in suggesting that Quine's argument proceeds invalidly from cases where the analytic-synthetic distinction has not been adequately explained to the conclusion that there is nothing to explain." Putnam follows them and suggests that there must be something deeper to Quine's argument, otherwise "it is puzzling why this [argument] is supposed to be a good a r g ~ m e n t . "The ~ ~ logic of Quine's argument is this. Quine's essay has three polemical sections: "Definition," "Interchangeability," and "Semantical Rules." In the first, Quine considers explicit definition in all its forms, and in the second and third, he considers implicit definition in the two relevant areas of science, linguistics and logic. The point that Grice, Strawson and Putnam m'iss is that Quine's argument exhausts the relevant areas where a method for clarification of semantic notions might exist. Thus, if he can show that there is no method that can do this in one of these areas, it is reasonable to think the notion of meaning is not clear enough to ground a theory of a priori truth. Quine's dismissal of explicit definition is unobjectionable: notational abbreviation rests on arbitrary stipulation and both rational reconstruction and lexicographical definition assume prior synonymy relations instead of explaining them. Furthermore, Quine's dismissal of implicit definition in logic, namely postulate or rule specification, modeled on definitions of logical ''Quine, W. V. From a Logical Point oJ View, Harper & Row, Inc., New York, 1953, pp. 20-46. Grice, H. P. and Strawson, P. F. "In Defense of a Dogma," The Philosophical Review, Vol. LXV, No. 2, 1956, pp. 141-158. 22 Putnam, H. " 'Two Dogmas' Revisited," in Contemporary Aspects of Philosophy, G. Ryle, ed., Oriel Press, London, 1976, p. 203.

SEMANTICS AND CONCEPTUAL CHANGE

truth in first orde; logic, is also unobjectionable: specification of meaning postulates or semantical rules enables us to "understand what expressions the rules attribute analyticity to, but we do not understand what the rules attribute to those expresthe point that has been missed, Quine's argus i o n ~ . "Given ~ ment would go through if his treatment of implicit definition in linguistics were sound. Putnam compounds the original sin by misrepresenting Quine's argument that synonymy is hopelessly unclear. Putnam says, "The only evidence that Quine produced to support this remarkable claim was that he, Quine, could not clarify the notion in a few pages."24 In those "few pages," however, Quine mounts a very sophisticated argument against synonymy. Putnam's reconstruction of Quine's argument as a n induction from a sample of one does the argument a n injustice. Quine's argument is rather this: Quine poses the problem of clarifying a notion of synonymy in linguistics as that of showing that it is possible to separate synonymous pairs of expressions from nonsynonymous pairs on the basis of an interchangeability or substitution test. H e then observes that either the corpus containing the "substitution frames" is intensional, because it contains expressions like "Necessarily, bachelors are bachelors," or extensional. If the corpus contains such intensional contexts, the question has already been begged. If the corpus does not contain them, every attempt to state a n interchangeability test must beg the question. Such a test must cite the property to be preserved when and only when synonyms replace synonyms. But since the choice is between truth and analyticity, no acceptable test can be formulated. Requiring the substitution to preserve truth fails to separate synonymous pairs of expressions from nonsynonymous but coextensive pairs while requiring analyticity assumes meaning is unproblematic and begs the question. This argument. would show that the central notions of the theory of meaning cannot be made clear enough for linguistics I f Quine can legitimately assume that interchangeability tests are the proper criteria for determining what is clear enough for this science. Though Quine's assumption is mistaken, he had

" From a Logical Point 24 "

of View, p. 33. 'Two Dogmas' Revisited," p. 204.

J E R R O L D J. K A TZ

good grounds for making it a t the time. T h e assumption came from the then dominant theory of language in linguistics, the taxonomic theory of grammar. This theory not only had the authority of scientific orthodoxy, but, being behaviorist, empiricist, and physicalist, it also had the proper philosophical creden'~ himself is quite explicit about appealing tials for Q ~ i n e . Quine to the taxonomic theory of grammar for a conception of the way that linguistics explains linguistic notions: So-called substitution criteria, or conditions of interchangeability, have in one form or another played central roles in modern grammar [viz., taxonomic theory]. For the synonymy problem of semantics such a n approach seems more obvious still. '"

Quine also is quite explicit about the fact that it is this model of explanation that he used in his examination of the notion of synonymy. 27 T h e true reason why Quine's argument fails is that it had the ground cut from under it by the dissolution of the taxonomic theory of grammar. Logically speaking, the "vicious circle" shows no more than that either notions like synonymy cannot be made sufficiently clear or the standard of clarity used in judging them is not adequate. It is plausible to think that these notions are inherently unclear in linguistics if it is, in fact, plausible to think that the standard is adequate. But there is no valid inference to the conclusion that the central notions in the theory of meaning are hopelessly unclear f t h e assumed standard of clarity is discredited. Since Chomsky's refutation of the taxonomic theory as an acceptable scientific theory of language establishes the inadequacy of substitution tests in linguistics, Quine's argument is a casualty of Chomsky's revolution in linguistic^.^^ Not only does Chomsky's revolution bring about the collapse of Quine's argument against meaning, it introduces a new conception of the proper way to clarify linguistic notions. Taxo'Wompare remarks of Bloomfield, the founder of the taxonomic theory in linguistics, such as those in Bloomfield, L. "Language or Idea?" Language, Vol. 12, 1936, p. 93 with those of Quine's,such as From a Logical Point of View, p. 48. 'Tram a Logical Point of View, p. 56. 27 Ibid., pp. 56-57. See Katz, J. J. "Logic and Language: An Examination of Recent Criticisms of Intensionalism," in Language, Mind, and Knowledge, pp. 36-63.

SEMANTICS AND CONCEPTUAL CHANGE

nomic theory conceives of grammars as data-cataloguing devices, compact tabulations of the distributional regularities in a sample of speech. It thus conceives of clarification in terms of the operationalism of substitution tests. Chomsky conceives of grammars as idealizations of the knowledge of a language that speakers exercise in verbal behavior. This makes linguistic theorizing like theorizing in the advanced sciences. It thus legitimizes in linguistics the same standard of clarity employed in the advanced sciences. Notions of any sort can appear in the rules of a grammar so long as, first, they can be precisely stated, and second, their appearance can be justified as a theoretical posit, that is, as essential to the predictions and explanations of the simplest theory of the phenomena. In linguistics, this means accounting for grammatical properties and relations. Furthermore, Chomsky's theory contained a new way of defining syntactic properties and relations that could serve as a model for definitions of semantic properties and relations. The crux of the idea was to set up the syntactic component of a grammar as an explication of the syntactic competence of the ideal speaker-hearer, to explicate this competence in terms of syntactic rules that generate syntactic representations of sentences, and then to define syntactic properties and relations on the output of these rules. For example, a string of words is marked as well-formed just in case it has a syntactic representation; two strings are marked as constituents ofthe same syntactic type just in case they are categorized in the same way in syntactic representation. O n this model, semantic theory was set up as a theory of the semantic competence of the speaker. Semantic theory specifies semantic rules and definitions of semantic properties and relations stated in terms of their output. The semantic rules consist of a set of dictiona7y rules and a projection rule. The dictionary rules assign formal semantic representations to lexical items as descriptions of their senses. The projection rule combines semantic representations of lexical items to form semantic representations of the compositional meaning of syntactically complex constituents. Semantic properties and relations are defined, in analogy to syntactic properties like well-formed, on the output of these semantic rules. For example, a sentence

JERROLD J. KA TZ

or expression in a sentence is marked as meaningfiul just in case it is assigned at least one semantic representation, meaningless just if it is assigned none, and ambiguous just in case it is assigned two or more. Two expressions or sentences are marked as synonymous (on a sense) just in case they are assigned the same semantic representation. zs Given the possibility of a semantic theory, Quine's thesis of indeterminacy of translation has nothing to support it. Hypotheses about the semantic component of a grammar of a language can be justified by checking their predictions about the meaningfulness, meaninglessness, ambiguity, and other semantic properties and relations of sentences against the judgments of speakers of the language. Evidence about the semantic properties and relations of expressions in languages can be used to confirm the existence of synonymy or translation relations, even in the case of radical translation. Bilingual speakers can, in principle, provide evidence, both direct and indirect, about whether or not, for example, "gavagai" has the same sense as "rabbit" or "rabbit stage" or "undetached rabbit part." We can ask such speakers for their judgment about whether "gavagai" is synonymous with any of these three English expressions. We may get a clear, unequivocal response. We may not. If we do, we have evidence. If not, we can ask a somewhat more sophisticated question or a somewhat more sophisticated informant. For example, we might come back with the question whether