Charity, Interpretation, and Belief

  • 67 139 10
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Charity, Interpretation, and Belief Colin McGinn The Journal of Philosophy, Vol. 74, No. 9. (Sep., 1977), pp. 521-535. Stable URL: The Journal of Philosophy is currently published by Journal of Philosophy, Inc..

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at 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 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]. Sat May 12 00:13:42 2007



So much the worse for Bayes' theorem as applied to epistemic conditionalization.Wf Levi's principles, it is the principle of confirmational conditionalization-what I have lately called "epistemic conditionalization"-that must go. I have drawn attention to Levi's condition (2b), since that embodies the multiplication axiom, and it is the combination of the multiplication axiom and the principle of confirmational conditionalization that, o n my view, leads to trouble. Confirmational conditionalization imposes a certain interpretation on conditional Q-functions, and, o n this interpretation, I claim, (2b) fails. Condition (2b) holds when conditional Q-functions are defined as ratios of absolute Q-functions; credal conditionalization is satisfied. But, if we interpret conditional Q-functions that way and absolute Q-functions my way, the principle of epistemic confirmational conditionalization no longer seems to hold of them. HENRY E. ILYBURG, JR.

University of Rochester


T is generally agreed that a principle of charity should play some part in regulating the project of radical interpretation. But it is a question what status such a principle enjoys. Donald Davidson has urged that charity with respect to the beliefs and sayings of others is a sine qua non of successful translation; more, that unless we see to it that veracity preponderates in a creature's attitudes and utterances we cannot construe its behavior as that of a rational agent or psychological subject. T h u s he says: Since charity is not an option, but a condition of having a workable theory [of radical interpretation], it is meaningless to suggest that we might fall into massive error by endorsing it. Until we have successfully established a systematic correlation of sentences held true with sentences held true, there are no mistakes to make. Charity is forced on us;-whether we like it or not, if we want to understand others, we must count them right in most matters.1 4 T h e distinction between two sorts of conditionalization has also made its appearance in statistical literature. See James S. Williams, "An Example of the Misapplication of Conditional Densities," Sankhya, XVIII (1966): 297-300, and R. J. Buehler, "Some Validity Criteria for Statistical Inference," Annals of Mathematical Statistics, xxx (1959): 845-863. 1"On the Very Idea of a Conceptual Scheme," Proceedings o f t h e American Philosophical Association, XLVII (1973/74): 5-20, p. 19.



And there are many other passages in the same vein.2 T h e claim, then, is that charity as a methodological precept is to be insisted on because we know ill advance, by a transcendental argument of some sort, that most of what others say and believe is going to be true (according of course to our own view of the truth). We know a priori that there is no possibility of widespread and deep-going disagreement between interpreter and interpreted. I t is this strong modal thesis that I am here concerned to undermine. I shall review arguments that have been, or might be, given to support the thesis, concluding that none is probative. T h a t the thesis is unproved is, however, less important than why it is that Davidson's central argument does not prove it; and my chief interest will be to advocate a conception of mental states, particularly propositional attitudes, which argues the incorrectness of Davidson's premises and, indirectly, of his conclusion. As by-products, I shall indicate (a) how that conception bears upon some doctrines of Hilary Putnam, and (b) what impact rejection of Davidson's position on charity has on his method of radical interpretation. We can begin by dismissing the following line of reasoning (though the confusion in it is manifest, the point that emerges will be useful later). "The object of interpretation is the explanation of actions, linguistic and nonlinguistic. Rationally to explain a n action requires that it be so described as to seem reasonable. An action will seem reasonable only if it issues from desires and beliefs that are themselves reasonable. For these attitudes to seem reasonable they must accord with what you, the interpreter, take to be reasonable. Thus it is that rationalizing explanation requires charity, i.e., the ascription of attitudes the interpreter himself has." T h e argument equivocates on 'reasonable'. T o describe an action as done for a certain reason is indeed to acknowledge that were you to have that reason (desire and belief) you wozlld find yourself rationally disposed to act as its agent did; it is not to acknowledge that you actually have such a reason. You appicciate the reasonableness of an action by putting yourself into its agent's shoes, not by forcing him into yours. So, since rationaliz~tion calls for em2 E.g., Davidson, "Thought and Talk," in S. Guttenplan, ed., h l i n d and Language (Oxford: Clarendon Press, 1975), p. 21. 3 I do not say that Davidson argues like this. [But cf. "In our need to make him make sense, we will try for a theory that finds him consistent, a belicver of truths, and a lover of the good (all by our own lights, i t goes without saying)"; "Mental Events," in L. Foster and J. Swanson, eds., Experience and T h e ory (London: Duckworth, 1970), p. 97.1 I n fact, I have come across this reasoning explicitly only in conversatiotl.



pathy not charity, there is no argument from the character of it to Davidson's thesis. A second argument is discernible in Davidson's writings. I t starts from the reflection, not here in dispute, that there is no separating the enterprises of, on the one hand, semantically interpreting a man's language, and, on the other, reaching a satisfactory determination of his propositional attitudes.4 Davidson proposes to breach the impasse created by the interdependent contributions of belief and meaning to speech behavior by taking as antecedently detectable the attitude of holding a sentence true in publicly recognizable conditions. H e says: Since knowledge of beliefs comes only with the ability to interpret words, the only possibility at the start is to assume general agreement on beliefs. We get a first approximation to a finished theory by assigning to sentences of a speaker conditions of truth that actually obtain (in our opinion) just when the speaker holds these sentences true ("On the Very Idea. ..," 18/19). Here the claim seems to be that, in order to distill off meaning from belief, a procedure of interpretation, at least in its initial attributions of belief, will inevitably attribute t r u e beliefs to a speaker observed to hold a sentence true. Now, aside from the fact that this does not get us to the desired conclusion, since it applies only to the early stages of interpretation and does not, as stated, preclude preponderant error, it simply begs the question at issue. For we may equally provide a basis for deriving the meanings of sentences held true by ulzcharitably imputing false beliefs to our speaker. We simply suppose, with or without good reason, that he has made a mistake and is expressing a false belief with a correspondingly false sentence. Falsity holds belief just as constant as truth, and affords an equally systematic rule for correlating sentences of our language with sentences of theirs in such a way (it is hoped) that the former will serve to give the meanings of the latter. I am not saying that this is a good rule, in practice or theory; only that the point about the belief-meaning conspiracy does not itself enforce the charitable assumption. We need some i n d e p e n d e n t reason for preferring a priori to find the other right instead of wrong. And Davidson does offer a further reason; a reason, moreover, that would, were it cogent, serve to underpin the arguments I have 4 See, e.g., Davidson, "Belief and the Basis of Meaning," Synthese, XXVII, 3/4 (July/August 1974): 309-324.



thus f a r peremptorily rejected. H e has often asserted, by way of justification of his radical-interpretation procedure a n d as a n independently compelling thesis, that i t is only against a background of true belief that the imputation of error is so much as intelligible. B u t h e has not been altogether expllcit as to why h e thinks this. T h e following passage from a recent paper makes his position (relatively) clear, a n d represents I suspect his m a i n motive for urging the necessity of charity all along: We can.. .take it as given that most beliefs are correct. The reason for this is that a belief is identified by its location in a pattern of beliefs; it is this pattern that determines the subject-matter of the belief, what the belief is about. Before some object in, or aspect of, the world can become part of the subject-matter of a belief (true or false) there must be endless true beliefs about the subject-matter. False beliefs tend to undermine the identification of the subject-matter; to undermine, therefore, the validity of a description of the belief as being about tliat subject. And so, in turn, false beliefs undermine the claim that a connected belief is false. T o take an example, how clear are we that the ancients-some ancients-believed that the earth was flat? This earth? Well, this earth of ours is part of the solar system, a system partly identified by the fact that it is a gaggle of large, cool, solid bodies circling around a very large, hot star. If someone believes none of this about the earth, is it certain that it is the earth that he is thinking about? ("Thought and Talk," 20/1). T h e question is clearly intended rhetorically. A n d the doctrine that prompts i t is evidently this: that we cannot sensibly assign a n object i n the world to a man's belief as its subject matter unless h e is (or is taken to be) equipped with a collateral constellatioil of other true beliefs concerning that object. T h a t is: a n object x can be the subject of a belief Bo only if there is some set of beliefs S = {B,,. . .B,) such that x satisfies the "predicative components" of the majority of {B,,. . . E n ) , for some fairly large ("endless") n. (Davidson seems also to commit himself to the converse of this principle, which asserts sufficiency, b u t I shall be chiefly occupied with his thesis construed as a necessary condition.) Before scrutinizi n g the principle we d o well to formulate a n analogous condition concerning semantic reference, as follows: a n object x can be the referent of a term t i n a sentence so uttered by a speaker U only if there is some set T = {s,. . .s,) of sentences containing t such that U is disposed to affirm T a n d x satisfies the "predicative components" of the majority of {s,. . .sn), for some fairly large ("endless") n. O u r question is whether these principles concerning the objects




of propositional attitudes and the denotations of singular terms are acceptable, and, if they are not, why they are not: for upon their truth hinges Davidson's case for according to charity the status of a constitutive conditiori of intelligibility. Denying the principles would not require one also to reject the holism alluded to by Davidson. And there is indeed considerable plausibility in the idea that the content of a belief (etc.) is determined by its relations, logical and causal, with other attitudes; so that the primary bearers of object-directedness are not beliefs singly considered but batches of interrelated beliefs. It may even be admitted that there is no possessing one belief about an object without possessing further beliefs about it, and without clear limit. But it does not follow from this holistic conception of belief content that any of the beliefs (etc.) thus interrelated are true. I n the same way holism with respect to meaning, such as Quine espouses, does not require that the totality of sentences in which a given sentence is embedded and from whose over-all semantic content its meaning is derived, be mainly true. Such holism does not itself supply a motive for charity, though without it Davidson's transcendental argument would be frustrated. T o see what is wrong with Davidson's principle, consider the following example. I t is said of the ancients-some of them-that they believed the stars to be apertures in a vast dome through which light from a conflagration behind penetrated. Such beliefs are radically yet recognizably false, and it seems undeniable that it was of the stars that they believed these falsehoods. Similarly, it seems evident that it was of the earth that Davidson's ancients entertained their egregious misconceptions. T h e acceptability of these attributions is owed, clearly, to what Quine aptly calls their relational form.5 And what explains that acceptability is simply that, in assigning such attitudes to a person, there is no presumption that the concepts (predicates) the reporter calls upon to pick out the entity the reported belief is said to concern be themselvesthose concepts-credited to the believer. T h e reporter employs his concepts to identify some object he takes the believer to be cognitively related to. T h a t these concepts are not assumed to be possessed by the believer is indicated by keeping the vocabulary one uses to express the concepts outside the scope of the belief operator. 5 See liis "Quantifiers and Propositional Attitudes," this JOURNAL, LIII, 5 (Mar. 1, 1956): 177-187, reprinted i n L. Linsky, ed., Reference and Modality (Nerv York: Oxford, 1971).



We are thus able intelligibly to impute preponderantly false beliefs to the ancients in relation to the earth and the stars without the implication that they believed these things of those entities as they are conceptualized by us. I t follows that a person may be intentionally related to an object (in the Brentano sense) without being able to conceive it aright. If it is true that relational attributions, with all their transparency, allow the reporter to distinguish his own conceptual commitments from those of his subject, then possessing concepts appropriate to some object, or kind of object, can be pulled apart from the having of beliefs (etc.) about the object, or kind of object. It may still be maintained (with what plausibility I shall not inquire) that possession of a concept requires a certain minimum of true beliefs about members of its extension, so that there cannot be shared concepts without a measure of shared beliefs; but this falls short of what Davidson wants, because now we see that disagreement concerning an object is possible unmediated by common concepts with respect to that object. Nor is the point confined to a minority of beliefs, since in general if a notional attribution is true so is a corresponding relational. These remarks suggest that Davidson's purported necessary condition on the subject matter of a belief is not binding. But the important question remains: why does it fail, and what alternative can be offered? Think of perception. Perception is paradigmatically a type of relational mental state which does not require for its holding between a percipient and an object that the percipient have true beliefs of the object he is perceiving. T h e perceptual relation is prior to, and affords a foundation for, beliefs of objects; it cannot be undermined by arrant falsity in the beliefs thus based. (This helps make intelligible the predicament of the ancients.) Nor is it the case that which object a man is perceiving is fixed by finding that object which satisfies a (l~ossiblyweighted) majority of the predicates he would be disposed to apply to the object. What all this suggests is that relational attitudes of mind are autonomous with respect to truth. This is hardly surprising, given the point just made about perception, since relational beliefs about the external world are typically, perhaps necessarily, based upon perceptual contact with its denizens (often mediated, no doubt, by memory). If perception can bring a n object into our ken independently of our believing truths concerning it, then perception-based beliefs about objects seem to be possible compatibly with dominant error.


9 7

These (as I take it) platitudes are echoed in a parallel issue over naming. And the motivation for Davidson's principle has much in common with certain assumptions that (in part) prompt a description theory of names: viz., that denotation is fixed by a certain sort of semantic fit between an object and the predicates a speaker associates with a name and supposes true of its bearer. And what Kripke-type counterexamples to that theory show, though they do not themselves explain it, is precisely that reference is autonomous with respect to truth (which isn't yet to say it is conceptually prior to truth).Vt is a consequence of this autonomy that a scheme of reference for a given (natural) language cannot be adequately characterized as that total assignment of objects to singular terms which induces a certain distribution of truth values upon those sentences of the language to which its speakers are disposed to assent: e.g., the assignment that maximizes truth. If charity recommended such a n assignment as determining a scheme of reference, as it does, it would very probably deliver an incorrect scheme. T h e same lesson goes for object-involving propositional attitudes: any procedure for assigning objects to a man's total set of beliefs whose sole or governing constraint is stated in terms of truth cannot be relied upon to yield a correct specification of their subject matter. Nor, a fortiori, could such a method be expected to illuminate what is constitutive of the relation of denotation, or what it is for a mental state to involve a n object as its subject matter. Consider now the following pair of theses about object-involving mental states: (i) if a mental state is correctly described in respect of its intentional content by reference to an object as comprising its subject matter, then the identity and existence conditions of that state are dependent upon and fixed by those of that object; (ii) it is a necesJnry (but not sufficient) condition of a n object's being referred to in correctly specifying the intentional content of a relational mental state that it-that object-figure suitably in the causal genesis of that state. I shall not here defend these theses 8 See Kripke's and Necessity," in Davidson and G. Harman, eds., Semantics of Natural Languages (Boston: Reidel, 1972). 7 On the first, see John McDowell, "On the Sense and Reference of a Proper Name," Mind, LXXXVI,342 (April 1977): 159-185, esp. sec. 8. On the second, Gareth Evans, "The Causal Theory of Names," Aristotelian Society supp. vol. XLVII (1973): 197-200, and Richard Grandy, "Reference, Meaning and Belief," this JOURNAL, LXX, 14 (Aug. 16, 1973): 439-452. I do not, in this paper, help myself to a full-fledged causal tlzeory of belief, i.e., a set of necessary and suficient conditions stated in causal terms for the intentional relation to hold. T o do that one would need some restrictions on the type of causal chain connecting



fully, since others have; but I shall offer an articulation of their force and try to show their connection. Suppose you are undergoing a perceptual experience as of seeing an F. And suppose that the content of your experience-what you are perceiving-is correctly specified by mentioning an object a. Let a be uniquely F. Now suppose yourself in a world in which again you have an experience as of perceiving an F; but now your perceptual state is correctly specified by reference to an object b, where a # b though b is uniquely F in that world. Then plainly you are i n different perceptual states in these two worlds, since perceiving a is not the same state as perceiving b-no matter how it may seem to you. Again, suppose you are in a world in which a does not exist; then you cannot be in that world in the very same mental state as you instantiate in this world when you perceive a. Let us say, then, that the perceptual state of seeing a is rigid with respect to a; a perceptual state cannot be that state unless its content is specified by mentioning that object. I t does not seem that a theory of this relation in terms of an object's fitting or satisfying the qualitative features of the state can account for this rigidity. Similarly with belief. If you believe of a that it is G, where a is and you believe that it is uniquely F, then the belief of b that it is G, in a world in which b is and you believe it is uniquely F, is just not the same belief: for its content, and therefore it, essentially involves the object it actually involves, viz. a. But according to the satisfactional theory of intentionality, there should be no distinction between these belief states, since by that theory their content is given in terms of the purely genera2 predicates the person would volunteer as singling out the object of his belief. I n fact the defect is exactly parallel to the incapacity of the description theory to secure the rigidity of a name in respect of its bearer in terms of a set of descriptions possessed by the speaker and satisfied by the name's bearer. Such a theory is not equipped to explain why it is that the identity and existence conditions of the contents of mental states and of the senses of names are determined by the particular objects whose mention serves to specify those contents or senses. That the perceptual relation cannot hold between a man and an object unless that object figures suitably in some causal process rep

a mental state with an object. Since I do not require such ail analysis for my purposes here, I shall not undertake to supply one-though it seems clear that the type of causal chain one wants to isolate will be a matter of its being suitable for the acquisition of information about the object in question. Perception is the obvious model here.



sulting in a certain mcntal state suflcretl by the man is now generally accepted."t is very plausible that a comparable necessary condition obtains for belief: a man's relational belief cannot be reported as really abozit an object referred to by the reporter unless the belief was, in part, caused by the object. Indeecl, as remarked, this requirement may bc seen as a consequence of the fact that such beliefs are acquired through the faculty of perception. At any rate, there is a clear distinction between beliefs about an object acquired by causal (perceptual) interaction with it, and beliefs acquired (as we might say) by inference, as when one conjectures there to be an object uniquely answering to certain purely general conditions."n the latter type of case, tlle content of the belief seems not to be essentially tied to the iclentity of the object actually meeting those general conditions, so here the satisfactional theory is plausible enough. But in the former case, where the content of the mental state is individuated by its causal source in the state's intentional object, it is the iclentity of that object which fixes the identity conditions of the state. It is as if tlle "onus of match" between state and object is reversed in the two types of case. T h e two theses are connected, for the causal condition helps explain the point about individuation. Silppose a belief B has intentional object a. Then, by rigidity of B with respect to a, it is necessary that B is about a. We also have it that, if B is genuinely about a, then a must be suitably involved in the causation of B; there must be some appropriate causal chain leading from a to B such as to make B about a. So B is necessarily caused (inter alia) by a. Now we can say that the fact that B is necessarily about a is e x plained by the fact that B is necessarily caused by a. On the satisfactional theory, that rigid relation between B ancl a cannot be explained, since the predicative components of the beliefs associated, li la Daviclson, with B could have been satisfied by some object distinct from a, in which circumstance B would have been about that other object-and that contradicts rigidity. Further, on that theory B could exist in a world in which a didn't exist, just as a definite description can have a sense in a world in which it contingently fails to be satisfied. But again, as there cannot be such a thing as the state of perceiving a in a world in which a doesn't exist, so there cannot be such a thing as a relational belief, whose content is given 8 T h e locus classicus is H . P. Grice, "The Causal Theory of Perception" Aristotelian Society supp. vol. vxxv (1961): 121-168. 9 Recall David Kaplan's case of "the shortest spy" and associated discussion in "Quantifying In," reprinted in Linsky, op. cit.



by mentioning a, in a world in which a isn't. (If Quine had never existed, I would not have been able to have a belief about him.) These truths about relational mental states seem, however, to be explained by the thesis, itself independently plausible, that such states necessarily include their intentional object in their causal history.1° T h e upshot of these remarks, incomplete as they are, is that the conception of the intentional relation implicit in the passage from Davidson is mistaken at root. I t is because tve observe that people causally interact with objects in their environment in such ways as enable them to have thoughts concerning those objects, paradigmatically in perception, that we are prepared to assign those objects to their beliefs as comprising their subject matter, notwithstanding the amount of bad theiry they may bring to bear upon the objects. So Davidson's reasoning was valid-charity would be a condition of the possibility of intelligible ascriptions of propositional attitudes if the holistic-satisfactional theory of intentionality were true-but, the premise being false, the conclusion need not be accepted. T h e principle of charity still lacks a transcendental justification. Digressing now from charity, I wish to show the bearing of our reflections on the inclividuation of mental states upon some recent claims of Putnam.ll He contends that the mental state that constitutes one's understanding a term-mass term, natural-kind sortal, loIt is instructive to recall Russell's distinction between two sorts of "knowledge of things": there is knowledge of things by description, which is mediated by knowledge of truths, and knowledge of things by acquaintance, which requires no such mediation. Russell says, "Knowledge of things, when it is of the kind we call knowledge by acquaintance, is essentially simpler than any knowledge of truths, and logically independent of knowledge of truths.. .Knowledge of things by description, on the contrary, always involves.. .some knowledge of truths as its source and ground." Chapter 5 of Problems of Philosophy (1912). This is not to say that Russell characterized the acquaintance relation satisfactorily. Because he required that it could hold between a person and an object only if the object's existence was indubitable for the person, and because he did not really inquire what constitutes the relation, he made its range at once too wide and too narrow: too wide because he included abstract universals, and too narrow because the indubitability condition is seldom met by the ordinary particulars with which we suppose ourselves acquainted, e.g., in perception. TVe do better to construe acquaintance as a certain type of causal relation, and knowledge of things by description as a "semantic" relation. On the constitutive character of the intentional relation, compare a writer from a somewhat different tradition: "The for-itself [consciousness] is outside itself in the in-itself [external reality] since it causes itself to be defined by what it is not; the first bond between the in-itself and the for-itself is therefore a bond of being." J.-P. Sartre, Being and Nothingness, Hazel E. Barnes, tr. (London: Methuen, 1957), p. 177. 11 See "Meaning and Reference," this JOURNAL, LXX,19 (Nov. 8, 1973): 699-711.



or singular term-clocs not uniquely fix its extension. Since, as he thinks, it is implausible that meaning could fail to determine extension, he concludes that mental state-individual or social-does not determine meaning: "the meaning of one's words is not in one's head." T h e reason offered in support of the initial contention was that two sl~eakers,or two linguistic communities, could use (phonetically identical) terms whose extensions in their respective languages were disjoint and yet be in the same mental states with respect to the terms and their extensions. T h u s we are to imagine one language in which 'water' has H,O as extension and another in which the extension is a distinct chemical substance XYZ; however, speakers of the respective languages are scientifically unsophisticated and cannot tell these phenomenologically indistinguishable substances apart. I agree with Putriam that the extensions of 'water' are distinct in the two languages, and I agree too that users of the term might be thus unable to distinguish the two substances and suppose them iclentical. But I do not think it follows that their mental states with respect to these disjoint extensions are the same. For, on the conception of the identity conditions of mental states advocated above, a correct specification of the mental states of the two groups of speakers in respect of H,O and XYZ would mention those very substances; and their distinctness guarantees the distinctness of the mental states directed toward them. This is very clear for perception: since they perceive different substances, their perceptual states are different, even though they may not be aware of this. Analogously for belief: in specifying their relational beliefs we must, if we are to report adequately, refer these beliefs to the substances causal interaction with which makes them of the substances they are. TVe should not let this elementary observation be obscured by the fact that they suppose their mental states to be identical; that would be like agreeing with a man who thinks of identical twins that they are one and the same when he says that h e saw the same person twice.12 T o insist upon identifying such qualitatively indistinguishable states would be simply not to take intentional directedness seriously and literally. Nor would it help the doctrine to describe the states purely phenomenologically, using no terms that actually refer to the objects the beliefs are about, and then claim that those descriptions do not, in their neutrality, suffice to 1 2 T h i s point shows what is rvrong with certain doctrines o f "privileged access." Since we are o f t e n less well placed to individuate our own mental states than another is, it can hardly be maintained that rve enjoy authoritative access to our own mental descriptions.



fix the relevant term's extension. For that would be open to the charge of just refusing to individuate the states correctly at all. It is no good to claim that one kind of fact does not determine another kind on the ground that one does not have to describe the former facts as they are. Again, that would be tantamount to rejecting the very idea of a genuinely constitutive intentional relation between mental states and external objects. So the reason mental state uniquely fixes extension is precisely that it is fixed by extension.13 I n fact this point tallies with others of Putnam's doctrines. For he holds that the terms in question are covertly indexical, and indexical terms are the paradigms of those expressions grasp of whose sense requires a de re attitude on the part of the grasper: one does not know the contribution made to truth conditions of a token of 'that man' unless one knows that it is of that m a n that a remark is being made. It is doubly ironic that Putnam should also champion a version of the causal theory of reference, since use of a term, and grasp of its meaning, will be causally connected, in a way that makes for a reference relation, with the object or substance referred to. Putnam has only to apply his views on the contribution of the world to meaning, to the nature and individuation of mental states to reject his claims about the relation, or lack of it, between mind and meaning. That rejection would bc well motivated, for resisting Putnam's arguments permits the retention of two otherwise attractive theses: (i) that the theory of meaning is a theory of knowledge of meaning, that meaning is an essentially cognitive notion, that there can be no more to the semantics of a language than is comprised in competence with it 14; and (ii) that semantic facts are s u p e ~ v c ~ z i e n t 1 3 Actually Putnam's thesis seems to have undergone a significant reformulation, to judge from his "The Meaning of 'Meaning'," in Afind, Language and Reality, Philosophical Papers, vol. 11 (New York: Cambridge, 1975). For he norv restricts the claim that psychological state does not determine extension, to psychological states in what he calls the "narro~vsense" (see p. 221). T h e intention seems to be to limit psychological descriptions to the purely phenomenological. T h e restriction is surely question-begging, but let us grant it to him. T h e n it must be said that he seriously misrepresents his opposition to traditionalists. For now their mistake was, not to hold that meanings are in the head (mind), but that mental states are-i.e., that it is possible adequately to identify a mental state without mentioning extra-mental entities. They should rather have conceived mental states in the "wide sense" conceded by Putnam to be legitimate (220). 14 For an extended defense of that viewpoint see Michael Dummett's Frege: Philosophy of Language (London: Duckrvorth, 1973), passim, but especially p.



upon psychological, or sociological, facts: there cannot be two speakers (communities) alike in all psychological (sociological) respects yet differing in some semantic respect.15 Both theses are preserved on the advocated view of mental states, since (i) mention of an object is required both to give the meaning of a referential term and to specify knowledge of that meaning, and (ii) the reference of a term could hardly fail to supervene upon a mental state in respect of that term which was essentially identified by adverting to the object referred to. I n fact Putnam's speakers afford a good illustration of the general position here advanced. An interpreter, apprised of chemical theory, would assign to the sentences and beliefs of Putnam's two groups of speakers distinct substances, according to their observed causal interaction, via perception, with those substances. And he would not be tempted to identify the extensions of 'water' in the two languages just because they were indiscernible with respect to the predicates the speakers were disposed to apply to their actually distinct substances. Nor would he assign XYZ to 'water' in the community surrounded by H,O on the ground that, by some coincidence, the inchoate chemical theory developed by the latter for that liquid all around them happened to fit XYZ better. I n settling the subject matter of their sayings and beliefs, observed epistemic contact is the overriding criterion. I conclude with some brief remarks on the implications for a method of radical interpretation of the position we have reached on charity. (Strictly, I have no right to these remarks, since I have not shown that no justification can be given for the status Davidson wishes to bestow upon the principle of charity; only that, if I am right, he has not established its constitutive status. But it is of interest to see what might be the consequences for a theory of radical interpretation of allowing it to be conceivable that a community of speakers be wrong about most things.) T h e central idea of Davidson's proposal is that we extract a rich concept of translation from information about truth by pairing sentences of the native language with sentences of ours on the basis of shared truth values in such a way (to cut a long story short) that something close to interpretation is yielded.16 T h e step in the procedure I wish to question 92. (The view of semantic knowledge I presuppose would not, however, be congenial to Dummett.) 15 Cf. Davidson on the supervenience of the mental on the physical, "Mental Events," op. cit., p. 88. 1 6 See his "Radical Interpretation," Dialectics, x x v ~ t(1973): 314-328.



is the anterior move from holding true to truth. It is because Davidson thinks, as a result of his views on charity, that holding true could never fail to be good evidence for truth, at least on the whole, that he takes it as providing a generally reliable test of the truth of theorems of some candidate truth theory. But if it is possible that speakers be mainly in the wrong, then reliance on their holding a sentence true as evidence for its truth will simply lead to mostly false (i.e. materially false) T-sentences, since it will direct us to pair a true sentence of ours with what is, despite their assent to it, a false sentence of theirs. So Davidson's method would be undercut at base if speakers could have preponderantly false beliefs in a conceivable case. And this suggests that holding true cannot play the role in a radical interpretation procedure that Davidson has tried to cut out for it. What might be put in its place? We noted earlier that rationalization involves the attribution of attitudes that make certain kinds of behavior reasonable for the agent, and therefore apt to be followed. T h e attribution can be tested, at least u p to indeterminacy, by seeing whether the agent behaves as one would who possessed those attitudes; and of course the verification of such a conjectured set of attitudes will inevitably be holistic. This exercise is reminiscent of the so-called hypothetico-deductive method in science: crudely, think of a theory that would explain the observed facts if it were true, then test it by deducing its further consequences and seeing whether they hold.l7 I suggest that interpretation may proceed in just that way: conjecture a theory of meaning and propositional attitudes guided by whatever hunches and expectations one has, then test for whether the subject behaves as one would of whom that theory were true.18 I t will be complained that this sketch, even if filled out, could scarcely amount to a method, i.e., that issue with the minimum of imaginative a set of prescri~~tions ingenuity in the selection of a correct theory (modulo, as always, indeterminacy). T h e complaint is correct; the hypothetico-deductive "method" is not a method in that sense of the word. But the lack of such a procedure for extracting theory from evidence does not impede science; and indeed, it is to be assumed that radical 17See, for the elements, Carl G. Ilempel, Philosophy of Natural Science (Prentice-Hall, 1966), chapters 2 and 3. In drawing this parallel I intend no commitment to a deductive-nomologica1 construal of human action explanation; that is a separate question. 1s This general rubric seems close to the conceptions of radical interpretation favored by McDowell, op. cit., and David Lewis, "Radical Interpretation," Syntlzese, x x v i ~ ,3/4 (Jnly/August 1974): 331-344.



interpretation has often been successfully undertaken without benefit of Davidson's method. I t may well be, as Davidson's writings suggest, that there can be no method without charity: but there can be interpretation without either.l9 COLIN MCGINN

University College London


' A R T R Y FIELD'S recent paper on the indeterminacy of reference is so brilliant and so important that it merits criticism. I will argue that Field's main example fails because, contra Field, there is strong evidence in favor of one of the analytic hypotheses he rejects; namely, the hypothesis that the Newtonian term 'mass' has the same denotation as the special relativistic term 'proper mass'. +

T h e argument. Both Newtonian mechanics (NhJ) and special relativity theory (SRT) can be written in a four-dimensional, intrinsic (i.e., coordinate-free) form. This for; is (i) the most perspicuous form known for either theory in terms of making the underlying mathematical and physical assumptions clear and explicit, and (ii) the most useful form known for comparing the two the0ries.l I n this form three principles of NM appear as (N1)

(N2) (N3)

m N is a scalar invariant P, = nz,VN FN = mSAN

l Q I should emphasize that in this paper I have not: (i) claimed to refute every conceivable argument that might be offered in support of Davidson's thesis, though I know of none that seems to me convincing; (ii) considered all versions of the plinciple of charit) that have been proposed, though the others I know of (e.g., David Lewis's in "Radical Interpretation," op. cit., p. 336) do not seem incompatible with the attribution of preponderant error; (iii) discussed charity as to finding others consistent and rational, which seems to me highly compelling, though quite compatible with the negation of the Davidsonian claim I have discussed. * "Theory Change and the Indeterminacy of Reference," this JOURKAL, LXX, 14 (Aug. 16, 1973): 462-481. 1 For support of (i) and (ii) see J. Earman and M. Friedman, "The Meaning and Status of Newton's Law of Inertia," Ph~losofihyof h e n c e , X L (1973): 329359.