Back to online papers


Alex Byrne,

Department of Linguistics and Philosophy, MIT

[Published in A Companion to the Philosophy of Mind, ed. S.D. Guttenplan (Blackwell, 1994)]

Introductory texts in the philosophy of mind often begin with a discussion of behaviourism, presented as one of the few theories of mind that have been conclusively refuted. But matters are not that simple: behaviourism, in one form or another, is still alive and kicking.

'Behaviourism' covers a multitude of positions. Yet there is a common underlying thread. The behaviourist takes minds not to be inner psychic mechanisms merely contingently connected with their outer behavioural effects, but to be (at least to a significant extent) constituted by those outer effects. The behaviourist's motivation is often epistemological: on the picture of the mind as essentially inner, how can its outer effects provide us with the wide ranging knowledge of others' minds we confidently take ourselves to possess?

As an imperfect but serviceable analogy, consider a clock. A clock has visible moving exterior parts - the hands. To the behaviourist about clocks, a clock is simply something with such time-indicating exterior parts. The inner workings of any clock are entirely irrelevant to its status as a clock, provided they produce (or at least don't interfere with) the movement of the hands. The anti-behaviourist, by contrast, thinks of a clock as an inner mechanism which, in favourable circumstances, can cause some exterior parts to move in a way which reliably indicates the time. But there is no a priori reason, according to the anti-behaviourist, why those favourable circumstances should be even possible. There may be clocks to which, given the laws of nature, hands cannot be attached. Moreover, such clocks need not be abnormal ones: they could even be paradigm examples of clocks.

Of course there are intermediate positions. One might hold that typical clocks must have, or be capable of having, time-indicating hands, while acknowledging that there could be atypical clocks of which this is not so. These atypical clocks would count as clocks in virtue of sharing inner mechanisms with typical clocks. Or one might, while insisting on the importance of the hands, impose some minimal constraints on the innards of a clock. For example, one might say that something could not be a clock unless the big hand and the little hand were controlled by the same mechanism; but beyond that, anything goes.

Suppose, to press the analogy still further, that we never open up any clocks to examine their inner parts. Clock anti-behaviourism would then seem to give us an epistemological problem: how do we know that there are any clocks?

Behaviourism flourished in the first half of the twentieth century. Philosophers from that period with behaviourist leanings include Carnap, Hempel, Russell, WITTGENSTEIN, and RYLE. Arranging some contemporary philosophers on a spectrum from the most behaviouristically inclined to the least finds QUINE at the behaviourist end, and SEARLE at the other. DAVIDSON, DENNETT and DUMMETT are closer to Quine than Searle, with FODOR, DRETSKE (and many others) closer to Searle than Quine. Armstrong and LEWIS are squarely in the middle.


Let us say that any instance of physical behaviour is a physical change to an agent's body (perhaps in relation to his environment), such as the rising of the agent's arm.

Let us say that any instance of agential behaviour is something an agent does, such as raising his arm. (For a related distinction, see Armstrong, 1968, p. 84.) We can similarly define physical behavioural dispositions and agential behavioural dispositions.

The relationship between physical behaviour and agential behaviour is controversial. On some views, all ACTIONS are identical to physical changes in the agent's body. (However, some kinds of physical behaviour, such as reflexes, are uncontroversially not kinds of agential behaviour.) On others, an agent's action must involve some physical change, but is not identical to it.

Both physical behaviour and agential behaviour could be understood in the widest sense. Anything a person can do - even calculating in his head, for instance - could be regarded as agential behaviour. Likewise, any physical change in a person's body - even the firing of a certain neuron, for instance - could be regarded as physical behaviour.

Of course, to claim that the mind is 'nothing over and above' such-and-such kinds of behaviour, construed as either physical or agential behaviour in the widest sense, is not necessarily to be a behaviourist. The theory that the mind is a series of volitional acts - a view close to Berkeley's - and the theory that the mind is a certain configuration of neural events, while both controversial, are not forms of behaviourism.

So either the behaviourist needs a less inclusive notion of behaviour or, at the very least, if he does allow some loner processes to count as behaviour, he must minimize their importance.

Waving to someone, or the consequent movement of one's arm, are more-behaviouristically acceptable than calculating in the head, or the accompanying firing of neurons. If a philosophical theory of the mind emphasizes wavings or arm movements over silent cogitations and brain events, then it is, to that extent, behaviourist. Accordingly, 'physical behaviour' and 'agential behaviour' will henceforth be understood in a behaviouristically restricted way. Wavings or visible arm movements (together with, for example, blushing and standing still, which do not involve bodily movement) are included, inner goings-on are excluded. This is vague, but none the worse for that.


Eliminative behaviourism is a forerunner of the contemporary doctrine of eliminative materialism (see ELIMINATIVISM). Eliminativists about the mental repudiate all or most of our commonsense psychological ontology: beliefs, conscious states, sensations, and so on. One argument for eliminativism, in contemporary dress, is this. First, 'FOLK PSYCHOLOGY', taken to be, inter alia, our tacit theory of the behaviour of others, suffers various deficiencies: widespread explanatory failings, for instance. Second, there are much better theories of behaviour which do not quantify over mental states. Therefore, in accordance with good scientific practice, folk psychology should be replaced by one of these superior theories. Eliminative behaviourism takes the replacement theory to be couched in the vocabulary of physical behaviour. (In fact, there are a number of choices for the vocabulary of physical behaviour: the austere terminology of kinematics is one, a certain rich fragment of English another. But at the least the vocabulary will not contain mentally loaded terms.)

Eliminative behaviourism is a dominant theme in the writings of Watson (1930) and Skinner (see the papers collected in Skinner et al., 1984): two central figures in the development of the now unpopular doctrine of psychological behaviourism. Psychological behaviourism is primarily a claim about the correct methodology of a scientific psychology, and arose in the early part of the twentieth century as a reaction to the 'introspective' psychology of Wundt, James and Titchener. According to the introspective school, the subject matter of psychology is CONSCIOUSNESS, and the proper methodology for its study is INTROSPECTION. Against this, Watson argued that a scientific psychology should just concern itself with what is 'objective', and 'observable', namely, according to him, behaviour.

There was more to psychological behaviourism than this, of course. Watson and Skinner both thought that the behaviour of an organism could be explained by its history of stimulation together with relatively simple processes of behavioural modification. Skinner's introduction of operant conditioning as one of these processes marked an improvement over the crude stimulus-response behaviourism of Watson.

Watson found it more acceptable than Skinner to go inside the organism to find stimuli and responses, but Watson gave these inner events no special status. Further, Watson took the brain to be just one inner part among others of equal importance: 'the behaviourist [places] no more emphasis on the brain and the spinal cord than upon the striped muscles of the body, the plain muscles of the stomach, [and] the glands' (1930, p. 49).

Both Watson and Skinner shared unclarities on two connected issues. They were unclear whether stimuli and responses can be described in mentally loaded terms, or whether only purely physical descriptions are allowed. They also vacillated between endorsing: (a) eliminativism about the mental; (b) the claim that mental states exist but are irrelevant to the scientific study of human beings; or (c) the claim that mental terminology can be translated into vocabulary of physical behaviour. But they both had a strong tendency towards eliminativism. Watson, for instance, took 'belief in the existence of consciousness' to go 'back to the ancient days of superstition and magic' (1930, p. 2). And Skinner has expressed similar sentiments (e.g. 1971, ch. 1). (For recent commentary on Skinner, see Skinner et al., 1984; Modgil and Modgil 1987.)

QUINE is another eliminative behaviourist, but for quite different reasons. His behaviourism appears to be motivated largely by his verificationism. He gives two reasons for eliminativism. The first is that belief and desire talk resists regimentation in first order logic, which Quine takes to be the litmus test for complete intelligibility. The second is his argument for the thesis of the indeterminacy of translation, which purports to show that there is simply no 'fact of the matter' as to what someone's language means (Quine, 1960, ch. 2). Quine assumes a sufficiently intimate connection between language and belief for it to follow that there is also no 'fact of the matter' as to what someone believes.

Most behaviouristically inclined philosophers are not eliminativists. The most powerful and straightforward kind of (non-eliminative) behaviourism is:

analytic (or logical) behaviourism: statements containing mental vocabulary can be analysed into statements containing just the vocabulary of physical behaviour.

Skinner can be interpreted as a part-time analytic behaviourist (e.g. Skinner 1971, p. 24). And Hempel, stating a view common to many logical positivists, wrote:

All psychological statements which are meaningful, that is to say, which are in principle verifiable, are translatable into statements which do not involve psychological concepts, but only the concepts of physics. (1949, p. 18, italics omitted)

Hempel derived this strong thesis from two premises. First, he held (but later abandoned) the verificationist theory of meaning, namely that 'the meaning of a statement is established by its conditions of verification' (1949, p. 17, italics omitted). Second, he held that a person's physical behaviour was a large part of the evidence for ascribing to him particular mental states. Putting the two together, he concluded that statements about mental states were equivalent to statements (largely) about physical behaviour.

Largely, but not entirely. Hempel was not a thoroughgoing behaviourist in the sense of ignoring inner processes altogether. According to Hempel, the verification conditions (which amount to the meaning) of 'Paul has a toothache' include certain changes in Paul's blood pressure, his digestive processes and his central nervous system. But as gross bodily movements play a large role in the verification of psychological sentences, and hence play a large role in determining their meanings, Hempel's position is to a significant extent behaviourist.

We now briefly turn to Ryle's influential quasi-behaviourist polemic, The Concept of Mind (1949) (see RYLE). Somewhat confusingly, Ryle's position is often called analytic (or logical) behaviourism, but it is quite different, in content and motivation, from the positivist sort of behaviourism exemplified by Hempel.

Hempel's behaviourism is part of his PHYSICALISM. Ryle, in contrast, was no physicalist. He regarded the very question of whether the world is ultimately physical as conceptually confused. Accordingly, Ryle spoke of agential behavioural dispositions, and showed little inclination to analyse this away in terms of physical behavioural dispositions.

Further, it is arguable whether Ryle's primary intention was to offer analyses of statements apparently about inner mental occurrences in terms of agential behavioural dispositions. At any rate, he does not supply very many. Ryle was chiefly concerned to deflate the idea that there must be complex inner mental processes behind a person's public actions and to show how this dissolved the problem of other minds. In doing this, he often did not analyse the inner away, but merely derided it. 'Overt intelligent performances are not clues to the workings of minds, they are those workings. Boswell described Johnson's mind when he described how he wrote, talked, ate, fidgeted and fumed. His description was of course incomplete, since there were notoriously some thoughts which Johnson kept carefully to himself and there must have been many dreams, daydreams and silent babblings which only Johnson could have recorded and only a James Joyce would wish him to have recorded' (Ryle, 1949, pp. 58-9).

Ryle was indeed, as he reportedly said, 'only one arm and one leg a behaviourist'.

Physicalists such as Place, Smart, Armstrong, and Lewis, were able to find much to agree with in Ryle, despite his antipathy toward physicalism. They took Ryle to have in effect shown that physical behavioural dispositions were of great importance in understanding the nature of the mind.

Geach (1957, p. 8) raised an important difficulty for any behaviourist analysis of the propositional attitudes, whether in terms of agential or physical behavioural dispositions. Geach noted that what someone does, or is disposed to do, depends not only on the fact that he holds a particular belief, but also on his desires (and, it should be added, on his other beliefs as well). Therefore there can be no question of a simple atomistic behavioural analysis: one which matches each belief with a different kind of behaviour (whether specified in the language of agential or physical behaviour). A given belief may issue in practically any sort of behaviour, depending on the agent's other attitudes (see BELIEF; PROPOSITIONAL ATTITUDES.)


Geach's observation that belief types do not have characteristic behavioural expressions is now regarded as a datum of folk psychological explanation, and consequently few contemporary philosophers are tempted by any sort of (atomistic) analytic behaviourism. But the failure of this kind of behaviourist analysis does not imply the failure of behaviourism, construed as a metaphysical thesis about the nature of the mind. All it does imply is that the most direct route to behaviourism - namely a simple analysis pairing beliefs and behavioural dispositions - is a dead end. A convincing argument for behaviourism must proceed down a less obvious path.

Contemporary behaviourist views derive from three sources. The first is the analytic FUNCTIONALISM of Armstrong, Lewis and others (see LEWIS). According to this view, the meanings of mental terms are determined by their role in our commonsense theory of behaviour: folk psychology. This tacitly known theory is taken to consist of generalizations linking perceptual input, (physical) behavioural output, and mental states. Some of these generalizations will link beliefs to desires, thus incorporating Geach's insight that a belief does not have a fund of behaviour to call its own. However, if analytic functionalism is correct, then the analyses of mental terms will have a significant behavioural component.

Note that analytic functionalism, while not underwriting an entailment from having such-and-such physical behavioural dispositions to having a mental life, nonetheless does appear to solve the problem of other minds. According to analytic functionalism, one may reasonably conclude that a creature has a mind on the basis of its behaviour, just as one may reasonably conclude that a car has an engine on the basis of its motion.

The second source is opposition to the functionalist idea that an organism needs a certain kind of inner causal organization in order to be a genuine believer.

These two incompatible lines of thought have their roots in Ryle, although he would only have expressed agreement with the second. The third source of contemporary behaviourism is one on which Ryle's influence was at best indirect, although he would have found it perfectly congenial. That source is Wittgenstein's (1958) attack on the possibility of a 'private language', which some take to show that meaning and belief must be 'manifestable' in behaviour (see WITTGENSTEIN).

Roughly corresponding to these three sources are three behaviourist theses. The first is:

Behaviour-as-necessary: necessarily, anything that has no physical behavioural dispositions of a certain kind and complexity does not have a mental life.

This vague formulation should be taken to express the behaviourist thought that some thing like a stone could not possibly have a mental life, because it is outwardly inert (not because there is nothing sufficiently complicated going on in the stone). It is a thesis which, incidentally, can with same safety be ascribed to Wittgenstein (1958, I, 281-4).

Analytic functionalism can evade the behaviour-as-necessary view to some extent, but not completely. Lewis's brand of analytic functionalism, for instance, has the consequence that anything which has no behavioural dispositions above a certain level of complexity either (a) has no mental life; or, (b) is an atypical member of its kind (Lewis, 1980). Therefore an extreme kind of permanent paralysis cannot be the typical condition of any population of thinking creatures: there cannot be a race of thinking stones.

A natural companion to the behaviour as-necessary view, although logically independent of it, is:

Behaviour-as-sufficient: necessarily, anything that has physical behavioural dispositions of a certain kind and complexity has a mental life.

According to functionalism, whether of the analytic variety or not, minds must have some specific type of inner causal structure. Functionalism in general is therefore inconsistent with the behaviour-as-sufficient view. However, DENNETT - on whom Ryle had a direct influence - has been a longtime opponent of the idea that to have a mind is to have a specific type of inner causal structure. On the one hand, Dennett is unmoved by thought experiments - often involving Martian scientists controlling hollow anthropoid puppets - which purport to show that having a mind involves some restriction on inner causal order. (See below.) On the other hand, Dennett is impressed, as was Hume, with the remarkable predictive power of everyday psychology. Adopting 'the intentional stance' (Dennett, 1971, 1981) - the predictive strategy of ascribing intentional states such as belief and desire to a system - is indispensable in many cases, its utility not diminished by any discoveries of bizarre inner causal organization.

Now if we can be assured that the assumption that a system has beliefs is not going to be defeated by future evidence - although specific belief attributions may well be - and that assumption is useful, why insist that there is some further question to be raised concerning whether the system really has beliefs?

Thus, according to Dennett, 'any object whatever its innards - that is reliably and voluminously predictable from the intentional stance is in the fullest sense of the word a believer' (1988, p. 496).

The two behaviourist theses considered so far are certainly controversial, but both are weak forms of behaviourism. Even their conjunction does not imply that any two behavioural duplicates necessarily share the same mental states. That last claim is our third behaviourist thesis:

Supervenient behaviourism: psychological facts supervene on physical behavioural dispositions: necessarily, if x and y differ with respect to types of mental states, then they differ with respect to types of behavioural dispositions (see SUPERVENIENCE).

(Note that analytic behaviourism entails supervenient behaviourism, but not conversely.)

Supervenient behaviourism may be broadened to include the supervenience of linguistic meaning on behavioural dispositions, or narrowed to exclude, say, sensations. Supervenient behaviourism can accommodate the view that content is not entirely 'in the head' by taking the supervenience base to comprise physical behavioural dispositions together with facts about the subject's environment. (The possibility of this latter refinement will be taken for granted in what follows.)

Supervenient behaviourism is quite compatible with eliminativism (according to which no x and y ever differ mentally!). And indeed Quine appears to need the former to derive the latter via the indeterminacy thesis (see e.g. van Cleve, 1992). However, some philosophers who repudiate eliminativism are at least attracted by supervenient behaviourism. More precisely, these philosophers held a doctrine about the mind which, together with assumptions they would probably accept, entails supervenient behaviourism. This doctrine may be explained as follows.

In the writings of DAVIDSON, and to some extent Dennett and Dummett, there is a strand of thought which amounts to a kind of third-person Cartesianism (see HISTORY). Where Cartesianism holds that someone has, under ideal conditions, complete and infallible access to his own mental life (see FIRST-PERSON AUTHORITY), the third-person version says that someone has, under ideal conditions, complete and infallible access to the mental life of another. Or, in its typical formulation, complete and infallible access to the PROPOSITIONAL ATTITUDES (and/or language) of another. Picturesquely: under ideal conditions a subject's belief-box and desire-box become transparent. This kind of claim is sometimes called interpretivism (Johnston, 1991).

We can dramatize interpretivism by introducing the familiar device of the ideal interpreter: an idealization of a human being, in ideal epistemic circumstances. The ideal interpreter is capable, according to interpretivism, of discovering exactly what a subject believes, desires (and/or means). The powers and data of the interpreter are a matter for dispute, but at least three points should be noted.

First, the interpreter's powers are not to be construed - on pain of triviality - to be just whatever powers are necessary in order to deliver the facts.

Second, the interpreter's official evidence does not include the contents of the subject's mental states, or the meaning of his utterances, for that is what the interpreter is supposed to be finding out.

Third, interpretivism does not offer a reductive analysis - at least of the attitudes -for the reference to our best judgments (captured in the heuristic device of the ideal interpreter) is supposed to be an ineliminable part of the story. Lewis (1974) makes use, in effect, of the ideal interpreter, but intends this to be a staging-post en route to a fully reductive account. That is not a version of interpretivism.

Bearing in mind the three points above, we could formulate interpretivism in the particular case of belief as the thesis that all biconditionals of the following form are a priori:

x believes that p if and only if, if there were an appropriately informed ideal interpreter, he would be disposed to attribute to x the belief that p.

Something like this seems to be clearly expressed by Davidson when he writes:

What a fully informed Interpreter could know about what a speaker means is all there is to learn; the same goes for what the speaker believes. (1986, p. 315)

Davidson is making a substantial claim, and so 'fully informed' should not be understood as 'fully informed about what a speaker means and believes'. He here has in mind his 'radical interpreter' (Davidson, 1973), starting the process of interpretation with no knowledge of the meaning of the subject's language or the content of his beliefs (although Davidson allows the interpreter's initial data to include intentional facts, namely what (uninterpreted) sentences the subject 'holds true').

Dennett comes to a similar conclusion:

[A]ll there is to being a true believer is being a system whose behaviour is reliably predictable via the intentional strategy, and hence all there is to really and truly believing that p (for any proposition p) is being an intentional system for which p occurs as a belief in the best (most predictive) interpretation. (1981, p. 29)

Now an interpretation requires a (hypothetical) interpreter, and so Dennett would appear to be saying that some ideal interpreter could deliver all and only the facts about what a subject believes. Moreover, as the failure of reductions in the philosophy of mind has been one of Dennett's major themes, he very likely thinks that this talk of interpretations and interpreters cannot be analysed away. But then Dennett is an interpretivist.

Interpretivism does not entail supervenient behaviourism. For it may be that the interpreter must take into account certain non-behavioural facts (e.g. facts about inner causal organization) when determining what someone believes. It seems likely, however, that Dennett's and Davidson's ideal interpreters would not ascribe different beliefs to two subjects unless they differed in their physical behavioural dispositions: their respective interpreters would judge any two behavioural duplicates alike. And as the ideal interpreter's word is gospel, this entails that no two subjects would differ in beliefs (or, we may assume, desires) unless they differed in physical behavioural dispositions. But this is supervenient behaviourism with respect to the propositional attitudes (and linguistic meaning, at least in the case of Davidson).

Dennett appears to arrive at interpretivism by an ambitious extension of his route to the behaviour-as-sufficient view. That is: why not take the ideal interpreter at his word, if we have no clear conception of how he can be mistaken? Davidson's reasons for interpretivism are less straightforward. The main ones emerge in the following passage, where Davidson argues for a behaviourist version of interpretivism from a Wittgensteinian premise about the 'publicity' of meaning:

As Ludwig Wittgenstein, not to mention Dewey, G.H. Mead, Quine, and many others have insisted, language is intrinsically social. This does not entail that truth and meaning can be defined in terms of observable behaviour, or that it is 'nothing but' observable behaviour; but it does imply that meaning is entirely determined by observable behaviour, even readily observable behaviour. That meanings are decipherable is not a matter of luck; public availability is a constitutive aspect of language (Davidson, 1990, p. 314).

Dummett has expressed similar views (e.g. 1973, pp. 217-18).

Davidson's 'readily observable behaviour' is intended to be understood as 'behaviour observable as physical behaviour' (or, perhaps as: 'behaviour observable as the holding true of sentences'). For suppose that the relevant deliverances of observation include linguistic behaviour such as saying that snow is white. Then Davidson's claim that meaning is determined by this sort of behaviour is, while perhaps not entirely trivial, at least extremely weak. Alternatively, suppose that the deliverances of observation include psychological behaviour such as intending to induce a certain belief in one's audience, but exclude any kind of behaviour described in a way which presupposes linguistic meaning. Then Davidson's claim would certainly not be trivial, but would violate his well-known insistence that a speaker's propositional attitudes and the meaning of his language are settled together, neither one having priority over the other (e.g. Davidson, 1973).

It seems, then, that lying behind Davidson's argument is a controversial epistemic thesis: our warrant for saying that someone speaks a certain language, or has certain beliefs, is ultimately founded on behaviour, observed as physical behaviour (or, perhaps: observable as the holding true of sentences). But even if that is found acceptable, Davidson's apparent move from 'our judgements are founded on such-and-such data' to 'the truth of our judgements is determined by such-and-such data' is questionable. We have learnt much about the world that is not 'determined' by what we have learnt it from.


To refute the view that a certain level of behavioural dispositions is necessary for a mental life, we need convincing cases of thinking stones, or utterly incurable paralytics, or disembodied minds. But these alleged possibilities are to some merely that.

To refute the view that a certain level of behavioural dispositions is sufficient for a mental life, we need convincing cases of rich behaviour with no accompanying mental states. The typical example is of a puppet controlled, via radio links, by other minds outside the puppet's hollow body (Peacocke, 1983, p. 205; Lycan, 1987, p. 5; see also Block, 1981). But one might wonder - and Dennett does - whether the dramatic devices are producing the anti-behaviourist intuition all by themselves. And how could the dramatic devices make a difference to the actual facts of the case? If the puppeteers were replaced by a machine (not designed by anyone, yet storing a vast number of input-output conditionals) which was reduced in size and placed in the puppet's head, do we still have a compelling counterexample to the behaviour-as-sufficient view? At least it is not so clear.

Such an example would work equally well against (the anti-eliminativist version of) the view that mental states supervene on behavioural dispositions. But supervenient behaviourism could be refuted by something less ambitious. The 'X-worlders' of Putnam (1965), who are in intense pain but do not betray this in their verbal or non-verbal behaviour, behaving just as pain-free human beings, would be the right sort of case. But even if Putnam has produced a counterexample for pain - which Dennett for one would doubtless deny - an'X-worlder' story to refute supervenient behaviourism with respect to the attitudes or linguistic meaning will be less intuitively convincing. Behaviourist resistance is easier here for the reason that having a belief, or meaning a certain thing, lack distinctive phenomenologies.

There is a more sophisticated line of attack. As Quine has remarked, some have taken his thesis of the indeterminacy of translation as a reductio of his behaviourism (1990, p. 37). For this to be convincing, Quine's argument for the indeterminacy thesis has to be persuasive on its own terms, and that is a disputed matter.

If behaviourism is finally laid to rest to the satisfaction of most philosophers, it will probably not be by counterexample, or by areductio from Quine's indeterminacy thesis. Rather, it will be because the behaviourist's worries about other minds, and the public availability of meaning, have been shown to be groundless, or not to require behaviourism. for their solution. But we can be sure that this happy day will take some time to arrive.



Armstrong, D.M. 1968. A Materialist Theory of the Mind. London: Routledge and Kegan Paul.

Block, N., ed. 1980. Readings in Philosophy of Psychology, vol. 1. Cambridge, MA: Harvard University Press.

Block, N. 1981. Psychologism and behaviourism. Philosophical Review, 90, 5-43.

Davidson, D. 1973. Radical interpretation. Dialectica, 27, 313-28. Reprinted in Davidson, Inquiries into Truth and Interpretation. Oxford University Press, 1984.

Davidson, D. 1986. A coherence theory of truth and knowledge. In Truth and Interpretation: Perspectives on the Philosophy of Donald Davidson, ed. E. LePore. Oxford: Basil Blackwell.

Davidson, D. 1990. The structure and content of truth. Journal of Philosophy, 87, 279-328.

Dennett, D.C. 1971. Intentional systems. Journal of Philosophy, 68,87-106. Reprinted in Dennett, Brainstorms. Brighton: The Harvester Press.

Dennett, D.C. 1981. True believers. In Scientific Explanation, ed. A. F. Heath. Oxford University Press. Reprinted in Dennett, The Intentional Stance. Cambridge, MA: MIT Press, 1987; page references are to the latter.

Dennett, D.C. 1988. Précis of The Intentional Stance. Behavioural and Brain Sciences, 11, 495-504.

Dummett, M.A E. 1973. The philosophical basis of intuitionistic logic. In Logic Colloquium '73, ed. H. E. Rose and J. C. Shepherdson. Reprinted in Dummett, Truth and Other Enigmas. London Duckworth, 1978; page references are to the latter.

Geach, P.T. 1957. Mental Acts: Their Content and their Objects. London: Routledge & Kegan Paul.

Hempel, C.G. 1949. The logical analysis of psychology. In Readings in Philosophical Analysis, ed. H. Feigl and W. Sellars. New York: Appleton-Century-Crofts. Reprinted with revisions in Block 1980; page references are to the latter.

Johnston, M. 1991. The missing explanation argument and its impact on subjectivism. MS.

Lewis, D.K. 1974. Radical interpretation. Synthese, 23, 331-44. Reprinted with postscripts in Lewis, 1983.

Lewis, D.K. 1980. Mad pain and martian pain. In Block 1980. Reprinted with postscript in Lewis, 1983.

Lewis, D.K. 1983. Philosophical Papers, vol. 1. Oxford University Press.

Lycan, W.G. 1987. Consciousness. Cambridge, MA.: MIT Press.

Modgil, S., and Modgil, C., eds. 1987. B. F. Skinner: Consensus and Controversy. London: Falmer Press.

Peacocke, C. 1983. Sense and Content. Oxford University Press.

Putnam, H. 1965. Brains and behavior. In Analytical Philosophy, Vol. 2, ed. R. J. Butler. Reprinted in Block, 1980.

Quine, W.V. 1960. Word and Object. Cambridge, MA.: MIT Press.

Quine, W.V. 1990. Pursuit of Truth. Cambridge, MA.: Harvard University Press.

Ryle, G. 1949. The Concept of Mind. New York: Barnes & Noble.

Skinner, B.F. 1971. Beyond Freedom and Dignity. New York: Alfred A. Knopf.

Skinner, B.F., et al. 1984. Skinner: canonical papers. Behavioural and Brain Sciences, 7, 473-701.

Van Cleve, J. 1992. Semantic supervenience and referential indeterminacy. Journal of Philosophy, 89, 344-61.

Watson, J.B. 1930. Behaviourism. University of Chicago Press.

Wittgenstein, L. 1958. Philosophical Investigations (2nd edn.). Oxford: Basil Blackwell.

Thanks to David Lewis, Michael Thau, and the editor for helpful comments.