Back to Home The Language of Thought Entry

Lawrence Kaye - University of Massachusetts at Boston  


Note: items marked (*) are mentioned in the tour, but they are not primarily concerned with the linguistic thought issue (some do not discuss it at all.)

Also note that the following compilation does not include articles on the following topics: Fodor vs. Dennett's instrumentalism, Fodor vs. Stich's syntacticism, Fodor vs. the Churchland's eliminativism, Fodor and Pylyshyn vs. connectionism unless such articles explicitly discuss the linguistic thought issue.



*Ryle, Gilbert (1949), The Concept of Mind, London: Hutchinson

Chapters 1-2 contain the classic presentation of logical behaviorism: Ryle attacks "Descartes' Myth"--not just the myth of the non-physical ghost in the machine, but the myth that mentalistic language refers to hidden, inner states and processes. That is a "category mistake". Rather, mental terms are to be understood as referring to complex sets of dispositions to behave. Much of what we do involves knowing how to do something rather than explicit propositional knowledge--knowledge that. This book was highly influential in its day and for several decades after--logical behaviorism was the dominant view in philosophy of mind through the early 70's. But just as rapidly as it rose it vanished--present-day adherents are few and far between. Essentially, this is anti-explanationism in psychology, for once we ask how it is that we are able to perform the behaviors that constitute knowing how to do something, we find that we are back to the postulation of inner, perhaps non-conscious, representational states and processes.

*Chomsky, Noam (1965), Aspects of the Theory of Syntax, Cambridge, MA: MIT Press.

Chomsky's seminal work on generative grammar begins with a philosophical development of his unique and brilliant conceptions of knowledge of language and linguistic methodology--Chapter 1 is thus essential reading for anyone interested in the philosophy of language. Includes explications of the competence/performance distinction, the idea of tacit knowledge, surface vs. deep structure, universal grammar, the innateness hypothesis, and connections with rationalism (e.g., Descartes and Leibniz).

*Chomsky, Noam (1975) "On Cognitive Capacity," Chapter 1 of Reflections on Language. New York: Pantheon. Reprinted in Block 1981.

A slight update of Chomsky (1965), Chapter 1: The study of language is the study of knowledge structures. There is no available learning theory to explain how humans achieve these structures. The most plausible approach is to postulate an innate universal grammar that contains the "essence" of human languages. The development of language in persons is thus viewed as a specific, programmed cognitive capacity--an organ that grows in the brain. These points are set in a broad discussion of knowledge and learning theory that generalizes Chomsky's linguistic methodology to other cognitive domains.

Fodor, Jerry A. (1975) The Language of Thought. New York: Crowell.   

Development and defense of the view that there is an innate language of thought, distinct from all spoken languages, that expresses all semantically primitive concepts (viz., most concepts, in Fodor's view). This "Mentalese" is supposed to be what we think in, and what we use to learn the semantics of spoken languages. Some main points/conclusions: in order to have psychological computations, there must be representations that get computed. 2) Public languages won't work 3) by the (recently so-called) Standard Argument, the system of representations constitute a very conceptually rich innate language. 4) Imagistic representations cannot play the role that Mentalese does, but rather rely on Mentalese for their meanings. Strong, clear philosophical argumentation mixed with discussion of examples and evidence from psychology and linguistics. Though the cognitive science is somewhat dated, the arguments remain forceful.

Dennett, Daniel (1978) "A Cure for the Common Code?" in Brainstorms, Cambridge, MA: MIT Press. 

Review of Fodor's The Language of Thought 1975. Dennett presents a series of doubts about whether beliefs correspond to explicit representations. Similar concerns are voiced in "Brain Writing and Mind Reading" (same volume), where Dennett actually develops a(n admittedly crude) version of a LOT hypothesis. Useful articles for understanding the differences between Fodor's and Dennett's conceptions of the mind/brain.

Field, Hartry (1978) "Mental Representation," Erkenntnis, 13: 9-61. Reprinted in Block 1981). 

Long and at times somewhat overly technical paper that provides a broad conception for explaining linguistic meaning in terms of (postulated) linguistic mental representations. Field's main line is that the only way that he can see for physicalism to solve Brentano's problem is by treating propositional attitudes (e.g., belief) as (probably functional) relations to sentential internal representations. The semantics of the inner sentences is to be given, ultimately, via a Tarski-style theory of truth based on a causal theory of reference. In other words, sentences are the only known physical things that can account for meaning; without the postulation of mental sentences we are stuck with Brentano's problem, viz., physically irreductable intentionality. Part I develops the view. Belief is analyzed into two-components: the belief that p is understood as being a matter of standing in a psychological relation to a sentence that means p. (N.B. Fodor agrees completely with this part of the approach.) Problems for developing each component are then discussed, including the issues of animals and of core beliefs. Part II dismisses a potential alternative to Field's view, namely a Lewis/Stalnaker-style functionalist account that provides a possible worlds semantics for the functionalized psychological theory, apparently without postulating inner presentations. Field argues that the only way that such a view can fail to postulate inner representations is if in states such as John believes "snow is white" and John believes "snow is white and snow is cold" the two occurrences of "snow is white" are treated as mere orthographical accidents, but this is implausible. (To put it another way, everyone thinks there is at least some compositionality to thoughts, e.g., truth-functional compositionality, but according to the envisioned functionalist view, there is absolutely no compositionality whatsoever.) Parts IV and V anticipate some subsequently raised issues about linguistic thought. Contra some current critics of LOT, Part IV suggests that the LOT theorist can turn to empirical psychological theories for the specification of mental syntax. Part V can be read as a counter to Stich (1983)--the role of truth in psychology is defended. Part VI returns to a recurring theme of (Parts I and III); it is argued that a possible worlds account of meaning will not be fine grained enough, e.g., it will not be able to distinguish the meaning of "Clay is Ali" from the meaning of "Ali is Ali." Overall, this paper does a nice job of showing that the postulation of linguistic thought is I) separate from Fodor's specific innate Mentalese hypothesis and that ii) there are broad philosophical considerations that support the idea of linguistic thought, independent of specific research programs in empirical psychology.

*Block, Ned, ed. (1981) Readings in the Philosophy of Psychology, Vol. 2, Cambridge MA: Harvard University Press. 

Collection of papers on four topics: (1) mental representation, (2) imagery, (3) the subject matter of grammar, and (4) innate ideas. Fair to good introductions written for the volume by Rey for (2), by Fodor for (3) and by Block for (4). Though much has been written about all of these topics since the volume came out, it is still useful as a source for significant early papers in these areas. Includes essays, available elsewhere as well, by Geach, Harman, Fodor, Dennett, Field, Kosslyn, Pylyshyn, Stich, Chomsky and Katz, among others. Also includes Chomsky-Putnam exchanges on Chomsky's linguistic innateness hypothesis. 

*Churchland, Paul M. (1981) "Eliminative Materialism and the Propositional Attitudes," Journal of Philosophy LXXVIII(2), pp. 67-90 

Belief-desire explanation, i.e. "folk psychology", is a theory. As such, it is a colossal failure: it fails to explain all sorts of mental phenomenon, it has stagnated in development for thousands of years, and it is not part of the synthesis of contemporary physical/chemical/biological/evolutionary explanations of humans. There are no convincing arguments to show that it can't be eliminated, so we should serious contemplate the possibility that it is false, and look for alternatives. (Churchland now maintains that the connectionist/neural networks view is the successor theory to folk psychology.)

Fodor, Jerry A. (1981a) "Propositional Attitudes," in his Representations, Cambridge, MA: MIT Press.   

Five main constraints on a successful theory of propositional attitudes are explained and defended: 1) The attitudes should be analyzed as relations, 2) the parallel between PA verbs and verbs of saying should be explained, 3) opacity should be accounted for, 4) the objects of attitudes must have logical form and 5) The theory should "mesh with" scientific psychological theories of mental processes. Fodor then shows how Carnap's "disposed to utter natural language sentences" view fulfills all of 1-5, but fails in that it gets the taxonomy of attitudes wrong (and fails in several other ways too.) The diagnosis is that it must be Mentalese sentences rather than natural language sentences that the attitudes are relations to. (Thus, it's a critique of the natural language view of thought. However, Carnap's view is a bit of a strawman--see Kaye 1995a or the LOT tour for discussion of the natural language view and taxonomy.)

Fodor, Jerry A. (1981b) "The Present Status of the Innateness Controversy", in his Representations, Cambridge, MA: MIT Press.

A very significant investigation of concepts. Defends the "Nativist" view of concepts--Fodor's radical concept nativism, which claims that most concepts are innate, and are triggered--against Empiricism--the definitional reductionist view that postulates a small set of innate concepts and see acquisition as a matter of combining concepts. Includes a clear presentation of what Fodor now terms "The Standard Argument" for radical concept nativism. The stereotype view is also sharply critiqued.

Sterelny, Kim (1983) "Mental Representation: What Language Is Brainese?", Philosophical Studies 43, pp. 365-82.

The title is somewhat misleading--the posed question is dealt with only very briefly at the end. Most of the paper involves an examination of (Fodorian/scientific) representational realism about belief-desire psychology. The main discussion concerns the taxonomy of belief--it is argued that representational realism, specifically linguistic realism, about folk psychology appears to provide the right causal taxonomy for beliefs and desires. Sterelny then defends realism against a number of Dennett's pro-instrumentalist criticisms that claim that realism gets the taxonomy wrong. Helpful in understanding/reviewing Fodor's RTM vs. Dennett's instrumentalism.

Stich, Stephen (1983) From Folk Psychology to Cognitive Science: The Case Against Belief. Cambridge, MA: MIT Press.

Part 1 is a mixture of investigation into the nature of folk (belief-desire) psychology and critical theorizing. The mental sentence view (roughly, Fodor's LOT hypothesis) is fleshed out and problems are raised about belief-typing. Externalist and holistic aspects of reference are used to critique a narrow causal-role account of belief. An alternative view of belief ascription is developed, and more reference concerns are raised. In Part II, all of this is brought to bear on representational theories of mind (Fodor being the main target here.) Both a strong and a weak form of representationalism are critiqued and rejected in favor of a purely syntactic theory of mental (roughly, computational) states. (In hindsight, if this was meant to be a normative correction to cognitive psychology, then it has not been obeyed, and if was meant to be a prediction, then it appears to be mistaken, since cognitive psychological theories continue to attribute representational (i.e., contentful) states and processes.)

*Block, Ned (1986) "Advertisement for a Semantics for Psychology," in P. French, T. Uehling and J. Wettstein, eds., Midwest Studies in Philosophy, Vol. X, Minneapolis: University of Minnesota Press, pp. 615-78. 

A very substantial defense of the view that conceptual roles semantics (CRS) is the most plausible view of meaning. (The conception is defended in the abstract--no specific CRS is proposed--that's why it's an "advertisement"!) Seven desiderata for an account of meaning are presented. It is argued that no other view of meaning satisfies all of them, whereas a two-factor (in the head plus connections to the world) CRS does have reasonable prospects for satisfying all seven. The middle part of the paper also serves as a brief (albeit biased) overview of the various philosophical conceptions of meaning. 

*Chomsky, Noam (1986), Knowledge of Language, NY: Praeger.   

The conception of language and of methodology for the study of language that was developed in works such as Chomsky (1965 and 1975 is re-developed and defended against the outlooks found in analytic philosophy. Under the typical philosophical picture (e.g. Quine, Davidson), language is conceived as a set of sentences with truth conditions. Chomsky argues that this external language (E-language) is not the primary linguistic phenomenon; rather it is derivative. It is the product of each person's internal language (I-language). The I-language consists of tacit knowledge of largely syntactic rules. It is only through studying such rules (grammars) and their innate basis that we are able delineate the E-language at all--that is, our very conception of the E-language relies on our tacit knowledge of the I-language . Thus, properties associated with the E-language such as reference and truth cannot be the primary basis for the study of language; indeed, Chomsky doubts that any systematic theory of these aspects of the E-language is possible. (N.B., although this view may seem to fit in with the wide/narrow content issue in the philosophical literature, it is actually a very different outlook. (It is not based on twin-Earth thought experiments or the like.) Chomsky thinks that some aspects of semantics, e.g., anaphors, will be included in theories of the I-language, but not referential truth conditions of the sort that philosophers are wont to focus on.) Also includes a defense of knowledge of rules against Kripke('s Wittgenstein) and against Dummett. Unfortunately, many readers have been mislead by what is mostly a terminological shift ("competence" = "I-language", "performance" = "E-language", "poverty of the stimulus" = "Plato's problem") into thinking that this marks a radical change in Chomsky's outlook. As Chomsky informed me circa 1989, it doesn't.                

Samet, Jerry (1986) "Troubles with Fodor's Nativism," in P. French, T. Uehling and J. Wettstein, eds., Midwest Studies in Philosophy, Vol. X, Minneapolis: University of Minnesota Press, pp. 575-94. 

In order to block Fodor's (1975, 1981b) "Standard Argument" for radical concept nativism, Samet suggests the metaphor of "catching" (a disease) to show that "unlearned" does not imply "innate." However, there is little elaboration to cash in the metaphor. Also contains some useful exposition of alternative, weaker (Chomskian) nativist arguments, and some discussion of the "what is a concept?" issue.

*Dennett, Daniel (1987) The Intentional Stance. Cambridge, MA: MIT Press.  

Ten separate essays, six previously published, with new "reflections" added. The main theme is the development of Dennett's instrumentalist view of belief-desire psychology: we don't literally have beliefs and desires but it is often useful to interpret behavior as though we do. Intentional concepts invoke a unique explanatory stance that is distinct from the (scientific) physical stance. Includes: critical discussion of Fodor's outlook, a treatment of the representational error problem (for Dretske) and considerations involving evolutionary biology--Dennett thinks that assumptions of adaptationist optimality parallel the intentionalist assumption of rationality.

Fodor, Jerry A. (1987) Psychosemantics. Cambridge, MA: MIT Press.

Defense of Fodor' preferred brand of scientific belief-desire realism. Chapter 1 contains the defense of b-d realism in general: the explanatory success of intentional psychology is best explained by postulating corresponding content-appropriate causally efficacious psychological states. Chapter 2 defends an individualistic view of content, Chapter 3 attempts to avoid meaning holism, Chapter 4 marks the initial presentation of Fodor's causal theory of content, and the Appendix provides reasons for thinking that thought is linguistic. There has been much subsequent discussion of the topics of Chapters 2-4 by Fodor and his co-authors and his views have substantially developed and changed; however, Chapter 1 and the Appendix are not outdated.  

*Jackendoff, R. (1987), Consciousness and the Computational Mind, Cambridge, MA: MIT Press. 

(Cognitive Psychology) A theory of cognitive architecture (interesting enough in its own right) is developed towards beginning to explain the phenomenological mind. The architecture essentially synthesizes a Chomskian view of grammar, Jackendoff's conceptual structure semantics, Marr's work on vision, and Jackendoff and Lerdahl's views of musical cognition; each of these components is thought to form (roughly) a separate faculty, containing a number of specialized modules or (largely) modular processes. (Mathematical knowledge is a glaring omission in this project.) This view is then brought to bear on explaining consciousness: making use of Marr's notion of levels of explanation, various aspects of phenomenal experience are explained by citing aspects of the computational architecture. (MIT computationalism meets consciousness--a very ambitious project!)  

*Schiffer, Stephen (1987) Remnants of Meaning. Cambridge, MA: MIT Press. 

Schiffer thinks that the only plausible view of meaning is one in which the meanings of utterances are Gricean-reduced to the meanings of thoughts, and the meanings of thoughts are type-reduced to brain states. He looks at all of the possible type-reductive views that he is aware of--propositionalism, Mentalese, conceptual roles, and Davidsonian sententialism are the main ones--and finds them all inadequate and unpromising. He then proposes an alternative to type reduction, namely token reduction. The resulting outlook is that although we have meaningful thoughts and utterances, there's no such thing as a theory of meaning; in particular, there is no such thing as a compositional semantics for any language. (Some may find that the highly analytic writing style tends to obscure the main points.)

Clark, Andy (1988) "Thoughts, Sentences and Cognitive Science", Philosophical Psychology, 1, no.3, pp. 263-78.

Big picture theorizing that attempts to defend connectionism against Fodor's LOT belief-desire realism; the result is largely confusion. Clark maintains that belief-desire explanation is holistic, which apparently makes it unsuited for processing explanation (why exactly?--we're not told), but apparently it is not implausibly holistic (why?--we're not told.) Clark develops an emergent realism about belief-desire psychology and a processing realism for connectionist theories. But in the end, we're briefly told that realism about high-level, linguistic conscious thoughts will be required anyway. But doesn't this by itself require a LOT hypothesis? (Ditto for language comprehension, by the way). And how can it be maintained that an explanation of conscious judgement will be completely divorced from an explanation of belief? All the important questions go unexplained and unacknowledged.

Sterelny, Kim (1989) "Fodor's Nativism," Philosophical Studies 55, pp. 119-41. 

A response to Fodor's main argument for radical concept nativism (now termed "The Standard Argument"). Sterelny argues against Fodor's nativist triggering account, and then sketches an alternative account using both conceptual roles and an externalist/causal view of reference-fixing. Prototypes are also brought in to help, but there's no clear explanation of how all of these ingredients fit together to yield a view of concept acquisition. (Incidentally, Fodor (1998a) appeals to triggered prototypes and a causal theory of content to get around the Standard Argument).

Braddon-Mitchell, David and Fitzpatrick, John (1990) "Explanation and the Language of Thought", Synthese 83, pp. 3-29.

The authors are not concerned with Mentalese, but with the general view that thought is linguistic. They try to show that there are alternative types of explanations (mainly connectionist) of the productivity and systematicity of thought that do not require the postulation of linguistic thought. However, there discussion is couched entirely in generalities; when one looks at specific examples of what they are proposing, it is absurd. They claim that evolutionary explanation can partially diachronically explain the matters that the hypothesis of linguistic thought seeks to synchronically explain. Does this mean that they think that our ability to comprehend an infinity of sentence contents can be explained by giving an infinite (or near infinite?) set of evolutionary explanations of our abilities to comprehend specific contents (e.g., "grass is green", "snow is green", etc.)? The point makes little sense vis-a-vis productivity. They also argue, citing Fodor (1975), that causal psychological explanations might be made purely in terms of lower-level instantiations laws. But this does not necessarily show that the high level representations are not compositional, and on the other hand, if they aren't, one is stuck with Schiffer's problem--one hasn't actually explained productivity or systematicity (see Kaye 1993a). And finally, it is not clear that the authors really understand productivity and systematicity--these properties have to do, after all, not just with apparent processes, but with the types of contents we can entertain. While the former may, as the authors note, seem like question begging in favor of symbol processing views, the latter is based on commonsense attributions and on introspection, and has (an extremely implausible) eliminativism as its only alternative.

Crane, Tim (1990) "The Language of Thought: No Syntax Without Semantics", Mind & Language, 5, no. 3, pp. 187-212.

In defense of Fodor, contra Stich (1983), Crane argues that there's no way of determining mental syntax without semantics, so Stich's vision of a purely syntactic psychology can't be right. Essentially, the idea is that there's no way of defining or theoretically specifying 'syntax' (e.g. "Non-semantic properties" includes way too much--e.g., biochemistry) so the only way to specify syntax is as that which gets interpreted semantically. Not really a very convincing line of argument: the best candidate for what makes syntax is form/shape. To counter this, Crane provides a cute little example of productive, systematic composition in the make-up of railway trains. The conclusion is supposed to be that its absurd to say that the trains have a syntax, but that's not absurd at all. The defender of Stich's view might simply point out that, not surprisingly, it's psychological/brain syntax rather than railroad syntax that psychology and linguistics are concerned with. (The mention of Mentalese at various points is misleading--it is general linguistic representational realism that Crane is concerned with, and really just RTM--not Fodor's specific Mentalese hypothesis. Otherwise, a good exposition of RTM and the linguistic thought view.)

Dunlop, Charles (1990) "Conceptual Dependency as the Language of Thought", Synthese 82, pp. 275-96.

Critical evaluation of Shank's Conceptual Dependency Theory. The interesting part, in this context, is that Dunlop, with some plausibility, treats this as an example of a specific LOT hypothesis. While it is not clear that the problems raised for Shank's view are general problems for LOT hypotheses, this paper does usefully show the types of problems that any specific LOT hypothesis (i.e., more specific than Fodor has ever ventured) will face.

Bonjour, Laurence (1991) "Is Thought a Symbolic Process", Synthese 89, pp. 331-52.

Argues against the view that thought is symbolic (e.g., Fodor/MIT computationalism--Bonjour calls it the "linguistic conception" but nothing about language actually enters into his arguments) on the grounds that this conception cannot account for our ability to have access to the contents of our thoughts. The argument turns on the assumption that to say that thought is symbolic is to say that its syntax and semantics are not inherently related--just as external words and symbols may be used to stand for anything at all, e.g., the syntax of 'green' does not figure in what the word means. But this is simply a mistake--this is not what Fodor and others mean to imply by saying that thought is linguistic or symbolic (see Crane 1990, pp. 192-3). Rather what is meant is that thoughts have the same sorts of syntactic structure (usually understood causally) as do corresponding types of symbols. The believer in linguistic thought is free to maintain that the syntax of thoughts (partially) determines (or participates in determining) their contents (as the Mentalese theorist will undoubtedly hold). Bonjour also seems to assume that our knowledge of our own thoughts' contents is such that thoughts must have intrinsic content. It is true that our thoughts' contents seem intrinsic, but without the dubious assumption that introspection reveals the true nature of the mind/brain, this point can't be used against the symbolic view. While the main argument is thus (probably doubly) unsound, serious problems about first-person knowledge of content are raised for both the causal view of content and the inferential role view. (The concerns apply equally to a historical/evolutionary account; however a non-inferential causal role view may avoid the problem raised for the inferential role view.) Generally, Bonjour is to be credited with drawing attention to the fact that naturalistic theories of content have failed to address, let alone explain, first-person knowledge of content.

Davies, Martin (1991)"Concepts, Connectionism and the Language of Thought" in Philosophy and Connectionist Theory, Ramsey, Stich and Rumelhart (eds.), Hillsdale, N.J.: Lawrence Earlbaum Associates, pp. 229-57.

Virtually identical in content to Davies (1992).

Egan, Francis (1991) "Propositional Attitudes and the Language of Thought", Canadian Journal of Philosophy, 21, 3. pp. 379-88.

Criticisms of two arguments for LOT (linguistic thought, actually) from Fodor (1987). The first is Fodor's appeal to "Principle P", roughly that constituently complex effects have constitutently complex causes. Egan rightly points out that there is no basis for thinking that this is a general principle of explanation. The second argument from Fodor is his appeal to parsing theory and computational perceptual theories to support linguistic thought. Here there is some confusion. While it is not entirely clear that the representations postulated by these theories are linguistic in nature, Egan thinks that Fodor's view requires that the theories describe something like commonsense propositional attitudes, and they clearly do not. But thought might still be linguistic, and there might be an innate Mentalese, even if belief-desire states and explanations are not part of ultimate psychological theory--representational content is the key issue here (so, N.B. realism about conscious occurrent thought serves the linguistic though arguments just as well as realism about beliefs and desires.) It is important to note that thus far, successful psychological theories are not ones that vindicate the commonsense attitudes. But it is also not clear if this is contrary to Fodor's outlook--if his realism implies that scientific psychology be coextensive with belief-desire psychology then it is, but, more plausibly, Fodor probably maintains only that scientific psychology will include b-d psychology and that's still an open question. In any case, Fodor may agree that these are not good arguments for linguistic thought, since he doesn't include them in his recent (1998a) list of arguments for LOT.

Schiffer, Stephen (1991) "Does Mentalese Have a Compositional Semantics?" in Meaning in Mind: Fodor and His Critics, Cambridge, MA: Blackwell.

In what is essential a slight addendum to Schiffer (1987), it is argued that the infinity of belief-contents does not entail that there is a compositional semantics for Mentalese since it is possible to have a merely token-reducing theory of Mentalese tokens.

Davies, Martin (1992) "Aunty's Own Argument for The Language of Thought" in Cognition, Semantics and Ontology, J. Ezquerro and J. Larrazabal (eds.), Norwell, MA: Kluwer, pp. 235-71.

A very careful little argument for linguistic thought: 1) If a cognitive process is systematic, then its inputs and outputs must have syntactic structure (a nice example of a drinks machine is presented to illustrate.) 2) Having thoughts involves mastering concepts which, in turn, involves knowing systematic (linguistic) inference relations. Thus, thoughts have (linguistic) syntactic structure. (2) is based on a neo-Fregean conception of content, thus, this argument is targeted at contemporary Oxonian outlooks--it, in effect, corrects Evans' (brief) rejection of mental syntax. (But see Rey 1995 who notes some apparent counterexamples to Evans' generality principle). The article concludes with a discussion of connectionism--why it doesn't undermine the argument, and related points.

DeWitt, Richard (1993) Vagueness, Semantics, and the Language of Thought", Psyche, 1(1).  avalaible online

A strawman is made to look hopeless. Part of Fodor's position is that semantics should "mirror" syntax in the language of thought, because he thinks that only syntax is causally efficacious--in order for content to be casually efficacious it must "piggyback" on syntax. Much has been written on the causal efficaciousness issue; DeWitt does not cite any of it, but presents his own very strange twist on the matter. He thinks that in order to avoid having meaning (given, as he puts it, by a "mature psychology") merely roughly correspond to syntax in a way consistent with Dennett's instrumentalism, Fodor's realism requires that there be completeness and soundness relations between syntax and semantics as there are in formal logical systems. He then argues that vague predicates undermine the possibility of such relations. But Fodor need not take the wild leap DeWitt proposes; in particular, Fodor has never made the absurd claim that semantic psychology will develop independently from syntactic specifications of Mentalese. Rather, I presume, Fodor would maintain that the connection between semantics and syntax will be given by psychology itself, as it already is in non-mature cognitive psychology. That is, psychological theories will specify both the syntactic and semantic aspects of representations and processes, thus explaining their relations.

Kaye, Lawrence J. (1993a) "Semantic Compositionality: Still the Only Game in Town," Analysis, 53, 1, pp. 17-23.

Schiffer (1987, Chapter 8) claims that a compositional semantics is not needed to explain how we can comprehend infinitely many sentences, but I argue that it is needed to explain a certain sort of productivity, viz., how our capacities for beliefs can exceed our sentence storage capacities. And Schiffer (1991) maintains that a token reductionist materialism can explain how it is possible to have infinitely many beliefs (in Mentalese) without implying the need for a compositional semantics, but, using an analogy with alphabet soup, I show that the token reductionist theory does not actually explain belief productivity.

Kaye, Lawrence J. (1993b) "Are Most of Our Concepts Innate?", Synthese, 95, 2, pp. 187-217. 

Exposition and critique of Fodor's (1975, 1981b) "Standard Argument" for radical concept nativism, viz., that because concept acquisition relies on the use of concepts already possessed by the learner, all concepts that cannot be definitionally reduced are innate, and since very few concepts do appear to definitionally reduce, it appears that most concepts are innate. After noting the reasons why we find such radical concept nativism implausible, I explicate Fodor's argument, showing that anyone who is committed to mentalistic explanation should take it seriously. Three attempts at avoiding the conclusion are examined (Samet 1986, Sterelny 1989 and Block 1986) and found to be unsuccessful. I then present an alternative way around Fodor's nativism: I maintain that concepts at a given level of explanation can be semantically primitive, yet at least partially acquired if some of the conditions at a lower level of explanation that are responsible for the concept's presence are themselves acquired.

Lycan, William (1993) "A Deductive Argument for the Representational Theory of Thinking", Mind & Language, 8, no. 3, pp. 404-22.

The representational theory is that the brain has actual linguistic representations (not necessarily Mentalese, but Lycan doesn't make this clear)--that's (1) + (2) from the LOT tour. The argument is that thought has the same unboundedness as language does. Given physicalism, the "only game in town" is to postulate a set of representational primitives that, Chomsky-style, can be used to generate the unbounded realm of possible thoughts. That is, the only going account that explains the unboundedness of thoughts is a view that postulates a set of linguistic representations. This is essentially a dialectic variant on the productivity argument in Fodor 1987, but put more forcefully (as well it should be.) What follows is a clever analogy defending the methodological outlook (incidentally, held by most analytic philosophers) that broad, roughly functionalist, speculative empirical psychology is possible based only on commonsense knowledge of humans--worth reading in its own right. There is also a good critical discussion about why connectionism does not have any basis for resisting linguistic representational realism. High recommended reading.

Millikan, Ruth Garrett (1993) "On Mentalese Orthography", in Dennett and His Critics: Demystifying Mind, B. Dahlbom (ed.), Cambridge, MA: Blackwell, pp. 97-123.

(Includes response from Dennett, pp. 222-7.) Somewhat rambling discussion about what a slightly more detailed LOT hypothesis would amount to. Completely unsympathetic to Fodor--ignores many of his main points, thus this amounts to contemplation of a series of strawmen, by someone who really should know better. The basis for postulating compositionality (productivity, systematicity, etc.) is completely ignored. Much is made of the supposition that the primary idea behind the LOT hypothesis conception is an analogy with a formal (rule-governed) deductive system. But that is false--the primary idea is that of data structures with linguistic semantic and syntactic structure. Problems are also raised for typing individual terms in a LOT. Only physical and meta-rule typing are considered, and both are found wanting--functionalist syntactic typing vis-a-vis processors is largely ignored. It is argued that there is no basis for distinguishing a=b from a=a in a LOT. But consider this computational possibility: each unique term has an associated file of a set of descriptions. Identifying two distinct terms amounts to, say, treating the contents of their respective files as mutually interchangeable.

*Pinker Steven (1994) The Language Instinct: How the Mind Creates Language. New York: W. Morrow and Co. 

An exposition of Pinker's linguistic, psychological and philosophical views of language written for a general audience. The outlook merges a Chomskian view of innate universal grammar, Fodor's hypothesis of innate Mentalese and Dawkin's evolutionary adaptationism. The result: "evolutionary psychology".

Schiffer, Stephen (1994) "The Language of Thought Relation and Its Implications", Philosophical Studies 76, pp. 263-85.

Lots of exposition of the framework developed in Schiffer (1987) (and Schiffer 1991), with a slight new twist: it's possible to have a physical supervenience theory that explains the systematicity and productivity of thought contents but not have a compositional semantics for thought (=Mentalese sentences) if propositions (that are the meanings of thoughts) are not compositionally determined things. No explanation is given of what sorts of things non-compositional propositions could be. Schiffer still does not understand that a supervenience theory of X is not thereby an explanation of all of X's properties--see Kaye (1993a).

Yagisawa, Takashi (1994) "Thinking in Neurons: Comments on Steven Schiffer's 'The Language of Thought Relation and Its Implications'", Philosophical Studies 76, pp. 287-96.

Followed by reply by Schiffer. Nine specific critical points about Schiffer's framework are raised; Schiffer's replies to some of them help to clarify his view a little bit. A few others constitute a failure on Yagisawa's part to fully appreciate the Mentalese/linguistic thought hypotheses (as the title reveals).

Hauser, Larry (1995) "Doing Without Mentalese", Behavior and Philosophy, 23, 2, pp. 42-7.

"Point and Counterpoint" exchange with Abbott (1995). There's great confusion here: Hauser runs together, on the one side, linguistic representational realism, and the specific Mentalese hypothesis and on the other side, some sort of neo-behaviorist view of thought and the hypothesis that we think in natural languages (e.g., see Kaye 1995a). He raises and briefly discusses a number of (seven) arguments that are supposed to undermine the former view(s), in favor of the latter, but, given the extreme lack of clarity in the delineation of a target, it's hard to make sense of any of the discussion.

Abbott, Barbara (1995) "Thinking Without English", Behavior and Philosophy, 23, 2, pp. 49-55.

"Point and Counterpoint" exchange with Hauser (1995). Abbott replies to the arguments against Mentalese that Hauser offers, and sketches or mentions some arguments and evidence that favor the Mentalese hypothesis. A good job of clearing up the many confusions in the Hauser paper.

*Chomsky, Noam (1995) The Minimalist Program, Cambridge, MA: MIT Press.

In the last essay of the volume, Chomsky's transformational/computational theory gets radically revised: S(surface)-Structure and D(deep)-Structure get abandoned. Only PF (phonetic form) and LF (logical form) remain. There are many resulting changes in postulated operations as well. A little philosophy here and there, but mostly just linguistics.

Kaye, Lawrence J. (1995a) "The Languages of Thought," Philosophy of Science, 62, 1, pp. 92-110.  

Various forms of the language of thought (LOT) hypothesis are critically examined. Many considerations, including the complexity of representational content and the systematicity of language understanding, are raised in support of the view that some, but not all, of our mental representations occur in a language. Next, several arguments for LOT are evaluated: considerations involving sententialism and the propositional attitudes, Fodor's arguments concerning infant and animal thought, and Fodor's argument for radical concept nativism. It is argued that none of them provide good reasons for postulating a LOT that is innate or otherwise distinct from spoken languages. Instead, it is suggested that we maintain the more conservative hypothesis, supported by introspection, that some of our thoughts occur in the languages that we speak.

*Kaye, Lawrence J. (1995b) "A Scientific Psychologistic Foundation for Theories of Meaning," Minds and Machines, 5, 2, pp. 187-206.  

I propose, develop and defend the view that theories of meaning--for instance, a theory specifying the logical form or truth conditions of natural language sentences--should be naturalized to scientific psychological inquiry. This involves both psychologism--the claim that semantics characterizes psychological states--and scientific naturalism--the claim that semantics will depend on the data and theories of scientific psychology. I argue that scientific psychologism is more plausible than the traditional alternative, the view that a theory of meaning is a priori. After defending scientific psychologism against several objections, I offer a speculative proposal that shows how a theory of meaning can be integrated into scientific psychology.

Pessin, Andrew (1995) "Mental Syntax: Between a Rock and Two Hard Places", Philosophical Studies 78, pp. 33-53.

An interesting issue is raised, viz. what sort of theory of syntax can the Mentalese theorist (really just the linguistic thought theorist give), a number of possibilities are raised and rejected, and the author concludes that LOT hypotheses (and, apparently, all symbolic psychology) are no good. This is a bad paper for at least four reasons: 1) Pessin assumes the falsity of type identity for mental states, but i) the general status of the identity theory is an open question and ii) it's always possible that some mental states are neuronally type specifiable even though others aren't. 2) An extreme holism in syntactic roles, of a sort that effectively rules out functional specification is also assumed. But 3) the computer analogy strongly suggests that mental symbols will somehow be specified via a combination of type identity and functional role, yet Pessin fails to even mention computer symbolism (does he take his arguments to refute the existence of symbols in computers too?) Finally, 4) he fails to see what's really operative here: both general linguistic thought and the specific Mentalese hypotheses are broad empirical views. So it is perfectly reasonable for Fodor, or anyone else holding such views, to maintain that scientific psychology will ultimately be responsible for specifying the syntax, either generally, or, more plausibly, for each specific domain. Indeed, we can see this occurring already in areas such as computation vision theory (e.g., Marr) and parsing theories. (Such views tend to be functionalist, with some attempts at neural identity too.)

Rey, Georges (1995) "A Not 'Merely Empirical' Argument for a Language of Thought", Philosophical Perspectives, 9, pp. 200-22.

The argument is that we are able to engage in first-order logical reasoning, but since such transitions occur between thoughts in virtue of the thoughts' logico-syntactic structure, thoughts therefore must have such structure, which is clearly linguistic structure. The argument is carefully exposited; potential objections (e.g., Davidsonian) are considered. It is also made clear that Rey is not arguing that we always think logically, but only that we have this capability. This is a good argument, and stands alongside productivity, systematicity and the complexity of content as the strongest support for linguistic thought. However, one might question the billing: it is not clear why the fact that we are able to reason logically is anything other than a broad empirical fact, as are productivity, systematicity, the complexity of content and general compositionality phenomena (Rey indeed notes that the argument does not engage an instrumentalist such as Dennett.) Rey might also have discussed the varieties of linguistic inference, e.g., our ability to infer "that's a deer" from "that's a doe", which appear to support more than just the thesis that (some) thoughts have first-order logical structure.

Carruthers, Peter (1996) Language, Thought and Consciousness: An Essay in Philosophical Psychology. New York: Cambridge University Press. 

In the first half of the book, Carruthers defends the view that occurrent thought occurs in spoken languages against Fodor's Mentalese theory. He then turns to consciousness and proposes a "reflexive thinking" account which maintains that in order for a thought to be conscious it must be available to further acts of thinking; that is, it must be possible to have thoughts about the thought. Together, these views support the Thesis of Natural Necessity: given the (contingent) way that our cognitive architecture is designed, we require public languages in order to be conscious.

Clapin, Hugh (1997) "Problems with Principle P", Pacific Philosophical Quarterly, 78, pp. 261-77.

In the Appendix to Fodor (1987) it is argued that we can infer the compositionality of mental states from the compositionality of behaviors together with what Fodor terms Principle P, roughly that complex effects have complex causes. Clapin raises a number of problems for this principle, the main one being that skills do not segment in a way appropriate for Principle P to apply. However, there is very little convincing discussion about the obvious case in point, viz., verbal behavior and its causes. (Note that Fodor seems to have dropped appeal to Principle P as a basis for linguistic thought-- Fodor (1998a) makes no mention of it.)

Cole, David (MS 1997) "Hearing Yourself Think"    avalaible online

Defends the view that (occurrent) thought consists of acoustic images of natural language sentences. Cole replies to Fodor's animals argument and the Standard Argument (both for LOT), offers some evidence against Mentalese, replies to the ambiguity objection to natural language thought, and offers a greatly weakened version of the Whorfian hypothesis.

Garson, James (1997) "Syntax in a Dynamic Brain", Synthese 110, pp. 343-55.

Critical discussion about LOT that's really just discussion about symbolic thought in general. (This is a response to a paper by Horgan and Tienson that is primarily about GOFAI; Garson treats their sketchy remarks about high-level symbolic representations as a LOT hypothesis.) There's lots of talk about how syntax is not causally real in connectionist networks, but this ignores arguments for linguistic thought that also imply that such networks don't explain thought at all, e.g., because they don't explain its productivity or systematicity.

Laurence, Stephen and Margolis, Eric (1997) "Regress Arguments Against the Language of Thought", Analysis 57.1, pp. 60-6.

The authors detect the following regress argument against the general hypothesis that thought involves linguistic representations (N.B. not Fodor's specific Mentalese hypothesis): Natural languages are learned, and linguistic thought is postulated to explain this, but either its learning needs to be explained or the explanation might have been applied to natural language in the first place. The authors note how Fodor (1975), and also Crane, respond to (variations on) this argument and then present what they think is a better response: simply that there are independent reasons to postulate linguistic thought. This paper involves a very bad dialectic misunderstanding (which may stem from Blackburn or from Crane, both cited by the authors) of Fodor (1975). He actually raises regress as a general problem (for any view) that only the (specific) Mentalese hypothesis can solve: To learn a language you need to know a language; there can't be an infinite regress, so some language must be not learned (=innate). Natural languages are learned. Solution: postulate innate Mentalese. Since the paper solves a problem that was not there to begin with (vis-a-vis Fodor), it is best avoided, especially by those seeking to develop an initial understanding of the LOT issue.

Rantala, Veikko and Vaden, Tere (1997) "Minds as Connoting Systems: Logic and the Language of Thought", Erkenntnis 46, pp. 315-34.

Another strawman goes down hard. The authors argue, focusing on Fodor and Pylyshyn's attack on connectionism, that the LOT hypothesis requires that all syntactic transitions involve the application of truth-preserving (or other-epistemic-quality-preserving) rules. They argue correctly that this is implausible, sub-conclude that Classicism is really not in conflict with connectionism, and then suggest that a semiotic system approach to reasoning be taken instead of the Classical approach. The authors do not seem to understand what productivity and systematicity amount to--in several places they seem to think that this primarily concerns processes rather than contents. More importantly, they do not grasp that it is the mere hypotheses that some representations are linguistic and that some transitions are truth-preserving that explain productivity, systematicity and our ability to reason. The latter clarification undermines just about all of the points in the paper.

Weller, Cass (1997) "Bonjour and Mentalese", Synthese 113, pp. 251-63.

Otiose but interesting exposition of, commentary on, and development of Bonjour (1991). Weller shows how Bonjour's argument against symbolic thought assumes both the intrinsic nature of content and a perceptual model of first-person knowledge of content, which no symbolic/naturalist will grant. But at the same time, Weller develops a Kantian reading of Bonjour's argument (and really, the view in question is the representational theory of self-knowledge, not the symbolic view of thought) along the lines that the apperceptional aspect of self-knowledge--the "I think" of self-awareness--is what naturalistic/representational views of self-knowledge fail to explain. (The title is misleading: Fodor's specific Mentalese hypothesis is not discussed.)

Aydede, Murat (1998) "The Language of Thought Hypothesis", in Stanford Encyclopedia of Philosophy  avalaible online

This entry does a good job of showing how the LOT issue is in imbedded in surrounding issues, viz., belief-desire realism, computationalism, intentional realism, naturalism, and "Classicism" vs. "Connectionism", but does a lousy job of explaining LOT itself (the issue of which language we think in is not mentioned, nor is nativism, nor is Fodor's "Standard Argument" for innate Mentalese.) As such it is best read as a supplement to the present tour; it will also be very useful to those whose primary interest is in understanding Fodor's Weltanschauung. An expanded, in-progress version is available avalaible online

Fodor, Jerry A. (1998a) Concepts: Where Cognitive Science Went Wrong. New York: Oxford  

Much of the book is devoted to a very strong critique of the two standard views of concepts: the definitional view and the prototype view. In the end, Fodor offers an alternative account: concepts are "lockings" to properties--roughly, causal correspondences, such as (but not limited to) the ability to pick out instances of the concept. Typical concept acquisition (yes, you read correctly!) involves a property (not reducible, else the concept would be too) causing a prototype, which in turn leads to a locking to the property. This non-cognitivist account of acquisition avoids Fodor's own Standard Argument for radical concept nativism (see Fodor (1981b)). Natural kind concepts are also treated, but it's argued that they are less widespread and important than philosophers typically assume. This is Fodor at his provocative, creative and architectonically precise best.

Fodor, Jerry A. (1998b) "Do We Think in Mentalese? Remarks on Some Arguments of Peter Carruthers" in In Critical Condition, Cambridge, MA: MIT Press, pp. 63-74.   

Criticisms of the argument from introspection (in support of the natural language of thought view), an explanation of why, contra Carruthers, the Mentalese hypothesis is not tied to the Gricean program, and some criticisms of Carruthers' non-atomistic semantics.

Kaye, Lawrence J. (1998) "Another Linguistic Turn?: Review of Language, Thought and Consciousness: An Essay in Philosophical Psychology by Peter Carruthers," Psyche, 4(2).  avalaible online

Review of Carruthers Language, Thought and Consciousness (1996). His reflexive thinking theory of consciousness is criticized on the grounds that conscious thinkers need not have conceptions of thoughts as thoughts. While indexical availability may be a key aspect of (human) consciousness, it does not follow, as Carruthers thinks, that language is essential to thought, or that children and animals are not literally conscious. Also includes critical discussion of Carruthers' attempt to develop an alternative to Fodor's Mentalese hypothesis.

Cole, David (MS 1998) "I Don't Think So: Pinker on the Thinker"  avalaible online

Critique of Pinker's (1994, Chapter 3) defense of a Mentalese (LOT), as opposed to the view that we think in natural language sentences. Cole discerns five arguments from Pinker in support of the claim that there must be a Mentalese, and five more arguments in defense of the view that Mentalese must be the exclusive medium of thought and provides strong critical responses to each argument. Includes discussion of imagery and animal thought, coining new terms, co-reference, context and ambiguity, and inexplicitness.

Knowles, Jonathan (1998) "The Language of Thought and Natural Language Understanding", Analysis 58.4, pp.264-72.

A response to Laurence and Margolis (1997). Knowles does a good job of defending and really just properly explaining the regress argument(s) for LOT. He argues that innateness does matter to the argument, and, even bracketing that, it is notapparent that LOT's ability to explain natural language understanding can be applied equally well to natural language itself.

Teng, Norman (1999) "The Language of Thought and the Embodied Nature of Language Use", Philosophical Studies, 94 (3), pp. 237-51.

Teng maintains that there is even more systematicity to thought than is usually mentioned by those such as Fodor who cite systematicity as evidence for linguistic thought. Teng suggests that this additional systematicity involves the interaction of perceptual and motor systems with the environment and is thus grounded in the environment itself. His conclusion is that the systematicity of the environment and these systems account for the systematicity of thought, thus undermining support for linguistic thought. But there's two gaping holes here. First, Teng fails to acknowledge that LOT theorists see the systematicity of perception and action as evidence for the postulation of representational linguistic perceptual and motive states. Second, he does not argue (or even recognize the need to show) that perceptual/environmental systematicity also explains the logical systematicity that Fodor cites in defense of linguistic thought.

Top of page

E-mail