I - Some central points:

(1) Concepts are intimately related to word meanings.

(2) The semantics of natural languages is the semantics of Mentalese.

[All points of emphasis in the quotations from Fodor are in the original]

"English inherits its semantics from the contents of the beliefs, desires, intentions, and so forth that it's used to express, as per Grice and his followers. Or, if you prefer (as I think, on balance, I do), English has no semantics. Learning English isn't learning a theory about what its sentences can mean, it's learning how to associate its sentences with the corresponding thoughts. To know English is to know, for example, that the form of words 'there are cats' is standardly used to express the thought that there are cats; and that the form of words 'it's raining' is standardly used to express the thought that it's raining; and that the form of words 'it's not raining' is standardly used to express the thought that it's not raining; and so on for in(de)finitely many other such cases". ("Concepts", 1998. p.9)

(3) We adopt the Representational Theory of Mind (RTM). "Mental states and processes are typically species of relations to mental representations, of which latter concepts are typically the parts" (p.7)

"The idea that there are mental representations is the idea that there are Ideas minus the idea that Ideas are images" (p.8)

"MOPs have to be sliced a good bit thinner than senses. Individuating MOPs is more like individuating forms of words than it is like individuating meanings." (p.17)

MOPs are "entertained" or "grasped". This does not mean that you "think about a MOP". Just as for rocks and numbers, there are innumerably many ways of thinking about a MOP. Thinking with a MOP is not thinking about it. There are many (innumerably many) ways of thinking about water. This does not mean that the concept WATER has innumerably many meanings (there are not innumerably many concepts of WATER).

Fodor wants to establish a uniqueness relation from a MOP to a concept. There is always more than one way to think about a referent. But Fodor wants the referent plus a MOP to determine a concept uniquely.

"MOPs are mental objects and referents aren't" [...] "Mental objects are ipso facto available to be proximal causes of mental processes; and it’s plausible that at least some mental objects are distinguished by the kinds of mental processes that they cause; i.e. they are functionally distinguished. Suppose that MOPs are in fact so distinguished. Then it’s hardly surprising that there is only one way a mind can entertain each MOP; since, on this ontological assumption, functionally equivalent MOPs are ipso facto identical" (p.19)

"Satisfying the metaphysically necessary conditions for having one concept never requires satisfying the metaphysically necessary conditions for having any other concept." (p.14)

"Wherever mental states with the same satisfaction conditions have different intentional objects (like, for example, wanting to swallow the Morning Star and wanting to swallow the Evening Star) there must be corresponding differences among the mental representations that get tokened in the course of having them." (ibid)

For some concepts (RED, DOG, DOORKNOB etc.) we are directly connected with their extensions, via personal experience. For other concepts (METAL, HYDROGEN, NEUTRINO etc.) the connection is indirect, inherited along a chain, by deference to other persons (the experts), books, eyewitnesses etc. The point is that their connection to the extension is direct (no indefinite regress).

In spite of considerable imperfection and vagueness, a notion of information à la Fred Dretske fits the bill. Objects, sets, events, situations etc. connect causally, and nomologically, to the mind via a suitably abstract power to convey information. (Fodor accepts this notion in its intuitive acception: the appeal to "information theory" proper is neither denied, nor presented as crucial).

II.1 RTM

Ever since "Psychological Explanations" (1974) Fodor insists that no external object can enter a psychological explanation (can be a cause of behavior, or thought) unless it is internally, mentally, represented somehow. Only objects "under a description" can be relevant to psychology. "No (mental) causation, without (mental) representation." That’s where Behaviorism failed miserably. "Giocasta" and "The mother of Oedipus" have the same extension, but we know well what tragedies can ensue from the intensional difference between these "descriptions".

II.2 A truth-functional semantics of mentalese

Conventions of "truthsaying" are as central to psychology as trajectories and forces are central to physics. The typical contents of propositional attitudes (such as believing, inferring, desiring, claiming etc.) are things that are intrinsically capable of being true or false. They are propositions (In a non-Platonist sense. They, in fact, typically possess syntactic properties).

Therefore:

Something like a Tarskian theory of truth applies. To the right of Tarski’s scheme we find "thoughts": Propositions in mentalese.

To the left, as usual, expressions of English (Chinese, Swahili etc.).

The criteria for inter-translatability are left to the syntactician, and are tied to the forms "standardly used to express the thought that p".

Mentalese has expressive powers "at least" as great as those of any natural language. To learn English is to learn the forms of words and expressions standardly used to express thoughts. Our access to thoughts (our access to formulae in Mentalese) is un-mediated.

The ontological status of Mentalese is prior to the ontological status of any natural language. It has many of the properties of natural languages (notably recursiveness, generativity, compositionality). It lacks (of course) phonology and morphology. It lacks expressiveness. It is a medium of computation, not of communication. It is a metaphysical possibility that we could communicate directly in Mentalese. Being what we are, of course, we cannot, but there is this distinct metaphysical possibility.

II.3 The strict relational character of propositional attitudes

John’s belief that Rome is beautiful is literally a relation between John and an internal mental representation that has the content "Rome is beautiful". The formula of Mentalese "Rome is beautiful" must somehow be tokened in John’s mind. By compositionality, the concepts "Rome" and "beautiful" must also be tokened in John’s mind. There cannot be anything we are allowed to call John’s belief that Rome is beautiful unless these necessary conditions are satisfied. The objects of desires, beliefs, doubts etc. must be "sliced as thin as" sentences are. They are sensitive to forms (possess syntactic properties). They are sliced "thinner than Fregean senses". Mental representations are syntactically structured. Mental processes are computations, and are therefore syntactically driven.

II.4 Mental contents are ontologically prior (not parasitic on capacities dispositions, skills, knowing how)

That concepts are capacities (say, possibilities of experience) is the standard Logical Empiricist position. The ontology of capacities (mechanisms, dispositions etc.) supplants the ontology of mental particulars. Fodor fights against this thesis.

Fodor (contra Dummett -page 4) does not contend with the position that the ontology of concepts (or of meanings) is intimately tied with the problem of what it is to have concepts (and meanings), and to acquire them. He is strongly against the idea that the epistemic problem of possession and identification is prior to, and ontologically dominant over, the problem of what concepts and meanings are. It's not language use, it's not capacities or abilities that are primary. If having concepts is having capacities (to recognize or to draw sound inferences), then concepts are not mental particulars, they are not things at all, and therefore a fortiori they are not mental things.

"Concepts are categories, and are routinely employed as such". (Page 24) Things in the world "fall under them".

"The thesis that concepts are mental particulars is intended to imply that having a concept is constituted by having a mental particular, and hence to exclude the thesis that having a concept is, in any interesting sense, constituted by having mental traits or capacities" (p.3)

Whatever a behaviorist semantics had that was wrong, one thing it had that was right (according to Fodor): It was a quintessentially atomistic semantics.

II. 5 Metaphysics versus cognitive science

"The questions with which theories of meaning are primarily concerned are metaphysical rather than epistemic. This is as it should be: understanding what a thing is, is invariably prior to understanding how we know what it is" (p. 5)

In his critique of Cowie (1999, unpublished) Fodor goes as far as saying that even concept acquisition in not a task for cognitive psychologists at all.

"What bestows content on mental representations is something about their causal-cum-nomological relations to the things that fall under them: for example, what bestows upon a mental representation the content dog is something about its tokenings being caused by dogs" (p.12)

II.6 Fodor’s long fight against IRS (Inferential Role Semantics)

There are meanings that are quintessentially inter-defined with other related meanings (Tuesday is inter-defined with Monday, Wednesday etc. and with "weekday"). IRS maintains that all meanings are inter-defined (see the impact of Kuhn’s theses on concept acquisition - Carey, Gopnik, Keil etc.). But this cannot be. Definitions are hopeless (see Fodor’s old paper on why not derive "kill" from "cause to die").

Other accounts of IRS won’t work either. Why?

First: Buying IRS would lead to circularity: "I can’t both tell a computational story about what inference is and tell an inferential story about what content is." (p.13)

Second: "an inferential role semantics has holistic implications that are both unavoidable and intolerable" (ibid) (see his book with LePore)

Third: atomism. "Satisfying the metaphysically necessary conditions for having one concept never requires satisfying the metaphysically necessary conditions for having any other concept." (p.14)

This accounts also for his polemics against Pustejovski, Jackendoff, and Hale and Keyser. No inference can be constitutive of the meaning of a concept. It is metaphysically conceivable that a mind can exist that possesses only the concept (the meaning) DOG, and nothing else (no "cat", and, most of all, no "animal" concept).

Some meanings are functionally individuated. But this is epistemic, not metaphysical. It is how you control their application, not what they are.

"Denying, as a point of semantics, that 'believe' has a functional definition is compatible with asserting, as a point of metaphysics, that belief [notice: no quotation marks] has a functional essence". ... "Ditto, mutatis mutandis, 'capitalism', 'carburetor', and the like". (p.8)

II. 7 Fodor on Putnam on "Twin Earth" and meanings being, or not being, "in the head" (See his "Psychosemantics", 1987)

The same "intension" can be connected with the wrong extension. Since meaning has two factors (an intension and its causal connection with an extension) intension cannot determine extension. Meanings are not in the head. (See also Tyler Burge on "arthritis" and "brisket")

Old Fodor’s solution: A complete counter-factual supporting account clears the way. The connection with the wrong extension is parasitic on the connection with the right extension, but not the other way round. Meanings are in the head, pace Putnam, and Burge.

New Fodor’s solution: MOPs and their counter-factual supporting nomological powers. MOPs are in the head, and are functionally individuated.

Another Problem: How to avoid the unwelcome consequence that coreferring expressions are ipso facto synonyms? That WATER means H2O?

Answer: Coextensionality is not sufficient for synonymy, after all. What's the extra ingredient? IRS (Inferential Role Semantics) has its own answer: What inferences one is prepared to draw. The concept H2O has the inference "contains hydrogen", but the concept WATER has no such inference. Co-extensionality plus identity of inferential relations constitutes synonymy, for IRS.

Fodor does not want this solution. Meanings are not even in part constituted by inferences. Thinking is in part constituted by inferences (as Turing had it). Thoughts reduce to computations (operations) on symbols. But not meanings (concepts). Content is exhaustively constituted by symbol-world relations. No inferences enter the picture, though they do enter for thinking, but not for meanings, concepts and contents.

"Concepts are distinguished along two (possibly orthogonal) parameters; viz. reference and Mode of Presentation" (p. 15)

WATER and H20 have the same reference, but not the same MOP. Different MOPs connect differently to an extension. You do not have to make WATER and H2O synonyms. This is a good thing, because you do not want to make them synonyms.

"MOPs are mental objects and referents aren't". "Mental objects are ipso facto available to be proximal causes of mental processes; and it’s plausible that at least some mental objects are distinguished by the kinds of mental processes that they cause; i.e. they are functionally distinguished. Suppose that MOPs are in fact so distinguished. Then it’s hardly surprising that there is only one way a mind can entertain each MOP; since, on this ontological assumption, functionally equivalent MOPs are ipso facto identical" (p.19)

This does not commit Fodor to holding that the individuation of thought content is functional. MOPs are needed, in connection with the referent, to individuate and activate a concept (one extension + one Mop = one concept, and one only). This process is functionally characterized, but the concept’s content is not.

"The Tarski theory is compositional assuming a lexicon; it says things like `a is F' is true iff the individual designated by a has the property expressed by F'. The lexicon tells you which individual and property these are." (EF)

In reality, it’s five plus one: (see Fodor and LePore, 1992)

III - 6 Concepts cannot be prototypes

Fodor’s main argument is that, if DOG is a concept organized around the prototypical DOG, because of compositionality, so must be NON-DOG. But this is crazy, because a bagel and Beethoven’s 5th Symphony, and zillions of other desperately heterogeneous things, are NON-DOG. Nothing can possibly count as the prototypical NON-DOG. On top of this, you have to add the notorious problems with the conjunction of prototypes (PET FISH and MALE NURSE etc.). The case is desperate.

Ever since "The Language of Thought" (1975) and the debate with Piaget (1976, published 1980) the thesis that has been attributed to Fodor is that "everything is innate". A better characterization is: "No (primitive) concept can be learned". As a consequence, since we do have many concepts, some of which, by necessity, have to be primitive, then it appears plausible to infer (non-demonstratively, of course) that they are likely to be innate.

IV.1 Target Thesis (in the Piagetian historical formulation, but this is not essential):

The child possesses at age X genuinely new and more powerful concepts than those already present in the child’s mind at all ages Y<X. These new and more powerful concepts are the result of constructive learning, i.e. creative and original cognitive manipulations of the less powerful ones, already present (thematization, assimilation, accommodation, abstractive reflection etc.).

(The empiricist would just suggest association. The connectionist progressive statistical convergence of an output vector onto the target concept).

IV.2 Fodor’s counter-argument (in synthesis)

(1) The only model of learning that we can make some sense of is inductive inference.

(2) Inductive inference is the tentative projection of a concept onto new instances, checking for confirmation and disconfirmation.

(3) In order to carry out this process, you have to have the concept already actually present in (not just potentially accessible to) your mind.

(4) If it is a genuinely "more powerful’ concept than the ones you have already, then you cannot project it at all (even less to "new instances").

(5) Therefore, either it is not "more powerful", or, if it is, it cannot be the result of learning. ("in any sense of ‘learning’ that we even begin to understand").

Consequence: Any concept that the child can learn by inductive inference is at most (not at least) as powerful as some concept that the child possesses already.

Immediate rebuttal (by Hilary Putnam): If Fodor is right, then also the concepts BUREAUCRAT and CARBURATOR have to be innate. And that’s an absurdity.

Fodor’s reply (around the late Seventies, and until recently): So be it! There may well be non-activated concepts (protoconcepts) in anyone’s mental lexicon. They are present, but they have not (yet) encountered the relevant and specific experiential connection to the world (they have not been "triggered" yet by the relevant piece of evidential data, assorted with a suitable morphophonological label).

Fodor’s actual (after ’98) reply: If (underlining if) there really is a valid argument to the effect that the innateness hypothesis is unstoppable, and it has to encompass BUREAUCRAT and CARBURATOR, so be it. But maybe this is not the case. The repertory of innate mental capacities and dispositions and contents has to be quite large, but maybe not that large.

IV.3 The central dilemma of concept acquisition.

Inductive inference would have had the advantage of being content-sensitive and evidentially based. Under this hypothesis, a concept would be acquired on the basis of a semantic evaluation, and a match between this evaluation and the relevant experiential data.

"According to the hypothesis-testing model, the relation between the content of the concepts one acquires and the content of the experiences that eventuate in one’s acquiring them is evidential; in particular, it’s mediated by content relations between a hypothesis and the experiences that serve to confirm it." (p.127).

Indeed it’s the only hypothesis that would have this advantage. On the contrary, a "trigger" can be efficient in "turning on" a concept, even if it has no nomological connection with the mental content of that concept. The relation is "brute causal". The sight of a giraffe might trigger the acquisition of the concept DOORKNOB, given suitable conditions (ethology offers many perspicuous examples).

But inductive inference is out, for the reasons we just saw. And mere triggers would constitute a weak, non-nomological, explanation.

Impasse! (Pop-Darwinism would offer a causal explanation for the trigger-concept connection, but that’s unacceptable for reasons of circularity. What, if not inductive learning, could be the mechanism by which evolution selects the causal relation between experience and concepts?).

"It can’t be that inductivism about the acquisition of primitive concepts is both circular and mandatory". (p.130)

IV.4 The way out

"The relation between the content of a sensory concept [like RED] and the character of its cause is not arbitrary when the cause is intentionally described. The thing to keep your eye on is that we typically get the concept RED from (or, anyhow, on the occasion of) experiencing things as red." (p. 131)

Psychophysical descriptions of the triggering experience may well look arbitrary, but the intentional ones don’t. Things in the world trigger the acquisition of concepts not for what they are, but in virtue of the power they have to "strike us as" exemplars of the concept. How they can do that is to be explained in terms of a property of our minds. Different minds could interact with things, and react to things, in quite different ways.

Causally interacting with doorknobs is necessary (metaphysically necessary) for DOORKNOB acquisition. But it is not sufficient. You also need a contingent, psychological mechanism to supply the sufficient conditions. Barring inductive learning, what can that be?

Hint: Generalize the "sensorium" hypothesis. Make "appearance concepts" encompass also non-sensory concepts. Doorknobhhood, i.e. the property of being a doorknob is "that property that minds like ours come to resonate to in consequence of relevant experience with stereotypic doorknobs" (p.137)

How can that happen? In virtue of the properties doorknobs have qua typical doorknobs, and in virtue of the properties our minds have.

"We have the kinds of minds that often acquire the concept X from experiences whose intentional objects are properties belonging to the X-stereotype." (p.138)

How much such experience? Under which conditions? This is for the psychology of cognitive development to "fill in". But that this general account has to be so, is a metaphysical necessity.

This applies to language acquisition too. From exposure to STEREOTYPICAL ENGLISH SENTENCE (in motherese) the child acquires the concept ENGLISH SENTENCE, not the concept STEREOTYPICAL ENGLISH SENTENCE. The child is not "locked onto" just the NVN types of sentences of motherese, but onto sentences of English in general, also those that are not of the form NVN.

Our minds are "functions from stereotypes to concepts". We typically acquire a concept "from instances that exemplify the stereotype".

What has to be innately given is whatever mechanisms determine that things strike us the way they do. A very, very rich kind of innate sensorium, but not limited to sensory data (things may strike us the way they do indirectly, in virtue of reports, deference to expert opinions etc.).

So, what’s innate is not Ideas, not intentional contents, not concepts as such, but mechanisms. Not even primitive concepts have to be innate.

Has Fodor turned into a profligate Empiricist?

"A lot of stuff that’s domain specific or species specific or both has to be innate in order that we should come to have the concept DOORKNOB (or for that matter, the concept RED)... [this is the case in anyone’s story that makes any sense].. The issue is whether it requires a lot of innate intentional stuff, a lot of innate stuff that has content." (p.143)

Once you finally and definitively give up the inductive inference model, and the hypothesis that meanings (conceptual contents) are constituted by inferential valences, you may give up the requirement for innate Ideas. Informational atomism does not require that there be any. It requires domain specific mechanisms for "locking onto" properties of things in the world.

"It entails that there can be no primitive concept without a corresponding property for it to lock onto" (p.165, final sentence of the book).

`learned' means something within hailing distance of induced. The reason I want to hold this is that I see no other way out of the `d/D' problem". [EF]

"So what's got to be innate, over and above the sensorium (and the mechanisms of inductive inference employed in prototype acquisition) is whatever it is that takes an X prototype as input and yields the X-concept as output. (I'm assuming, for all the familiar reasons [see III-6 above in this handout], that prototypes can't BE concepts.) Whether a theory that says that the doorknob prototype is learned but the concept DOORKNOB isn't counts as saying that the concept DOORKNOB is innate seems to me a terminological issue which one could well decide just to shrug one's shoulders at. I'm sort of hoping that the nativism issue, as traditionally formulated, will now become obsolete". [EF, Emphasis added]

Fodor’s opinion on all the data we have seen in this class:

There is a lot of confusion, and many attempted explanations just cannot be true. Immediate counterexamples come to mind (see his critiques of Pustejovsky, Jackendoff, Hale and Keyser). Many impossibilities of principle can also be thrown at these approaches.

The data that are really sound are all in the domain of syntax, not semantics.

A very revealing consideration: "give" allows the double dative construction, but not "donate". Fodor’s explanation: dative is trisyllabic. Period.

His hunch is that a lot of these data are entirely explicable in terms of some such morpho-syntactic considerations. No semantics is involved at all.