20th WCP:

1. Introduction: linguistic symbol and mental representation

A good deal of that important branch of contemporary philosophy which goes by the loose term of Analytical Philosophy lives, in my opinion, with a distorted representation of its own past, and consequently, with an inaccurate appreciation of the nature of its own achievements.

As a matter of fact the belief is widely spread among analytical circles that the birth of Analytical Philosophy is the result of a Linguistic Turn taken by its founding figures G. Frege and B. Russell and then expanded into various directions by their notorious or less notorious followers. (1) To put it in a nutshell, the Linguistic Turn — of which historians give in fact conflicting accounts — (2) can be characterized by saying that it turned (better said, intended to) every philosophical problem — and most of all every psychological and epistemological one — into a problem about language, or at least into a problem dependent upon problems about language. Accordingly, in the perspective opened by the Linguistic Turn, language is made the unique or at least the most central and fundamental object of philosophical investigation.

However it is no less commonly held that this bold and innovative stance on philosophical problems adopted by G.Frege and B.Russell was litterally overturned sometime in the late 50's and early 60's, when it appeared once again that the most central and fundamental issues in philosophy had to do with the cognitive part of the mind, and in particular that they were not only logically independent from, but also logically prior to, problems about language. Accordingly the Linguistic Turn is taken to have ended into a Cognitive Turn understood as a restoration of the primacy of philosophical psychology over the philosophical study of language, a restoration echoing the Cognitive Revolution which took place in the corresponding scientific fields.

Nowhere perhaps than in the case of the nature of mental intentionnality is this general picture of the evolution of Analytical Philosophy as driven by two successive and largely opposite turns better illustrated. In the wake of the initial Linguistic Turn the concept of mental intentionality is first supposed to have been in large part reduced to that of linguistic reference (the aboutness of linguistic units), thereby becoming at worst a superfluous notion, at best a derivative and secondary one. However, the Cognitive Turn is then believed to have reversed this reductive approach by making the notion of mental intentionality independent from, and logically prior to, that of linguistic aboutness.

This general picture is in my eyes extremely inaccurate. On the first hand, I do think that something like a Linguistic Turn occurred in the early days of Analytical Philosophy, but in a fairly complex way and at any rate posterior to Frege and (the first) Russell themselves. On the other hand I am not sure — and this is in part a consequence of this first remark — that to see the Analytical Turn as a Linguistic one is sufficient to capture its most essential features. As for the second essential element of the picture, namely the Cognitive Turn, it is precisely the contention of this paper that this notion should be modified.

Indeed I want to suggest that the claim to have reinstated the investigation of the mind as an area of Philosophy independent from, and logically prior to, the Philosophy of Language commonly put forward by most advocates of contemporary representationalism is a misleading one, and that much of this representationalism should better be seen as a radicalization of the Linguistic Turn. To put it in stronger terms, I want to argue that, under certain of its aspects at least and appearances notwithstanding, the Cognitive Turn should better be seen as the continuation by different means of the Linguistic one. My main reason for arguing in this way is that representationalists consider a mental representation to be nothing else than a mental word, even when they do not specifically believe in the existence of a language of thought.

Because of obvious time constraints I will have to make my point in a very limited form. I will concentrate on the restricted although central claim already mentioned that mental intentionality is a phenomenon independent from linguistic aboutness, and that linguistic aboutness is on the contrary derived from mental intentionality. Futhermore I will focus on a specific version of this claim to be found in Jerry Fodor's « Mental Representation: an introduction ». (3) The reason for my choice is that Fodor's well argumented position looks highly paradigmatic to me of the stand of contemporary representationalism on this issue, as well as of what goes wrong with it.

Behind its historical façade it should be clear that this problem has a purely theoretical dimension. What is first at stake is the real nature of the foundation relation introduced by a large body of contemporary philosophical theories of cognition between two central areas of the philosophy of mind and the philosophy of language, and consequently, between philosophy of mind and philosophy of language at large. Beyond this, the question of the correct direction to be given to such a relation is also naturally raised. In my opinion, this issue depends crucially upon the possibility of assimilating a mental representation to a symbol. Consequently, assessing this possibility is in my eyes the heart of the theoretical issues involved in the critical evaluation of the prevailing notion of the Cognitive Turn of Analytical Philosophy.

2. Fodor's double reduction:

I take it to be uncontroversial that Fodor's overall enterprise is an attempt to naturalize the intentional explanation of behaviour construed as a bona fide causal one. I will also assume without argument that this attempt is carried out along two main complementary lines of investigation: one of a 'purely philosophical kind' going from the naturalization of intentional states to the naturalization of intentional processes, and one belonging to 'speculative psychology' going in the reverse direction. Finally I will assume that in both lines of investigation, Fodor's strategy is twofold: it first reduces intentional explanation to a representational one, and then naturalizes the representationalist reconstruction of the intentional explanation thus obtained.

The reduction of the intentionality of mental states to the intentionality of mental representations offered in « Mental representation: an Introduction » belongs to the first phase of the first one of these two lines of investigations. It is in fact part and parcel of a more global reductive approach described by Fodor in the following terms: « The idea, to put it in a nutshell, is that it might be possible to pull off a double reduction: Firts derive the semantic properties of linguistic symbols from the intentional properties of mental states; then postulate a population of mental symbols — mental representations as one says — (4) and derive the intentional properties of beliefs and desires from these. Practically every major twentieth century philosopher, at least in the Anglo-American tradition has taken it for granted that such a double reduction is out of the question. But it increasingly appears that they were all wrong ».

A close examination of these two successive reductive steps is in order.

a) First step of the double reduction:

The first step of the double reduction starts with the empirically based claim that linguistic symbols and intentional states are endowed with very similar properties. Both have causal relations of different sort and both are about something else than themselves. Comparing the belief that the cat is on the mat with the English sentence expressing this belief, Fodor writes about the second one: « On the one hand it is a physical object...Qua physical object it has an indefinite complex of ...causal properties. For example it exerts a certain gravitational force upon the moon, it reflects light in various complicated ways... But on the other hand, this sort of physical object is special in that it also has a bundle of intentional porperties, much as we have seen that beliefs and desires do. So in particular it looks to be about something (viz the cat, or where the cat is, or the fact that the cat is on the mat.. ».In this case the (complex) linguistic symbol and the intentional state have the additional property of being about the same thing, whatever this thing really is. Fodor's claim can be represented with the following diagram:

Fodor then argues that such parallelisms between phenomena are too much of an exception in nature to be treated as a mere coincidence. Good rational science should try to make them appear as two manifestations of a single process. One way of achieving this is by trying to reduce one to the other. And this is the solution opted for by Fodor: « This parallelism between symbols and intentional states is simply too striking to treat as accidental; it cries out for some sort of theoretical solution ». In his eyes, such a reduction can be carried in two opposite directions: « Either show how the intentional properties of mental states derive from the semantic properties of symbols; or show how the semantic properties of symbols derive from the intentional properties of mental states ».

In the first case, mental intentionnality is taken as the fundamental relation and linguistic aboutness as the derived one. It should be noted that the reduction is not a total one: intentional content is considered to be identical in both cases, but linguistic symbols remain different from mental states. Consequently an expression relation has to be introduced between symbols and mental intentionality: linguistic symbols are about something because they express what mental states are about. Their aboutness is an « inherited » one, and it is inherited from mental states through an expression relation. This first reduction can be represented in the following way:

In short, when I say 'the cat is on the mat', my sentence is about a certain state of affairs because it expresses a belief which happens to be about this very same state of affairs.

In the second case, linguistic reference is taken as the fundamental relation and mental intentionality as the derived one. I believe that the cat is on the map, so to speak, because I say 'the cat is on the mat'. One version of this alternative reductive picture consists in identifying a belief with a disposition to utter a certain symbol endowed with meaning. The diagram of this second solution goes as follows, where the nature of the relation between T1 and linguistic aboutness is left unspecified:

According to Fodor, Figure 3 corresponds to « the standard treatment of the mind/body relation in modern philosophy »; it is also clear to him that it does not work and should consequently be replaced by the more traditional view that « the function of language is to express thought » represented by Figure 2. In other words, Figure 2 and 3 can be read as the respective embodiments of the Linguistic and the Cognitive turns, and the Cognitive turn is thus clearly interpreted as a reversal of the Linguistic one.

b) Second step of the double reduction:

However, according to Fodor, the Cognitive turn should be more than a mere reversal of the Linguistic one. Indeed he thinks that the reduction can be carried one step further, in the sense that « we can think of beliefs and desires as inheriting their semantic properties from those of mental representations, just as we have been thinking of linguistic symbols (e.g. English sentences) as inheriting their semantic properties from the intentional properties of the beliefs that they are used to express. » As a first approximation, this second step could be represented in the following way:

However, this diagram is not fully accurate.The second reductive step is not the mere repetition of the first one. They differ on two counts. On the one hand, the second reduction is not based on the observation that a parallelism runs between two sets of phenomena. It is simply « suggested » by the first reduction. Accordingly, mental representations, contrary to linguistic symbols, are not anything given, and the teminus ad quem of this second reductive step has therefore to be « postulated ». On the other hand an intentional state and a mental representation are not two different entities, while an intentional state and a linguistic symbol are. Indeed Fodor's polyadic approach to intentional states makes mental representation a crucial element of them (an intentional state of a certain type of an organism O being nothing else than a representation in a certain causal/computational relation with O). Consequently it seems more appropriate to represent the second step in the following way:

And the nature of the overall relation established between mental intentionality and linguistic reference according to Fodor can therefore be represented as follows:

There is no doubt that such a theory aims at reducing linguistic reference to mental intentionality and thereby reversing the Linguistic Turn perspective on the issue: « On this view, Fodor writes, the semantic properties of mental representations are at the very bottom of the pile. It is mental representations that have aboutness in the first instance; evreything else that is intentionnal inherits its intentionality, directly or otherwise, from the intentionality of mental representations ».

3. A fraudulent reduction:

In my opinion, however, the second step of this double reduction somehow annihilates the results of the first one. Although Fodor sees it as a way of carrying deeper the foundation of linguistic aboutness on mental intentionality, I am afraid it is more accurate to see it as a way of falling back into what he calls « the standard treatment of the mind/symbol relation in modern philosophy ».

As a matter of fact, there are a number of indications that the real achievment of the Fodorian strategy is the assimilation of representational aboutness to linguistic aboutness, much more than the reduction of linguistic aboutness to representational aboutness.

Firstly, one of the main epistemological goals of a reduction is to reveal something about the nature of the phenomenon to be reduced: the reduciens is supposed to bring into light the nature of the reduciendum. Fodor's reduction, however, reveals nothing about the nature of linguistic aboutness by deriving it from representational aboutness, because representational aboutness is ipso facto identified with the aboutness of a linguistic symbol of a mental kind. Nevertheless such an identification carries an important information about representational aboutness itself: it is nothing trivial to treat it as a kind of linguistic aboutness. Therefore, the main information provided by the double reduction is not, as it should be, that linguistic aboutness boils down to representational aboutness, but instead that mentally representing something is not different from linguistically referring to something. In other words, the informative import of the reduction is opposite to its logical order.

Moreover, when carrying out a reduction, it is a basic requirement that the concept(s) to be reduced should be replaced by different one(s). And Fodor explicitly agrees with such a requirement when he later makes a third reductive step in Psychosemantics and attempts to naturalize representational aboutness itself. Indeed he writes on this occasion: « I want a naturalized theory of meaning: a theory of meaning that articulates, in non semantic and nonintentional terms, sufficient conditions for one bit of the world to be about...another bit ». (5) Accordingly, the reduction of linguistic aboutness to representational aboutness should articulate in non linguistic terms what it is for a word to be about something, while in fact it does articulate what representational aboutness is in linguistic terms, in the sense that representational aboutness is considered to be nothing else than linguistic reference transferred at the mental level.

Finally, not only are the fundamental bearers of intentionality assimilated to linguistic symbols of a mental kind, but the system they constitute is taken to be nothing but a formal language (this is in my view the specific point introduced by the Language of Thought Hypothesis stricto sensu). As a result, the theory of intentionality is turned into the semantical investigation of a new language ; it is indeed nothing else than a « Psycho-Semantics ». Moreover the division of labor is identical for psychological language semantics and for natural languages semantics, as the article Where do truth-conditions come from? (6) makes it clear: the psychologist's job is to provide a « truth-theory » for mental representations, and the philosopher's one is to provide a « theory of meaning », that is, an explanation of « the relation between a mental symbol and a state of affairs in virtue of which the latter determines the truth value of the former ». Psycho-Semantics is thus linguistics and philosophy of language applied to the system of mental representations. Its only significant difference with natural language semantics is that it investigates the most fundamental of all natural languages.

Is it not surprising, to say the least, that a reduction of linguistic aboutness to representational aboutness should have as a main result the total 'semantization' of representational aboutness and of its investigation ? Is this not a serious indication that the actual direction of the reduction might in fact be fairly different from what it claims to be? As a matter of fact Fodor has made several comments on the relation between the philosophy of mind and the philosophy of language which can be viewed as giving support to support this analysis: on various occasions, for instance, he explicitly refers to a « merging » of the philosophy of mind and of the philosophy of language.

Assuming now that the interpretation advocated here is correct, the question naturally arises of understanding why and how the double reduction functions, so to speak, in reverse order. And as a consequence, whether it should be considered as logically flawed, and in particular as involving a vicious circle.

The answer to the first question is extremely simple. As already indicated, the second step of the reduction in fact nearly reverses the first one. Indeed, the first step reduces linguistic aboutness to the intentionality of a mental state, while the second one reduces the intentionality of a mental state to a kind of aboutness which is in essence identical with the linguistic one, because it is assumed that a mental representation is just a mental word. Nothing essential distinguishes a mental symbol from a linguistic one from the point of view of intentionality : the only distinctive feature of mental aboutness is its fundamental character. Therefore the reduciens and the reduciendum are basically identical. And this is the reason why what is really achieved through the second step of the Fodorian reduction is the identification of linguistic aboutness to a linguistic aboutness of a more primitive kind. Accordingly, more than a reduction of linguistic aboutness to representational aboutness, it is a projection of linguistic aboutness onto representational aboutness.

However, the Fodorian reduction is saved from circularity thanks to the secondary properties (primacy, symbolic substance) distinguishing a mental word from a linguistic one. But its validity is a purely formal one in the negative sense of the term. Formally there is no question that Fodor legitimately reduces linguistic aboutnes to representational aboutness; nevertheless the substance of the reduction is to transfer linguistic aboutness to the level of mental representation. It is not an invalid reduction, but it is a fraudulent one. Fodor pretends to eliminate linguistic aboutness in favor of representational aboutness, while the significant step he actually takes is just the opposite: it is that of assimilating representational aboutness to a form of linguistic aboutness. And this certainly does not count as a reversal of the linguistic « treatment of the mental/symbol relation »: it is a radicalization of it.

4. Conclusion: an unsupported identification

There are two main theoretical problems raised by this unveiling of the real nature of the Fodorian reduction : 1)Is the claim that mental representations are nothing else than linguistic symbols of a primitive kind a true one ? 2)Does the Fodorian strategy at any rate offer any support in its favor?

I will conclude with a very brief statement of my position about these two issues, without offering any real argument for it.

In the first place, I think that the identification of a mental representation with a symbol (and a wordlike one especially) is a major source of difficulties for contemporary representationalism. Among other things, I am skeptical about the possibility of offering an adequate reconstruction of intentionalism on the basis of such an identification. I also think that it probably lies at the root of many of the difficulties encoutered by modern attempts at naturalizing intentionality. Accordingly, I tend to believe that one of the most important challenges facing the naturalization of intentionalism is not to do without the notion of representation, but to elaborate a non symbolic notion of mental representation. (7)

At any rate, treating a mental representation as a symbol at least requires a susbtantial theoretical justification. But it probably is the most central prejudice of contemporary philosophy of mind. It has been taken over from classical theories of cognition (be they rationalist or empiricist) without questioning . Interestingly enough, Fodor himself does not offer any direct support in favor of such an identification. Indeed, I think that all the arguments given throughout his work in favor of the second step (such as the adequacy to the constituency and productivity of thoughts as well as to intentional causality) are arguments in favor of the language of thought hypothesis stricto sensu, that is, the thesis that a mental representations system has a combinatorial organization identical to that of a formal language, and not the thesis that mental representations themselves do not differ from symbolic elements. To the best of my knowledge, his only substantially developed argument to the effect that, if it is reasonable to assume that a system of mental representations is structurally identical with a formal language, then mental representations should be identified with linguistic symbols of a new kind, is to be found in the context of his criticism of J. Bruner developmental theory. (8) However, I think that it can be shown to be both questionable and too limited in scope.