From the ALWS archives: A selection of papers from the International Wittgenstein Symposia in Kirchberg am Wechsel

1. Introduction

The concept of Gedanke in Wittgenstein's Tractatus is a controversial one. The few elucidations he gives of it seem to give rise to two opposing interpretations. On the one hand, a thought is said to be a logical picture of facts (TLP§3) and a proposition with a sense (TLP§4). This suggests a non-mentalist reading of the concept, i.e. one according to which a thought is, in its essence, no different from a written or spoken sentence: both depict states of affairs in virtue of being combinations of signs that reflect the logical form of reality. A non-mentalist interpretation of thoughts has been advocated, e.g. by Mounce (1981), Winch (1987) and Carruthers (1989). On the other hand, Wittgenstein makes a distinction between a proposition and a propositional sign, and says of the latter that it needs to be projected onto reality before it is a genuine picture (TLP§3.12). The method of projection is "to think of the sense of the proposition" (TLP§3.11), which gives the impression that a mental process of thinking is needed to give life to signs; in other words, that thoughts (being the 'units' of thinking) are prior to, and necessary preconditions of, the possibility of linguistic representation. Such a mentalist or psychological interpretation has been advanced, among others, by Kannisto (1986), Summerfield (1992), Malcolm (1993) and Hacker (1996).

In this paper I attempt to defend the non-mentalist interpretation of Gedanke. My aim in doing this is, besides trying to contribute to a correct understanding of the Tractatus, to cut off certain connections that have been claimed to exist between the book and modern philosophy of mind. In particular, I try to refute Malcolm's and Summerfield's suggestion that, like Chomsky or Fodor, Wittgenstein needs to assume an underlying mental mechanism or a language of thought to account for the possibility of determinate representation. Therefore the paper is organized around a presentation and a critical discussion of (some of) Malcolm's and Summerfield's views. The main argument put forward against their views is that, whether such mental processes occur during language-using or not, speculation about them simply isn't relevant to Wittgenstein's picture theory of language. Far from taking psychological considerations to be essential to philosophical theories of linguistic meaning, the reading sketched here sees the Tractatus as criticizing all such theories and thus resembling, in this respect, his own later philosophy rather than contemporary philosophy of mind.

2. Malcolm's interpretation

The most straightforward example of a mentalist view of Tractarian thoughts is undoubtedly Norman Malcolm's. His interpretation relies heavily on Wittgenstein's reply to Russell's inquiry as to whether thoughts consist of words. Wittgenstein answered (with apparent impatience) "No! But of psychical constituents that have the same sort of relation to reality as words" (NB, p. 131). Malcolm draws from this the conclusion that "all thoughts are composed of mental elements" and "no thought consists of words, spoken or written." (Malcolm 1993, 32.). The relationship between thought and language is, accordingly, described thus:

"A thought is a structure with a sense. A meaningful sentence is also a structure with a sense. The view of the Tractatus would seem to be that when a thought is expressed in a sentence, what happens is that the sense of the thought is thought into the sentence. The physical sentence is given the same sense that the thought already has. Thus, there are two structures with the same sense. One structure is composed of mental elements, the other of words." (ibid., 32)

It follows that when a particular proposition is true, there are, according to Malcolm, three parallel structures: a state of affairs that consists of simple objects, a thought that is composed of mental elements, and a sentence that is a combination of perceptible signs. What these three separate structures have in common (by means of which the latter two can be pictures of the first) is the same logical form. Since the mental structure and the sign-structure represent the same state of affairs, they can be regarded as one and the same 'proposition' (and thus Wittgenstein's seemingly non-mentalist definition "Thought is a proposition with a sense" (TLP §4) is explained away) (Malcolm 1993, 33-4).

The second aspect in Malcolm's mentalist view of thoughts is that they are necessary so that our statements can have determinate meaning. Referring to TLP §4.002, he says that

"[t]he apparently vague statements of ordinary language are given their actually precise sense by processes of logical analysis that are largely unconscious. [. . .] processes of exact thinking take place at a subterranean, unconscious level as the Tractatus hints when it says that it is impossible to gather immediately from everyday language 'what the logic of language is' " (Malcolm 1993, 53-4).

It is this (assumed) feature of the Tractatus that Malcolm connects to Noam Chomsky's philosophy of mind - namely, to Chomsky's idea that behind language there are some underlying abstract representations which are related by swift mental processes to spoken or written words (although we can view neither the abstract forms nor the mental processes by introspection) (see ibid., 54).

Malcolm's three-layered model seems to involve redundancy: there is no need, in my view, to regard thought (mental elements) and language (physical signs) as constituting two separate and yet co-existent structures. A thought can, of course, remain unexpressed, in which case it consists of mental elements (whose nature is unknown); and in this sense a thought is, as Winch notes, a more general concept than a proposition (see Winch 1987, 12- 15). But when the thought is expressed, no identically structured psychical fact is needed behind the spoken or written words: the thought is then nothing over and above its expression - it is just the applied propositional sign (TLP §3,5), that is, a proposition with a sense (TLP §4).¹And most importantly, a thought can always be expressed: it has the same limits as language, it cannot do anything that couldn't be done with words, as is implied in many places in the Tractatus:

" 'A state of affairs is thinkable': what this means is that we can picture it to ourselves." (3.001)

"Everything that can thought at all can be thought clearly. Everything that can be put into words can be put clearly." (4.116)

"The limits of my language mean the limits of my world." (5.6)

"We cannot think what we cannot think; so what we cannot think we cannot SAY either." (5.61)

There is a clear identification of the powers of thought and language in Wittgenstein's letter to Russell, where he tries to explain the main point of the book:

"The main point is the theory of what can be expressed (gesagt) by props. - i.e. by language - (and, which comes to the same, what can be thought) and what can not be expressed by props, but only shown (gezeigt)" (Russell 1975, 350)

What can be expressed comes to the same as what can be thought - there is no need to make a fundamental distinction between them (such as their being on entirely different 'levels'). In short, the middle layer of Malcolm's model does not explain anything - after all, as Malcolm himself points out, the constituents of thoughts (whatever they are) stand in the same relation to reality as words. In the case where we have words, adding an identically structured level beneath them would be like buying several copies of today's paper in order to make sure that what it writes is true, to use a simile from Wittgenstein's later philosophy (PI §265). Multiplication of identical structures is of no avail, whether the problem is to find the essence of picturing or to test the reliability of the newspaper. I do not think Wittgenstein committed this kind of fallacy in the Tractatus.

What does seem to run counter to a non-mentalist view of thoughts is Wittgenstein's remark that "language disguises thought. So much so, that from the outward form of the clothing it is impossible to infer the form of the thought beneath it" (TLP §4.002). And yet he also says that "all of the propositions of our ordinary language are actually in perfect logical order, just as they are." (TLP §5.5563). These apparently contradictory remarks can be reconciled in the light of Wittgenstein's idea of analysis - i.e. his conviction that all propositions can be dissected into their ultimate constituents (elementary propositions, which are immediate combinations of names). In this form, all (possible) ambiguities in meaning are resolved and the exact sense of the sentence is clearly visible (see e.g. TLP §§3.25, 4.221). Now Malcolm, as we saw, takes this to mean that this analysis is something that actually occurs while we mean or understand something - that it is a mental process underlying the utterance (or interpretation) of propositions (Malcolm 1993, 54). Since Wittgenstein says that it is not humanly possible to grasp immediately what the logic of everyday language is (TLP §4.002) and since, in spite of this, we know exactly what we mean by our propositions, they must, thinks Malcolm, get their determinate sense at an unconscious level, in virtue of some rapid mental processes.

This postulation of a Chomskyan type of mental mechanism to account for determinate sense seems to me unnecessary in the context of the Tractatus. Wittgenstein only says that it is impossible to gather the logic of language immediately from our ordinary expressions, but this does not mean that their real logical structure could not even in principle be revealed through analysis. On the contrary, Wittgenstein explicitly says that "[w]hat a proposition expresses it expresses in a determinate manner which can be set out clearly" (TLP §3.251, e.a.). Thus the two-fold nature of our ordinary propositions (i.e. their being 'vague' and 'in perfect logical order' at the same time) means, rather than there being mental processes of meaning behind the words, that it is possible to reformulate the proposition so that its determinate sense can be clearly seen - to put the same thing in such a manner that all chances of confusion are excluded. That we do not actually need to perform this analysis in everyday conversation is explained by the fact that use reveals the exact, intended meaning: "what signs fail to express, their application shows. What signs slur over, their application shows clearly." (TLP §3.262). Ordinary language might contain such flaws as having only one name for two different things (like the word 'bank' in English), but the context in which the word is used makes it quite clear which meaning is intended. The understanding of everyday language requires familiarity with (often quite complicated) conventions.

Moreover, it should be remembered that Wittgenstein, unlike most modern cognitive scientists, made a sharp distinction between philosophy and psychology; and as he says in his letter to Russell from 1919, it is the matter of psychology, not philosophy, to find out what the relation between the constituents of thoughts is to the pictured fact (and also what the constituents of thoughts are) (NB, 130). For Wittgenstein's purposes, psychological investigations, although tempting, are just "unnecessary entanglements"; in TLP §4.1121 he explicitly warns us of getting mixed up with them. This, if nothing else, should speak against the connections that Malcolm (among others) has drawn between the Tractatus and modern scientifically-oriented philosophy of mind.

3. Summerfield's interpretation

Another commentator who sees Wittgenstein as offering a two-level theory of representation is Donna M. Summerfield. She thinks it is necessary to assume, in the context of the Tractatus, the existence of an underlying mental language - a language of thought - in order to avoid the infinite regress of interpretations, which ordinary signs lead us into:

"Ordinary linguistic signs can be interpreted in various ways. [. . .] If that to which we appeal in the attempt to determine the interpretation of one sign is itself a sign that can be interpreted in various possible ways, we risk launching an infinite regress of interpretations. [. . .] I argue [. . .] that the Tractatus assumes that, in order to stop the infinite regress of interpretations, there must be some representations, in some way within our grasp, that need no interpretation. I believe that the Tractatus assumes there are such ideal representations underlying written and spoken signs, that there is, in effect, a "language of thought." [. . .] it is by translating perceptible signs into a language of thought that we are able to interpret the ambiguous signs of natural languages." (Summerfield 1992, 224)

This is naturally linked to Jerry Fodor's language of thought -hypothesis - to the view that linguistic representation should be explained by appeal to mental representation; thus, like Malcolm, Summerfield sees the Tractatus as having "important parallels with contemporary work in the philosophy of mind and cognitive psychology" (ibid., 226).

Unlike the other commentators who stress the importance of thoughts in Wittgenstein's account of linguistic meaning, Summerfield makes a distinction between intrinsic and original intentionality, and ascribes to Tractarian thoughts only the latter. The difference between these two kinds of intentionality is this:

"If a state or event has original intentionality, its capacity to represent something other than itself cannot be explained by appeal to the intentionality of any other states or events (its "aboutness" is "first" or "original"), but its capacity to represent may nevertheless require explanation. If a state or event has intrinsic intentionality, its capacity to represent something other than itself cannot and need not be explained by appeal to anything other than itself." (ibid., 225)

That is, although Summerfield thinks that the regress of interpretations of natural language expressions stops at the level of thoughts, she still sees Wittgenstein as offering an explanation of how thoughts represent - namely, by being logical pictures of possible situations. More specifically, the intentionality of thoughts is explained by appealing to the linguistic features of thoughts, although it is not derivative from the intentionality of natural language expressions. (Summerfield 1993, 225.)

The question that immediately arises after this explanation is: what, then, makes thoughts so special, if it is only their logico-linguistic features that count? Why can't ordinary linguistic signs do the job? All the commentators who ascribe outstanding semantic powers to Tractarian thoughts should be able to show that thoughts are in some crucial respect different from ordinary language expressions. Now the first difference that naturally comes to mind (and to which e.g. Malcolm appeals) is that thoughts consist of psychical elements, sentences of physical elements, as Wittgenstein said in his letter to Russell. But Summerfield notes (correctly, I think) that it is not the mental features of thoughts that explain their intentionality: "it is only the logical features of psychical facts that are relevant [. . .] if thoughts have any other features, they are not features in virtue of which thoughts represent" (Summerfield 1992, 233). What else could the difference be, then? As far as I can see, Summerfield doesn't really answer this question. Her solution as to what makes thoughts of the Tractatus have superpowers is that they are, unlike our everyday expressions, logically perspicuous, obedient to the rules of logical syntax. Thus a sentence in a language of thought shows its sense immediately, so that no further interpretation is required (Summerfield 1992, 227-9). In other words, Summerfield's 'language of thought' corresponds to what Wittgenstein called 'elementary propositions'.

But can the logically transparent elementary propositions be said to constitute a language of their own (as Summerfield obviously means, since she talks, as we saw, of translation processes between the language of thought and natural languages)? I do not think so. After all, according to Wittgenstein, all our thinking and language-using must obey the same logical laws; it is not possible to think or speak illogically (TLP §§3.03, 3.032). This logic is not always immediately displayed in an ordinary language expression (if taken out of its context), but when it is rendered into a logically perspicuous form, i.e. analyzed into elementary propositions, it cannot be a matter of translating it into a different language, since "[a] proposition about a complex stands in an internal relation to a proposition about a constituent of a complex" (TLP §3.24) and "[t]he totality of propositions is language" (TLP §4.001). Both the analyzed and the unanalyzed form belong to the same logical space and are internally connected with each other; they must therefore belong to the same language - the only language which I understand ("der Sprache, die allein ich verstehe" (TLP §5.62)). The language of which Wittgenstein talks is not, of course, English or German or any other natural language, but rather what all languages must have in common in order to be able to represent states of affairs; and this common feature is not, for him, some underlying mental symbolism of the speakers of natural languages, but the common logic of all our pictorial modes of expression (cf. TLP §4.015: "[t]he possibility of all imagery [. . .] is contained in the logic of depiction"). An ordinary language expression and the elementary propositions that are its ultimate constituents are just two different ways of saying the same thing, but nevertheless using the same logic of depiction; and insofar as all meaningful propositions reflect it, they belong to the same language.

How should the problem of intentionality be solved, then? As can be remembered from previous discussion, Wittgenstein held already in the Tractatus the view that when lingustic signs are taken together with their application, with their logico-syntactical employment, their exact meaning reveals itself and no infinite regress of interpretations threatens. Thus, linguistic signs themselves have original intentionality - it need not be explained by appealing to the intentionality of some other state or event, such as an underlying mental sentence, but it does require reference to the use of signs accroding to linguistic conventions.

Summary

The aim of this paper has been to show that Wittgenstein makes no fundamental difference between thought and language in the Tractatus, because (1) a thought can always be expressed in words (its limits coincide with the limits of language) and (2) a thought represents states of affairs in virtue of precisely the same features as a proposition does - the constituents of (unexpressed) thoughts are similar to the words of language and stand in the same kind of relation to reality as spoken or written words. From the viewpoint of depicting, it is irrelevant whether the signs are mental or physical entities. I do not, however, mean to suggest that Wittgenstein denies the relevance of mental processes to language-using altogether. The point of the offered non-mentalist interpretation is simply that neither in Wittgenstein's early nor late philosophy do mental considerations help to answer the philosophical question of how representation through language is possible. Even if there do occur some rapid subconscious mental processes during meaning and understanding, they did not interest Wittgenstein and, insofar as they cannot even in principle be put into words, are not what he called 'thoughts'. Therefore his views in the Tractatus shouldn't be seen as bearing a resemblance to any theories - past or present - which involve speculations about the mental mechanisms of language-users. Nor does the emphasis of this interpretation on the logical features of thought and language mean that the whole book should be seen as being only about the logical preconditions of representation. The "fundamental idea" of Wittgenstein picture theory was, of course, that the shared logical form with reality (essential for any picture) can no longer be depicted - that "there can be no representatives of the logic of facts." (TLP §4.0312). That is, although whatever can be said or thought can be said or thought clearly, there are things which cannot be said (or thought) but only shown. It has not been my intention to undermine this distinction and the importance of the realm of the only showable (wherein perhaps the sole purpose of the book lies). I have only tried to show that, as far as thoughts and their role in the picture theory of language are concerned, it is only their logical or structural features that count.