Wittgenstein and Sellars on Thinking
Wittgenstein and Sellars on Thinking


I present Ludwig Wittgenstein’s criticisms of the idea that meaningful linguistic utterances have to be accompanied by mental acts of thinking their sense and the idea that thinking occurs in a language of thought consisting of mental signs. I try to show that Wilfrid Sellars developed an account of thinking on the model of language which can accommodate Wittgenstein’s challenges.

Table of contents

    When we speak about thoughts and thinking we have to be careful to distinguish three significantly different concepts associated with the terms ‘think’ and ‘thought’. First, there are so-called occurrent thoughts; dateable mental events constituting the mental activity of thinking. Second, ‘think’ is used in order to express or ascribe the propositional attitude of belief. Finally, we use ‘thought’ in a Fregean sense in which it is roughly equivalent to the technical term ‘proposition’. In what follows I will be exclusively occupied with the nature of occurrent thoughts. Accordingly, when I use the words ‘thought’ and ‘thinking’ they should only be understood as referring to occurrent thoughts and the mental activity of thinking.

    On one common interpretation the Tractatus logico-philosophicus presupposed an account of the nature of occurrent thinking and its relation to language that Wittgenstein criticized sharply in his later work. According to this account, thinking is a kind of speaking, which consists of ‘mental signs’ that correspond to the signs of our public language; and the meaningful use of language, whether written or spoken, consists of two parallel processes, operating with signs and mental acts of thinking the senses of these signs. Both of these ideas, the idea of a language of thought and what could be called a dual-process conception1 of meaningful speech, Wittgenstein later criticized pointedly.

    The dual-process conception of meaningful speech is quite natural. There seems to be a significant difference between the utterances of a competent speaker and the squawkings of a parrot, although both might produce exactly the same words. The difference, one might be inclined to say, is due to the fact that only utterances of the former kind are accompanied by acts of thinking, whereas the latter are a mere production of noise. It is these accompanying acts of thinking that make all the difference, without them the signs are ‘dead’ (cf. Wittgenstein 1958, 4). This impression is reinforced when we consider the fact that sometimes we speak with thought and sometimes we think without speech. It seems that in the latter case the process that accompanies meaningful speech simply goes on without its overt expression (cf. Wittgenstein 1953, §§330, 332). The motivating idea behind the dual process conception is the belief that signs and utterances are mere physical objects and events and therefore posses no more meaning or intentional content than stones, chairs and claps of thunder. Hence they have to be accompanied by some sort of mental process, for example sentences in a language of thought or mental images.

    The problem with this beguiling picture of the role of thought in meaningful speech is that whatever process we imagine thought to be, it won’t be able to achieve its supposed role of endowing otherwise dead signs with ‘life’ or meaning. In the Blue Book Wittgenstein asks his reader to imagine that we replace the inner process that is supposed to give our utterances meaning by an outer activity (cf. Wittgenstein 1958, 3ff, 33ff). If we think that, for example, the use of colour words has to be accompanied by mental images in order to be more than a mere production of noise, we should imagine that someone carries a colour chart with her, in which colour-samples are correlated with their names and which she consults whenever she speaks about colours (cf. Wittgenstein 1958, 3). If we are inclined to believe that the mere production of linguistic signs alone cannot determine their meaning, than we probably won’t think that producing sounds plus handling a colour chart will do the trick. Colour charts are just another means of representing colours and can be variously applied. The defender of the dual-process conception will have to assume that their use has to be accompanied by some further kind of mental process, let us say speaking to oneself in foro interno. But such an activity, as well, can go on with or without thought (e.g. when one absent-mindedly recites a poem in the imagination) which might induce us to postulate a further mental process and so on ad infinitum.

    Whatever parallel process or activity we imagine it does not seem to be any better to determine the meaning of our utterances than these utterances themselves. Wittgenstein, therefore, concludes that meaning an utterance or thinking its sense is not a distinct process or activity that runs parallel to the production of the utterance and is detachable from it (cf. Wittgenstein 1953, §339). What distinguishes meaningful from mere parroting speech is that only the former has a use (cf. Wittgenstein 1953, §43) in a language game. And what determines whether an utterance has a use or is a move in a language game is not something that accompanies the utterance but the manner and the circumstances of its occurrence. This is also the reason why it is perfectly coherent to imagine a people who only think out loud (cf. Wittgenstein 1953, §331).

    Closely related to the dual-process conception of thought and talk is the idea that the process of thinking occurs in some kind of inner or mental symbolism, in a language of thought, constituted by mental signs. In a letter to Bertrand Russell the young Wittgenstein himself appears to have endorsed this idea. There he replies to a query by Russell concerning his conception of thinking that ‘I don't know what the constituents of a thought are but I know that it must have such constituents which correspond to the words of Language’ (Wittgenstein 1979, 130).

    The problem with this conception of thinking is that it postulates an inner symbolism or language that has to be radically unlike any public language or symbolism we know. Whenever someone uses a public language, be it written or spoken, we can always ask what he meant by a certain sign or signs, say some name N that occurred in his speech. And the speaker will usually be able to further specify what he meant or to whom he referred. This is radically different in the case of thoughts. Contrary to public utterances which can be identified both as acoustical or visual occurrences and as intentionally contentful speech acts, acts of thinking cannot be separated from their intentional content. We cannot specify a thought independently of what it is a thought about; a thought is individuated by its intentional content; it is, as Wittgenstein says, the ‘last interpretation’ (Wittgenstein 1958, 34).

    Despite his poignant criticism of the dual-process conception of thought and talk and the idea that thinking occurs in an inner symbolism, even the later Wittgenstein repeatedly compared thinking to using language (cf. Wittgenstein 1953, §§319, 320, 331). It would be interesting to see whether there is a way of using language as a model to explain the nature of occurrent thinking, i.e. a way of conceiving of a language of thought, which is compatible with Wittgenstein’s criticisms. The account of occurrent thoughts Wilfrid Sellars first developed in his classic essay Empiricism and the Philosophy of Mind (EPM) seems to achieve exactly this.

    Sellars aims in EPM to develop a theory of thinking which retains the, as he calls it, ‘classical’ (Sellars 1963, 177) idea of thoughts as essentially episodic inner occurrences while departing from the classical tradition in several respects. Firstly, he rejects the Cartesian idea that all thoughts are, as it were, ‘self-disclosing’, i.e. cannot occur without the subject knowing them to occur. Thoughts, according to Sellars, are inner occurrences to which the subject has privileged access but which can occur without her being aware of it. Secondly, he sharply distinguishes thoughts from other inner occurrences such as feelings, sensations and mental images. Thirdly, he rejects the idea that self-knowledge is a kind of observation. Unlike observational knowledge, knowledge of our own thoughts does not involve sensations, such as visual or acoustical impressions. Finally, and most importantly he rejects the classical idea that ‘both overt verbal behaviour and verbal imagery owe their meaningfulness to the fact that they stand to… thoughts in the unique relation of "expressing" them.’ (Sellars 1963, 177) Instead he suggests an analysis of our semantic idioms that aims to show that semantical discourse about public linguistic behaviour does not have to be analyzed in terms of the intentionality of mental acts but “that the categories of intentionality are, at bottom, semantical categories pertaining to overt verbal performances.” (Sellars 1963, 180)

    At the time he wrote EPM Sellars didn’t have a fully worked out positive account of the semantic idioms (cf. Rosenberg 2007, 173). It was only later that he arrived at his mature view, which is probably best articulated in Meaning as Functional Classification (Sellars 2007, 81-100), published 18 years after EPM. According to this view the meaning of an utterance is determined by its functional role within a norm-governed linguistic practice. The rules governing this practice determine essentially three kinds of linguistic performances, corresponding to the categories of perception, inference and action. The first kind of performances or ‘moves in the language game’ are ‘language entry transitions’, that is linguistic responses to non-linguistic stimuli; the second are ‘intra-linguistic moves, i.e. transitions from one utterance to another, and the third are ‘language departure transitions’, utterances which are followed by non-linguistic performances such as raising one’s hand (cf. Sellars 2007, 87-88). All sentences that have an equivalent norm-governed role within the network of language entry transitions, intra-linguistic moves and language departure transitions within their respective languages have the same intentional content. It is this equivalence of normative-functional role that allows us to say that the German sentence ‘Schnee ist weiss’ means the same as its English translation ‘Snow is white’.

    Thoughts are, according to Sellars inner goings-on which have the same normative-functional role within a network of thoughts or, as he sometimes puts it, within the ‘game or reasoning’ (cf. Sellars 1963, 324) as the utterances which express them have within our language games. This equivalence of normative-functional role allows us to say that utterances and thoughts have the same content, similarly as equivalent statements in different languages share one and the same meaning. When we ascribe thoughts to a person we are ascribing inner episodes to her, which are characterized in purely normative-functional terms and which causally explain her behaviour. This explanation of human behaviour in terms of thoughts is, according to Sellars, similar to the explanation of the observable ‘behaviour’ of physical objects in terms of the postulates of scientific theories. However, unlike the postulates of scientific theories which can be literally in, i.e. be a proper part of, the object whose behaviour is to explained, thoughts are only ‘inner’ in a metaphorical sense, indicating that they are not directly observable, and are ascribed to the person as a whole and not to any part of her (cf. Sellars 1968, 169-70).

    Does Sellars’s elaboration of the analogy between thought and talk avoid the objections Wittgenstein levelled against the dual-process conception of meaningful speech and the idea that thought is a kind of speech in an inner symbolism? It is quite clear that Sellars doesn’t endorse the dual process conception of meaningful speech. His explicit aim in EPM is – as we saw – to ‘reconcile the classical idea of thoughts as inner episodes… which are properly referred to in terms of the vocabulary of intentionality, with the idea that the categories of intentionality are, at bottom, semantical categories pertaining to overt verbal performances’ (Sellars 1963, 180). This commits him to an explanation of the meaning of public utterances that doesn’t refer to inner episodes. He provides this explanation with his theory of meaning as normative-functional role in language games. Public utterances according to Sellars do not stand in need of an inner accompaniment in order to be meaningful. Indeed we ascribe thoughts to a person not in order to explain what she said but to explain why she said it.

    Let us now turn to the question of whether Sellars is committed to the problematic idea that thinking occurs in a language of thought which is constituted by the use of mental signs. It doesn’t seem so; for although Sellars claims that thoughts are analogous to public linguistic utterances, he does not subscribe to the idea that thoughts are realized by the operation of mental signs. The problem with construing thoughts as occurrences in an inner symbolism is, as Wittgenstein pointed out, that with respect to all forms of symbolic representation, be it spoken, written, gestured or painted, we can always distinguish the symbol from its content. We can always ask what a certain symbol, for example a word, or a set of symbols, for example a sentence, means. This cannot be done in the case of thoughts. Contrary to symbolic representations thoughts cannot be variously interpreted, they are, as Wittgenstein says, the ‘last interpretation’ (Wittgenstein 1958, 34). Accordingly Sellars characterizes thoughts in purely normative-functional terms, i.e. as occurrences which play a certain normative-functional role in the game of reasoning. Unlike linguistic expressions, which we characterize in semantical statements both with respect to their sign designs and the normative-functional role, which constitutes their meaning, thoughts, according to Sellars, are characterized only with respect to their normative-functional role. Since this role determines the meaning of linguistic expressions, an episode which is characterized only in terms of its normative-functional role, i.e. its meaning-determining properties, cannot be said to be variously interpretable.


    1. Hacker, P. M. S. 1990 Wittgenstein: Meaning and Mind - Vol. 3 of An Analytical Commentary on the Philosophical Investigations, Oxford: Blackwell.
    2. Rosenberg, Jay F. 2007 Wilfrid Sellars: Fusing the Images, Oxford: Oxford University Press.
    3. Sellars, Wilfrid 1963 Science, Perception and Reality, London: Routledge & Kegan Paul.
    4. Sellars, Wilfrid 2007 In the Space of Reasons: Selected Essays of Wilfrid Sellars, Kevin Scharp, Robert B. Brandom (eds.), Cambridge, Mass.: Harvard University Press.
    5. Sellars, Wilfrid 1968 Science and Metaphysics: Variations on Kantian Themes, London: Routledge & Kegan Paul.
    6. Wittgenstein, Ludwig 1958 The Blue and Brown Books, Oxford: Blackwell.
    7. Wittgenstein, Ludwig 1979 Notebooks 1914–16, G. E. M. Anscombe and G. H. von Wright (eds.), G. E. M. Anscombe (trans.), Oxford: Blackwell.
    8. Wittgenstein, Ludwig 1953 Philosophical Investigations, G. E. M. Anscombe and R. Rhees (eds.), G. E. M. Anscombe (trans.), Oxford: Blackwell.
    9. Wittgenstein, Ludwig 1961 Tractatus Logico-Philosophicus, D. F. Pears and B. F. McGuinness (trans.), London: Routledge & Kegan Paul.
    These terms are taken from Hacker 1990, 318-326.
    Stefan Brandt. Date: XML TEI markup by WAB (Rune J. Falch, Heinz W. Krüger, Alois Pichler, Deirdre C.P. Smith) 2011-13. Last change 18.12.2013.
    This page is made available under the Creative Commons General Public License "Attribution, Non-Commercial, Share-Alike", version 3.0 (CCPL BY-NC-SA)


    • There are currently no refbacks.