Wittgenstein’s private language argument is notorious for its elusiveness. In fact, there is little exaggeration in saying that the consensus among the commentators does not go far beyond the point that it amounts to a strand in Wittgenstein’s Philosophical Investigations which gains prominence aroundparagraph 243 and vanishes somewhere around paragraph 300. It is not the ambition of this essay to argue in favor of an answer to the question about how Wittgenstein’s remarks there are best interpreted. Instead, I will investigate the prospects of designing a convincing argument that may be thought of as inspired by Philosophical Investigations. I will do it by first explaining a version of the private language argument and then discussing some criticism leveled against it. 

The version which I would like to consider takes the following overall shape:

1.Sensations are private objects. (Supposition for reductio.)

2. Predicates applying to (so conceived) sensations can be defined only by means of inward ostensive definition. (Premise.)

3. If a sensation predicate is introduced by means of an inward ostensive definition, then the speaker of a (private) sensation language cannot draw a line between a correct and merely seemingly correct application of the predicate. (Premise.)

4. The speaker of a (private) sensation language cannot draw a line between correct and merely seemingly correct application of a predicate. (From 2 and 3.)

5. If the speaker cannot distinguish between correct and merely seemingly correct applications of a predicate, then statements containing such a predicate cannot play a role in communication. (Premise supported by considerations about language) 

6. A (private) sensation language cannot play a role in communication (between the past and the present self of the speaker), i.e. such a language is not possible. (From 4 and 5)

7. There is a language whose predicates apply to sensations and which is successfully used in communication. (Premise)

8. Sensations are not private objects. ((1) leads to (6), which conflicts with facts as described by (7) and thus is false.)

Since the validity of the argument does not appear to be open to criticism, its fate is hostage to the status of the premises. It is not difficult to notice, however, that some of them amount to intricate philosophical claims which will obstinately resist evaluation if not first subjected to explication. Consequently, the only way of arriving at a well-informed verdict on the argument leads via spelling out its premises in some detail. 

The notion of a private object, invoked in the supposition for reductio, (1), and, as a result, also in the conclusion, (8), is critical not only to the argument’s soundness, but also to the question of what, strictly speaking, it purports to show. Two interpretations of the notion seem to be and relevant here. Privacy may be understood as a de facto lack of epistemic access to some objects on the part of anyone but one person. Before the arrival of Man Friday Robinson Crusoe was surrounded by objects which were private in this sense. On the other hand, it may be taken to describe more a radical scenario – a lack of epistemic access on the part of the others which is not accidental but necessary. It is the latter conception of private object which is employed in the argument. For a sensation to be a private object is for its nature to lie beyond a possible epistemic reach of everyone except the person who possesses it.[1] Put slightly differently, sensations are conceived as neither publicly accessible themselves nor as revealed by their outer manifestations. Although sensations are causally responsible for the manifestations (which, being states of the body or pieces of behavior, are publicly accessible), the connection here is only contingent. The contingency is twofold. Firstly, a sensation may occur without being manifested in any manner at all. Secondly, even if it is manifested, the manifestation does not reveal its essential features. (Of course, it is not plausible to claim that all properties of a sensation are not revealed by its manifestation. Pretending aside, the property of causing the manifestation seems to be always revealed.) The private language argument presupposes that a speaker attempts to design a language by focusing on the features of his sensations which are not manifested. (For the sake of simplicity it may be assumed that the speaker is ‘cut off’ from any manifestations of his sensations.) 

In the background of the second step of the argument lies the question concerning conditions which must be met if sensations (conceived as private objects) are to belong to the field of reference of a language. At the very least such a language would require predicates denoting them. Their absence would mean that it lacks resources for making basic subject-predicate statements. On the contrary, apart from addressing this deficiency, their availability would provide a springboard to introducing singular terms which would escape the limitations of demonstrative expressions.If predicate ‘-is a pain’ is furnished with a meaning, definite description ‘the first pain which I had yesterday’ is able to perform the work of a denoting phrase. Thus the importance of the question of how the predicates of the private language are defined.

As it is well known, primitive predicates of a public language are assigned meanings by means of ostensive definitions. A few objects with a feature which is to correspond to a predicate are indicated in hope that the student will associate the property with utterances of the predicate. It seems that the only route to conferring meaning on sensation predicates leads through a mental (inward) counterpart of ostensive definition. A concentration of attention on a sensation would be analogous to pointing to a publicly identifiable object by means of a demonstrative accompanied by a gesture. However, in contrast with ostensive definitions of predicates of a public language, their inward cousin would not require pointing to a number of objects. An ostensive definition of a predicate of a public language requires the introduction of a number of objects precisely because the person who learns the meaning of a term is different from his teacher. External objects have a considerable number of properties and a presentation of only one of them is not sufficient to make it clear to the student which of its properties is the intended to correspond to the term. On the assumption that the teacher consistently uses the predicate to stand for the same feature the student needs at least a few objects of which the predicate is true in order to realize which of their features is to be captured by it. But when an inward ostensive definition is invoked, it is immediately transparent to the aspiring speaker of a private language which property of a sensation he wants a predicate to stand for.[2] Thus it seems that, in a sense, the task of introducing a predicate is less arduous here than in the case of a public language.[3]

The third step is the crux of the version of the private language argument under consideration. It rests on the observation that to apply successfully a sensation predicate introduced by means of an inward ostensive definition the speaker must recognize an occurrent sensation as being of the same type as the sensation which served to define the predicate. This is to say that he must be right, or correct, in judging that the present sensation (which is, to reiterate, assumed to be a private object) is of the same kind as the past sensation in the presence of which he introduced the predicate. In general, for a subject to in position to make a right judgment he must have a criterion of correctness. However, contends the premise, no such criterion is available in the case of the putative private objects. To make the judgment that the sensation I have now is of the same kind as the sensation I decided to call ‘a pain’ yesterday is to say that I have a sensation of a certain kind now and that the past sensation which I called ‘a pain’ was of the same kind. If I am to judge this correctly I must be aware of at least two things. Firstly, I must know what kind of sensation I am having now. Secondly, I must know what kind of sensation I decided to call ‘a pain’ yesterday. In the framework of the conception of privacy which the private language argument presupposes the former requirement does not engender a difficulty. By contrast, the latter requirement is taken to speak strongly against the possibility of a successful application of sensation predicates whose meaning was allegedly established by means of inward ostensive definitions. 

On the assumption that sensations are essentially private and transient objects the only route to obtaining knowledge about these of them which have waned leads via memory. Yet there is no criterion by which the correctness of the recollections of such sensations can be tested. The only way of being right about some fact concerning a past sensation, for example about its being called ‘a pain’, is to rely on what seems to be right, that is, to accept the deliverances of memory at face value. In other words, if a sensation predicate is to be correctly applied (which presupposes that it is used with the same meaning) a speaker must remember that the predicate was introduced in the presence of a sensation of the same kind as the one currently occurring. Of course, the failure of one’s memory means that one is not applying a predicate correctly. But this implies that the application of a sensation predicate introduced by means of a private ostensive definition is in quite a desperate predicament. It has to rest on a recollection the correctness of which cannot be verified. One using such a predicate must take what seems to be right (the predicate-sensation connection as portraited by memory) to be what is right (the actually established connection). The practice of using a sensation predicate defined by inward ostension forces the speaker to relay on unverifiable recollections and thus he must regard what, in their light, seems to have been the case as what really was the case. As a result, from the subject’s point of view the distinction between being right and seeming right about applications of a sensation predicate disappears.[4]

Step (5) draws a crucial consequence from the claim, (4), that, as far as the predicates of a private language are concerned, there is no distinction between their correct and only seemingly correct application. If sensations are private objects, the possibility of a failure of memory implies that the speaker has no means of determining whether in applying a predicate to a current sensation he uses it to denote the same kind of sensation as at the time when the predicate was first introduced. He has no verifiable access to the episode in his past which determined the meaning of the predicate and thus, in a sense, does not know the meaning. As a result, he can neither make meaningful statements nor, a fortiori, understand his past statements concerning sensations.[5] This immediately implies (step 6) that the predicates of a private language cannot be employed in conveying information about the sensations which he experienced in the past. Furthermore, since the conveyance of information is one of the fundamental functions of language, this failure inflicts a decisive blow to the proposition that a sensation language is possible. As it is easy to notice, this consequence is, to put it mildly, at odds with our actual linguistic practice: we commonly and, it seems, successfully communicate about sensations (step 7). Apart from (2) and (3), which are not controversial in eyes of the supporter of the argument, unacceptable (7) rests on (1). Hence the conclusion, (8), that sensations are not private objects.

It is important to appreciate the limited nature of this conclusion. Strictly speaking, it does not deny that there might be some entities which are private or that they may be known. What it claims is that the sensations to which we commonly refer are not such private objects. This is to say, if the private language argument is successful, it precludes only some kind of knowledge of the alleged private objects, namely knowledge presupposing the meaningfulness of sensation predicates. Yet, there may be kinds of knowledge that do not require the employment of a predicate. Imagine a solitary speaker, say Robinson Crusoe, confronted with a strange object the color of which is quite distinct from any color he has ever experienced. After running in his head through the colors predicates he knows, Crusoe concludes that he does not have a predicate to describe the puzzling thing. But it does not seem a cavalier thing to say that he knows that This has that, i.e. the object in question has the feature which he identified.[6] Thus, the adherent of the view that some mental items are private objects may grant the private language argument. In the light of the argument he will not be justified in holding that a private mental item has a feature which he encountered before. None the less, the lesser claim that the sensation has the feature which strikes him at the moment may be well founded. Of course, the private language argument implies that he will neither be able to preserve this piece of knowledge for his own future purposes nor to communicate it to others.

All this does not mean that, if successful, the private language argument does not undermine the contention that mental items are private. Although the view that there might be some private mental items is, strictly speaking, not shown to be false, its attractiveness is considerably weakened. If there were private objects, then they would not be the denotation of such words as ‘pain’ or ‘the feeling of relief.’ Furthermore, they would not be able to feature as a subject matter of one’s soliloquy, let alone of public communication. Their status would, therefore, be rather shadowy. Indeed, they would qualify for the role of things which, as Wittgenstein said in a different context, cannot be spoken of and thus must be consigned to silence. 

The contention that an aspiring speaker of a private language would suffer from a lack of resources to distinguish between correct and incorrect applications of sensation words is at the heart of the argument sketched above. It is, in turn, the inevitability of the appeal to memory which is deemed to be responsible for this predicament. If one misremembers how a sensation term was originally defined and applies it to a sensation of a different kind, he will not be able to realize his mistake. The assignment of such a pivotal role to the possibility of an unverifiable failure of memory has been often considered as the weakest point of the private language argument. The accusation has been that the argument would speak not only against the idea of a language pertaining to private objects but also against a solitary language, i.e. a language used by only one person to describe objects which are, in principle although not in fact, publicly accessible. Like a speaker of a private language, the solitary speaker will have to resort to his memory when he applies a term. Although his words will denote external physical objects and processes, he will have to remember how he defined them if he wants to understand his past statements or make new ones. And in the absence of other speakers who would correct him,[7] he will not have any check on his memory: an application of a term which will seem right to him will be right. The solitary language would share the dismal fate of the private language. But then, the criticism continues, the private language argument would not show that the very conception of sensations as private objects engenders a serious philosophical worry. It ‘would at most tend to show that the idea of a language of any kind used by only one person was an absurdity: and this conclusion would have no special immediate relevance to the case of sensation.’[8]

It seems that there is some room for doubt about the force of this criticism. Even if it is true that the private language argument speaks only against a language ‘used by only one person’, its success would have serious ramifications for the question of the nature of sensations. It would still support the claim that the conception of sensations as private objects implies that they are beyond a linguistic reach even of the person who possesses them. In other words, although it would not be limited to ‘the case of sensation’, its conclusion would remain unchanged as far as sensations are concerned. This may be why critics of the argument proceed to press further charges.

Maintaining that the private language argument implies the claim that both a private sensation language and a solitary language are not viable, they contend that this claim is false as far as a solitary language is concerned. Although memory would have to provide the ultimate test of the recognition of some objects as referents of the terms of a solitary language, this would not make such a language impossible. Of course, there may be certain psychological barriers in virtue of which human beings would fail to speak such a language (for example, their brain may need a linguistic input at an early stage of development), but the inadequacy of the memory with which an average human is endowed is not one of them. Such a memory is sufficiently reliable for establishing a successful linguistic practice.[9] But if a solitary language is possible and the private language argument implies that it is not, then the argument, and in particular its crucial step, i.e.(3), is erroneous. 

Interestingly enough, some of the critics of the argument, for instance Ayer, contend that although the presence of other speakers would provide a ‘check’ on a person’s use of words, a solitary speaker’s supposed failure of memory would not have any practical consequences.[10] On the other hand it has been maintained that the possibility of a solitary language has been declared much too hastily.[11]I will investigate these claims in turn and then consider how the status of the solitary language hypothesis affects the claim that a private language is impossible. 

Following Dorit Bar-on, imagine a hitherto non-linguistic Super-Crusoe who realizes that some berries are rather nutritious whereas their immediate and almost indistinguishable neighbors cause severe skin-rashes. Concerned with avoiding dermatological problems in the future, he marks the bush which produces good berries with a blue string and the other bush with a red string. It is natural to expect that, in time, Super-Crusoe ‘could come to possess a mini-science of his environment encoded in a notational system which serves as an instrument for avoiding trouble and better control of his surroundings.’[12] The plausibility of such an expectation seems to speak against Ayer’s contention that a failure of Super-Crusoe’s memory would go unnoticed. He will suffer physiological consequences if he fails to decode the statements of his sign language. One might think that this shows that he will have a way of knowing that he interprets his language incorrectly. Put in more general terms, the implication would be that a speaker of a solitary language would be in possession of standards for evaluating the correctness of his linguistic performance. Doesn’t it mean that a solitary language is not only possible, but also that a solitary speaker is much better equipped than his private cousin (and thus that the Strawson-Ayer criticism of the private language argument has been significantly weakened)? Maybe a little surprisingly, some of the supporters of the argument reply ‘No.’ They maintain that even a solitary language, let alone the private one, faces insurmountable difficulties. 

This claim seems to be to a large extent inspired by a thesis which lies at the heart of a Gricean account of the emergence of linguistic meaning. The account contends that if an expression is to possess a linguistic meaning, the speaker must have a ‘reflexive’ communicative intention. This is to say that in addition to entertaining an intention to induce a certain belief in the audience he must intend that the belief is induced in virtue of the audience’s recognition of his ‘intention to induce in them that belief.’ When applied to the solitary efforts of Super-Crusoe’s, the Gricean account implies that to have the linguistic meaning That is a bad bush ‘a sign must offer an audience intention-dependent evidence that that is a bad bush.’ More precisely, Super-Crusoe 

‘must intend to induce in his future self the belief that that’s a (the) bad bush, and, further, to remember that very intention. To form that intention, it is not enough for him to be able to think of his future behavior, and intend to affect it in a certain way. He must also be able to think of his future recognition and memory of a present belief and intention.’ 

Yet, goes the argument against the possibility of a solitary language, even if Super-Crusoe can form such complex intentions, his solitariness will render the ‘reflexive’ intention idle.The ‘reflexive’ intention to remember the intention with which he attaches a red string to a bush has no point if the latter intention is the intention to remind his future self about a belief he currently entertains. This is because if he thinks that the red string can ‘trigger memory of the [first-order] intention, why shouldn’t he think that it can trigger, more directly, memory of the bad bush belief?’ In general, if a sign is used as a reminder, the (first-order) intention to remind about a belief is remembered only if the belief itself is remembered. Thus, it does not serve any purpose to intend to remember the (first-order) intention to induce the belief, which means that if one wants that his future self remembers his current belief he does not have to intend that it is recognized via his first-order intention.[13]

But do we have to assume that Super-Crusoe remembers that it was himself who attached a red string to the poisonous bush? Even if he does not remember it upon encountering a bush marked with a red string, he may still be able to learn that his past self had the intention of inducing in him the belief that the bush is poisonous. Moreover, it is not entirely unreasonable to think that to (re)acquire this belief he will have to recognize the intention with which he attached the string. The train of thought passing trough his mind may run along the following lines. He will first notice that there is a red string attached to a bush and conclude that it was put there by someone. He will also know that in general he uses red strings to mark some bushes in order to induce in his future self the belief that they are poisonous. He will be fully aware, however, that he does not have any recollection of attaching a red string to the bush in front of him. Yet won’t he begin to circumvent this lapse of memory by making the conjecture that, in the absence of any evidence of other sapient creatures on his inland, it was himself who attached the string? Furthermore, he will have no reason to think that his intention in doing so wasn’t the usual one of inducing the belief about the poisonous character of a bush in his future self. On the basis of the recognition of this intention he will conclude that the bush is indeed poisonous. Thus to recover his former belief he will appeal to the first order-intention. And since the recognition of the first order intention is an indispensable step in reacquiring the former belief, he will have a reason for having the intention (when he attaches the strings) that his first order intention is recognized.

One step of my reconstruction of Super-Cruzoe’s considerations may invite serious doubts. Is it really the case that when he attaches red strings to bushes he does it with the intention to induce in his future self the belief that they are poisonous? Doesn’t he rather intend that they serve as a recollection of his current belief, i.e. that they cause him to remember the belief? Although it is true that Super-Crusoe will often, if not usually, be reminded by a red string of his past belief about the poisonous nature of a bush, it does not follow that he has an intention to be so reminded when he attaches red strings to some plants. While talking about Super-Crusoe we indulge in considering the possibility of a certain scenario of developing a solitary language. The point of this exercise is to show that his actions and beliefs may be shown to lead to acquiring a solitary language in a fairly natural way. And to assign to him the general intention of inducing a certain belief in his future self is not psychologically extravagant. Indeed, in light of the limitations of our memory such an intention seems to be more psychologically realistic than the intention to remember a presently entertained belief. After all, we are well aware that we have plenty of beliefs which we will not remember. But we also know that we will be able to infer that we had them. When we make notes, write entries in our diaries, etc. we allow, if not presuppose, that we will not remember the beliefs which we then entertain. We know pretty well that sometimes we will not even remember that we wrote it. This means that we write it the intention to induce the currently entertained belief in our future self rather than with the intention to remember this belief at some point in future. But if we commonly harbor such an intention, then why think that it is incoherent or even implausible to assign it Super-Crusoe?

The failure of the contention that a solitary language is impossible may not automatically translate into support for the claim that a private language is possible. As noted above, it seems that, unlike the speaker of a private language, Super-Crusoe would be in possession of resources which would allow him to distinguish between correct and incorrect uses of the expressions of his solitary language. In other words, it appears that the speaker of a solitary language would, after all, have a criterion for separating the uses of its expressions which are right from those which only seem so. 

However, it has been claimed that the aspiring speaker of a private language is not in such a desperate position as the Wittgensteinian enemies of a private language suggest. Ross Harrison has argued that even if sensations are not linked to behavior, they may still be governed by regularities. For instance, sensation S may be always succeeded by sensation T and preceded by sensation R. In such a case there would be a difference between correct and merely seemingly correct applications of sensation predicates.The subject would know that he applied, say, predicate S correctly if predicate R was applied to the sensation immediately preceding it and predicate T to the sensation immediately following it.[14]

One may doubt whether this criticism will persuade many supporters of the private language argument. To begin with, the question arises how the speaker is to discover lawful regularities among sensations. It seems that to do it he must be able to record their past occurrences, which, in turn, requires a language. Moreover, one may think that memory will still play a crucial role here. Imagine that the speaker is just describing a sensation as T, i.e. the third element in the RST series. According to Harrison’s proposal, to be correct in his application of the T predicate, apart from knowing that there is an RTS regularity, he must remember that T was preceded by R and S. But, of course, he has no way of verifying that his memory is correct. So, again, he appears to be unable to draw a well-justified line between correct and incorrect applications of a sensation predicate. 

It is difficult to resist the impression that the private language argument fends off various kinds of criticism quite well. To recapitulate, the Strawson-Ayer contention that the possibility of a solitary language implies the possibility of a private language seems to be called into question when it is granted that the speaker of a solitary language has some resources to distinguish between correct and incorrect uses of its expressions. This is especially apparent in light of the failure of the claim that a solitary language is impossible for Gricean reasons. On the other hand, Harrison’s attempt to show that the speaker of a private language can appeal to regularities in order to achieve a criterion of linguistic correctness is not likely to succeed. Yet all this does not mean that the opponents of the private language should abandon hope. While arguing for the possibility of a solitary language Strawson asks whether we ‘in fact find ourselves misremembering the use of very simple words of our language, and having to correct ourselves by attention to others’ use.’[15] This may inspire one to support the claim that we should not worry about memory failures in the case of sensations (conceived as private objects) by pointing out that our memory works quite well in areas where its deliverances can be checked. Of course, a natural response to this claim will contend that the reliability of one’s memory may vary quite significantly from area to area. For example, a person may have an excellent visual memory but a very poor acoustic one. Why should it be the case that the reliability of our memory in ‘publicly verifiable areas’ does not say anything about its quality in the ‘private sphere?’ The supporter of private language may have some difficulties with designing a persuasive response to this worry, but, again, his position does not have to be hopeless. For instance, he may receive some support from the dispositional theory of secondary qualities.According to the theory we perceive colors by being aware of sensations of color. Thus, the fact that we do not often misremember colors of public objects may be taken to indicate that we do not misremember what color sensation we were aware of. Of course, the dispositional theory of color is embroiled in an ongoing controversy the mention of which makes me think that I should stop here. 

Notes