Private Language Problem
PRIVATE LANGUAGE PROBLEM
The private language problem is essentially the question of whether or not a language as a system of symbols that are means of thinking is, of necessity, a language as a system of symbols that are means of communication. Defining "private language" as language (in the sense of means of thinking) which in principle the speaker alone can understand (so that it cannot serve as a means of communication), our question is roughly equivalent to: "Is a private language possible?" Many philosophers, following Ludwig Wittgenstein, have made the claim (here called the private language thesis, abbreviated PLT) that private languages are impossible. Armed with it, they have argued against solipsism, phenomenalism, the analogical or empirical view of one's knowledge of other minds, and against mind-body dualism. Some of them have gone on to argue for certain versions of philosophical behaviorism as well as for the view that the meaning of a word consists of its use or employment in a social practice and not in its referring to something or its designating a kind of entity.
Thus, the PLT has been a central principle in the cluster of Wittgensteinian doctrines. It is not clear, however, that exactly the same thesis figures in all the arguments in question, since the idea of a private language varies in different contexts. There is, therefore, a multiple problem: First, to differentiate the several propositions which pass as the PLT by clarifying the sense of "private language" being used; second, to determine which ones are true; and third, to explain why they are supposed to be intimately related. These problems differ from the question, debated around 1930, of whether or not it is possible to start with a private language about one's sensations or "raw" feelings and arrive at the intersubjective and communicable language of science. (On this question, see Rudolf Carnap, "Psychology in Physical Language," and J. R. Weinberg, An Examination of Logical Positivism.)
The Sense of "Impossible"
In all the interpretations of the PLT, the word impossible is understood in a strong sense that is not easy to characterize precisely. Some philosophers speak of "logical impossibility," but they do not necessarily mean that private languages are impossible in the sense that unbounded triangular figures are impossible. The expression "unbounded triangular figure" reduces to the formal self-contradiction "unbounded figures bounded by three lines" by means of a substitution allowed by the definition of "triangle." But few philosophers would suggest that there is a similarly ready definition of "language" by means of which we can produce a formal self-contradiction "private so-and-so which is not private." The impossibility at issue is like (1) the impossibility of unextended red things (that is, the impossibility that something be red and yet lack width or length), or (2) the impossibility of a cube with fewer than eight edges. These do not lead straightforwardly to formal contradictions, since there are no definitions for all the terms involved; they depend on implication relations that constitute the concepts involved in their statement. In the last analysis, the persistent rejection of (1) and (2) evidences the failure to understand the meanings of all the words involved, that is, the lack of some of the relevant concepts. But (1) is unprovable and obvious, and (2) only needs a trivial argument, while the PLT (if true) requires careful reasoning. We shall speak of conceptual impossibility to refer to any formal self-contradiction, to any impossibility which entails a formal self-contradiction, and to any a priori impossibility such as that found in the above examples (1) and (2).
The Private Language Thesis
The most important propositions often discussed as the PLT, each embodying a different idea of private language, are the following:
PLT*: It is impossible for a man to use a word with a meaning that nobody else could, even in principle, understand.
PLT-1: It is impossible for a man to use words that refer to private objects, that is, objects that nobody else could—even in principle—know. (For subtheses arising out of the ambiguities of "know," see H.-N. Castañeda, "The Private-Language Argument.")
PLT-2: It is impossible for a man who has always lived in isolation to possess a language, even if his sounds are understandable by another person.
Here the expressions "could not in principle" and "impossible" are meant to express conceptual impossibility. PLT* allows that a man may use words with meanings that nobody else in fact understands, provided that they are understandable to other people in the appropriate circumstances. PLT-1 allows that a man may refer to objects that, in fact, he alone knows, but again others must be capable of knowing them in the appropriate circumstances. PLT-2 allows that a man, like Robinson Crusoe, keeps possession of a language he learned previously while living in a community of speakers.
Many philosophers assume that it is conceptually impossible for two persons to share one and the same immediate sensation. Many also hold that, in a strict sense of "know," others do not really know whether one has a certain immediate sensation or not, precisely because they cannot share it. On these assumptions, a language about one's own immediate sensations would be a language of the sort that PLT-1 claims to be impossible. Indeed, such a language is customarily regarded as the would-be prototype of private language.
In general, on the assumption that (direct) knowledge of the referent of a word is required for understanding the meaning of the word in question, PLT* entails PLT-1. On this assumption, a language about one's own immediate sensations is also private in the way that PLT* claims to be impossible.
PLT-1 does not entail PLT*. A word might have a meaning understandable to only one person because the word itself is a private object in the sense of PLT-1, even though everybody may be acquainted with the physical objects it refers to. For example, the words of a person's language might all be mental images of German written words, so that all his thinking would be a sort of mental reading of German. In this case, the referents of the words would be public, but the words themselves would be private and hence unintelligible to others.
PLT-2 neither entails nor is entailed by PLT-1. If PLT-2 is true, then if on the previous assumptions about sensations, one's language about one's own sensations is private in the sense of PLT-1, then one could still, in principle, invent such a language. Conversely, the truth of PLT-1 does not by itself make it impossible for an isolated person to invent a language about physical objects. Similarly, PLT-2 neither entails nor is entailed by PLT*.
applications of private language theses
The important claims made with the help of the PLT do require other assumptions, which in their turn play roles, as we shall see, in the defense of the PLT itself. The most natural and pervasive of these assumptions is the following:
(A ) In the sense of "thinking" in which one can both have a false (or true) thought and draw inferences from what one thinks, it is conceptually impossible to think without possessing a language that is a means of thinking.
From this assumption and PLT-2, one can conclude that the fact that one thinks, guarantees the existence of other persons, namely, one's fellow speakers of the same language. Thus, the solipsist who merely asserted that it is possible that he alone exists at the time he is thinking would be contradicting himself (an argument of this sort can be constructed with premises suggested by Rush Rhees in "Can There Be a Private Language?"). Of course, many philosophers have serious objections to (A ).
The existence of hallucinations, illusions, and visual perspective leads many philosophers to characterize every case of perception in terms of our apprehension of sense data or immediate impressions. Some have proceeded to espouse a phenomenalistic program of "logical reconstruction" of physical objects and minds as systems of sense data; others, however, have subscribed to some form of realism, that is, the complete irreducibility of physical objects and minds to sense data. But all of them have recently been criticized on the ground that the language of sense data is private in either the sense of PLT* or the sense of PLT-1. Here, in addition to (A ), the critics need the following assumption:
(B ) If it is conceptually impossible that there be a language about entities of a sort T, then there are no entities of sort T.
Again, some philosophers would claim against (B ) that if PLT* or PLT-1 is true, then sense data or the given in experience are simply ineffable.
Many philosophers have subscribed to some form or other of a principle of verification, for example:
(C ) It is conceptually impossible to understand a sentence without knowing what state of affairs would verify the statement made with it.
Assumption (C ) leads to the view that language about states of consciousness is private, if we add to it and (A ) and (B ) the following principle:
(P ) Only the person himself can verify conclusively and directly that he has certain experiences.
On this view, for instance, when someone else speaking about me says, "He is in pain," he cannot understand or mean exactly the same thing that I understand and mean when, of myself, I say, "I am in pain." But if PLT* is accepted, one is involved in a contradiction. Here many philosophers have given up (P ), and in order to guarantee that everybody else can know what somebody is feeling or thinking, some philosophers have espoused some form of behaviorism, that is, a view according to which every description of a person's experiences or mental states is really shorthand for (synonymous with) a description of his bodily movements, his relations to other bodies, and his abilities to perform further movement. This is often supplemented with the supposition that first-person utterances like "I have a headache" do not make statements of direct knowledge but are, rather, learned responses, analogous to the natural responses of moaning, crying, and so on, which are said to constitute the person's ache. As is to be expected, other philosophers have preferred to keep (P ) and reject one or more of the other premises, in particular (C ) or PLT*. (See Castañeda, op. cit., Part B, for a discussion of the privacy of experiences.)
the main arguments for the plt
There are many arguments seeking to prove that being private makes it impossible for a language to have a property required for the existence of a language. Most of the arguments depend on the following assumption:
(D ) A language is a system of rules, and to speak or write a language is to follow rules.
On this assumption, it suffices to establish the PLT to show that a man (say, Privatus) cannot be following rules when he is using a private language (to be called Privatish). This is, in fact, what a series of arguments suggested by Wittgenstein purports to do. The gist of the argument is as follows: A rule is, by its very nature, the sort of thing that can be misapplied (or disobeyed), but Privatus cannot misapply the rules of Privatish; hence, when speaking Privatish, Privatus is not following rules. The specific arguments are meant to support the crucial premise:
(1) Privatus cannot misapply the rules of Privatish.
A fair objection to (1) is that Privatus can certainly make slips; he may call something of kind A "B," whatever "A " and "B " may mean in Privatish. Slips of the tongue are precisely ways in which one violates the rules (if there are such) of natural languages. For instance, if there are rules of English governing the application of color words to physical objects; whenever one commits a slip of the tongue and calls a red object "blue," then one misapplies either a rule governing the use of "red" or one governing the use of "blue."
This reply to (1) is often met by several rejoinders. The first claims both that a slip counts as a misapplication of a linguistic rule only if there is a way in which the speaker can in principle detect and correct his slip and that Privatus cannot detect or correct his slips. This rejoinder, however, changes the issue, since premise (1) says nothing about verifying the existence of a misapplication of a rule. Nevertheless, the rejoinder has a point, for if to use words is to apply rules, then one must at least sometimes be able both to know of one's misapplications of the rules for the use of one's words and to know how to make the appropriate corrections. The question of whether or not Privatish allows this is discussed below under premise (2).
The second rejoinder is that to obey a rule is a custom (use, institution), but Privatus's actions cannot constitute a custom (see Ludwig Wittgenstein, Philosophical Investigations, Sec. 199). This rejoinder would establish PLT-2 but not PLT* or PLT-1. For it may be a custom in a tribe that people use words which they alone understand in the ways required by PLT* or PLT-1. But as an argument for PLT-2 the rejoinder is by itself question-begging. It must be supported by an argument which shows that obeying a rule is indeed a custom.
The third rejoinder is that Privatus's slips do not count as violations of the rules of Privatish because we cannot be corrected or taught by others what is the correct thing to say (see Wittgenstein, op. cit., Sec. 378, and Norman Malcolm, "Discussion of Wittgenstein's Philosophical Investigations," pp. 536f.). If the "cannot" here is taken to mean conceptual impossibility, the rejoinder does not apply to PLT-2. If it is taken in a weaker sense, that is, a sense in which a person may be in the position of being in fact corrected by other persons, then the rejoinder supports PLT-2, but it would not allow that there be just one language-user in the universe. Besides, it is not clear that it would allow that Antonia Udina, for example, used language when, as we normally say, he spoke Dalmatian as the last speaker of Dalmatian. Although a person who uses words must be capable of self-correction, it is not immediately obvious that a person's sounds cannot count as utterances of words if nobody else can (in some sense) correct him. The need for others' possible corrections has to be established by an argument. Thus, we are again thrown back to the other lines of reasoning.
The fourth rejoinder is that Privatus's slips do not count because another person, by noting Privatus's behavior and circumstances, cannot discover that his use of the word is correct or incorrect (adopting Malcolm, op. cit., p. 537). This rejoinder also leaves PLT-2 unsupported if "cannot" is understood as expressing conceptual impossibility. While it must be conceptually possible for Privatus to know whether his uses of language are correct or incorrect, it is not at all clear that it must be possible for others to know this fact. The principle that it must be possible for others to know whether his uses of language are correct or incorrect requires an independent argument to support it. However, the present rejoinder has a point. It reminds us that if there is no way at all of telling, for any word of Privatish, whether or not Privatus used it correctly (however coherent the concept of a private language is), it would be a completely gratuitous hypothesis that Privatus spoke a private language. Although our topic here is only the conceptual possibility of private language, we should note that the claim that somebody's entire language is of the type described in PLT* is certainly gratuitous. Yet the claim that someone has a mixed language, part of which is private in the sense of PLT*, does not seem gratuitous.
The fifth rejoinder dismisses mere slips on the ground that they show at most a breakdown of a linguistic habit. The rejoinder asks us to consider the case of Privatus trying deliberately to apply a rule of Privatish and failing to comply with it. The rejoinder claims that, for Privatish, "thinking one was obeying a rule would be the same thing as obeying it," but "to think one is obeying a rule is not to obey a rule. Hence it is not possible to obey a rule 'privately'" (Wittgenstein, op. cit., Sec. 202). This rejoinder does not require that every utterance of a word be a case of deliberately attempting to obey the corresponding linguistic rule(s). Conjoined with assumption (A ), this view would lead to a vicious infinite regress. For then, in order to say something, one would have to be aware of the rules governing the words one intends to utter, and these rules in their turn would be formulated in some words the rules governing which one would have to be aware of through some other words, and so on ad infinitum. Therefore, to use language is, of necessity, to use most of the words from habit, not in intended obedience of the linguistic rules. The rejoinder cannot even demand that Privatus sometimes be aware of the rules of Privatish: A being might speak a language without ever rising to the level of formulating any of his rules. But if, by assumption (D ), languages are made up of rules, then if it were conceptually impossible for Privatus to be at least sometimes aware of the rules of Privatish, Privatish would be a very defective language indeed, incapable of discharging the philosophical duties that private languages are alleged to discharge. Thus, the rejoinder is right in urging that
(a ) For every rule R of a language L and every speaker S of L, it is conceptually possible that sometimes R applies to S 's situation while S thinks that he is obeying R without S 's actually obeying R.
Presumably, a rule of language is here of the form "If x is ϕ, you may (must) call it '…,'" but the meaning of "call" is difficult. In one normal sense of "call," slips of the tongue are, again, ways in which (a ) is true. Clearly, a person may think that he is calling a thing "red" in deliberate compliance with the English rule for "red," without realizing that he actually called it "blue" because he is deaf or because he simply did not hear what he said. In the same sense of "call," (a ) can be true because the speaker deliberately calls a red thing "blue," if he thinks that the rule in question allows (or prescribes) his calling it "blue." In particular, suppose that the rule R allowing (or prescribing) that one call a thing "red" is the rule Gaskon typed yesterday and that today, confusedly, Gaskon thinks that the rule he typed yesterday allows (or prescribes) that a certain thing be called "blue," and he calls the thing in question "blue," thinking that he is complying with the rule. Here, in spite of his deliberately calling a certain thing "blue," Gaskon's use of "blue" and the rule he thinks he is complying with both satisfy (a ). Both ways of satisfying (a ) are open to Privatus. It might be argued that Privatus's deliberately calling one of his private objects "A " instead of "B " has no point or "function" (see Wittgenstein, op. cit., Sec. 260), since he is not talking to others. This is, however, false. Privatus may very well play word games involving miscallings of things. But more importantly, whether or not there is a point in Privatus's flouting of the rules of Privatish has nothing to do with the issue about the possibility of private language.
The rejoinder often uses a stronger sense of "call." In this sense, by a natural development of assumption (A ), to think that something is, for example, red is to call it "red." This stronger sense appears in an argument given in support of PLT-1. As said above, language about one's own immediate sensations is often regarded as the paradigm of private language in the sense of PLT-1. Now, one knows incorrigibly that one's sensations have immediately sensible qualities. That is to say, if one believes that one has a pain (itch, tickling, feeling of discomfort), then one knows that one has a pain (itch, tickling, feeling of discomfort). So it is impossible to have no pain while one thinks that one has a pain. Thus, if one thinks that one is obeying the rule of the form "If x is a pain, you may (must) call it 'pain,'" one surely thinks that one is in pain and the rule cannot fail to apply. Similarly, since one also has incorrigible knowledge of the absence of one's immediate sensations, if the objects that Privatus can think about in Privatish are only his immediate sensations, then when he thinks that a rule of Privatish does not apply, the rule does not, in fact, apply. But if "call" is taken in its normal sense, neither of these two features of the rules of Privatish implies that Privatus cannot think that he is obeying a rule (which then applies) without actually obeying it, since slips and deliberate miscallings are still available as violations of the rule. However, if "call" is taken in the strong sense (in which thinking can be calling), then if Privatus thinks that he is obeying a rule of the form "If x is A, you may (must) call it 'A,'" he surely thinks that the rule applies, that is, he thinks that the object x is A ; if A is a sensible property of Privatus's immediate sensation x, then x is A, and Privatus is both calling x "A " and unavoidably obeying the rule. Thus, if Privatish is a private language about Privatus's immediate sensations and their sensible properties, then (a ) above and (b ) below are both false:
(b ) For every rule R of a language L and every speaker S of L, it is possible that sometimes S thinks that he is obeying R while he is not.
Since (a ) is true, Privatish is not a private language.
This argument does not by itself support PLT-2; it may or may not support PLT*, depending on how one interprets the phrase "knowing the meaning of a word."
There is, however, a difficulty with the above argument. Consider the rule of English: "If x is a cat, you may (must) call x 'cat'; that is, you may (must) think that x is a cat." This rule differs from the above rule for the Privatish word "A " in that thinking that one is obeying the rule for "cat" does not imply that the rule for "cat" applies to the situation in question. For to think that one is obeying the latter rule implies that one thinks that it applies, and this implies that one thinks that some object x is a cat. But surely one can be mistaken about x 's being a cat. Yet the rule for "cat" also fails to satisfy condition (a ). Suppose that the rule applies; then the object x in question is a cat. And suppose that one thinks that one is obeying the rule; then it is true that one thinks that if x is a cat one may (must) think that x is a cat, and that one thinks that x is cat. Thus, one is in fact obeying the rule! Therefore, the strong sense of "call" included in the concept of language rule R makes (a ) an impossible condition.
Now, in the case in which a rule R does not apply to a man's situation, we are often reluctant to say that when such a man thinks that he is obeying R, he is not obeying R. But we could say this with no great distortion, and if we did, we could say that the above rule for the English word "cat" satisfies condition (b ). For in a situation in which an object x is not a cat and the rule does not apply, we may very well both misperceive or otherwise think that x is a cat and think that, in accordance with the rule, we may (must) think that x is a cat. Thus, if we raise (b ) as the crucial condition that linguistic rules must satisfy, then we can claim that PLT-1 is established in the sense that a pure language of sensations is impossible. But this answer is inconclusive. Besides the small amount of distortion involved, there is the fact that (b ) is not a general condition of rules. This is shown by the following rule which a man might give to his son: "If you think that you need to delay your action, think that 1 + 2 + 3 + ··· + 24 = 300." Since to think that one thinks that p entails that one thinks that p, if the boy thinks that the rule applies, he thinks he needs to delay his action, and the rule applies. If he thinks that he is obeying the rule, he thinks both that it applies and that 1 + 2 + 3 + ··· + 24 = 300; hence he thinks that 1 + 2 + ··· + 24 = 300; hence, the rule applies and he obeys it. Thus, to defend PLT-1 by means of (b ) requires an independent argument showing that rules of language must, in any case, comply with (b ), distorted as suggested.
Let us turn now to a subtler line of argument. Some defenders of the PLT do not argue for (1) but for
(2) Privatus cannot distinguish his correct uses of Privatish words from his incorrect uses.
Suppose, then, that Privatus is debating whether something is A or not. Suppose that Privatish is private in the sense of PLT-1. Here the defenders of the PLT adduce (a ) that Privatus lacks a criterion of correctness, that is, "something independent of his impression" that he is correctly using the Privatish rule governing the use of "A " by means of which he can "prove his impression correct" (Malcolm, op. cit., p. 532), and (b ) that his impression that he remembers what objects of kind A appeared like before is of no help, since memory "is not the highest court of appeal" (Wittgenstein, Sec. 56) and the "process [of checking memories] has got to produce a memory which is actually correct " (Sec. 265). Now, these points exaggerate Privatus's predicament. Privatus's private objects may be related among themselves by entailment, by coexistence, by similarities, by causal relationships, and so on. Privatus can resort to any of these to test whether he is, on the present occasion, using the term "A " correctly. For instance, in Privatish, "being A " may be logically equivalent to "being B and becoming C in the presence of another C." Indeed, Privatus may even employ paradigms. The very first object he calls "A " may very well be enduring, so that he can compare the next objects of kind A with it. The same applies to languages of the type mentioned in PLT-2. Furthermore, memory is the highest court of appeal when it comes to our knowledge of the past. True, we have records and other historical evidence, but all of this only provides inductive evidence, not a proof, and our inductions involve the acceptance of unchallenged memories.
Nevertheless, Privatus is not only in no position to question the correctness of all of his uses of words, but he also cannot prove that the uses he questions are correct unless he is allowed the ability to identify certain properties of objects without criteria and without challenging his memory. But exactly the same happens with the speakers of any language. In the case of terms like "red" and "straight," for instance, there is nothing at all to which an English speaker E can resort in order to "prove" that he has correctly called an object red or straight. His fellow speakers may all utter in unison, "Not red but blue." Yet this choral utterance is not a proof; the speakers may be lying, may all be victims of a hallucination, or may just be rehearsing a new song—or the whole proceedings may be just E 's hallucination. In any case, for E to accept the correction, he must correctly identify the words expressing it without the use of criteria and remember correctly the meanings of these words. A vicious infinite regress would ensue if E were required to have a proof that he both remembers this correctly and identifies the objects the words apply to.
Moreover, there is nothing to prove each corrector's use of words correct. Suppose, for example, that one corrector learned the meaning of "blue" with the help of object O and that he continuously stares at O during the preceding two minutes before correcting Privatus. He still must remember correctly that O has the same color it had two minutes before, that the color of O is called "blue," that the name of the color sounds "b–l–u–e," that the noise "red" uttered by E has the same meaning that makes red and blue incompatible, and so on. Thus, either somebody just identifies some words or objects correctly and remembers some qualities of objects and the meanings of some words correctly, or else nobody can be corrected by another speaker. In sum, demands (a ) and (b ) cannot be adduced against the possibility of a private language.
Often it is claimed that a private language cannot have logical words or syntactical rules, both of which are necessary for the existence of logical relationships. Clearly, if a private language is allowed no implications or entailments, it would certainly be no language. But if "private language" is meant in the sense of PLT* or PLT-1 or PLT-2, this contention appears to be false. Often this contention is defended on the ground that a really private language does not have words with meanings in common with the words of another language (Wittgenstein Sec. 261; Malcolm p. 537). Now, private language in this sense is impossible. A language is a system of words of which some refer to objects, some signify properties or relations, and some express logical connections; the words expressing logical connections must be capable of being understood by anybody else and must, therefore, be common to all languages. This is an important result. But it is not the same as PLT*, which requires that every single word of a language must be understood by persons other than the speaker. Likewise, the impossibility of languages without logical words does not imply that a language cannot have some nonlogical words which refer to private objects, that is, it does not imply that PLT-1 is true. Again, that a language must have logical words implies nothing about the possibility of a single man developing a language for and by himself, that is, does not imply that PLT-2 is true.
Apparently Wittgenstein knew that there are no criteria (in the sense of something independent) which prove that words have been used incorrectly. He also knew that the correctness of an application of a word is not determined by a rule whose formulation serves as a recipe or canon. His fundamental opposition to private language derives from his profound investigations into the nature of concepts and his strong inclination toward an extreme nominalism. This opposition is never crystallized in a definite argument, but its gist is, in crude form, as follows. Postulate:
(E ) The similarities and samenesses we find in things do not exist in rerum natura, that is, do not exist in things as we find them, independently of our finding them or of our referring to them in the way we do; they "come from the language" (Rhees p. 80) and at bottom consist of the fact that we "call" the things in question the same (Wittgenstein Secs. 146, 149, 185–190, 208–223, 348–352).
On a rigorous interpretation of (E ), we find a rationale for assumptions (A ), (B ), and (C ), as well as for the fact that the PLT has a chameleon-like and pervasive character. If we take (E ) literally, then to find a property in several things is to find that we "call" the things in question "the same" or refer to them with the same word. Thus, it is impossible to think that something is such-and-such without a language in which there is an expression (even if a very long phrase) which "constitutes" the such-and-such in question. This is assumption (A ). Also, (B ), without an expression "constituting" a type T, there is no type T for things to belong to. Similarly, to understand an expression is not to apprehend an independently existing (or subsisting) property but simply to know how and to what to apply it, and this includes knowing how to call certain utterances "true" in which the expression is correctly applied. This is, in fact, a generalization of assumption (C ).
We cannot say that a man in doubt about whether or not he used a word correctly must simply identify certain features of things without criteria and, armed with these identifications, test his uses of words. For on the extreme interpretation of assumptions (A ) through (E ) to identify a feature is to "call" a thing something. So, when the use of a word is at issue, the identification and nature of the thing is precisely what is at issue. The referents of one's previous uses of the word, as well as the uses themselves, are irrelevant. If one "calls" something "A," then it is A and a fortiori similar to the previous A 's; if one withholds the name "A " from it, then it is not an A, and a fortiori it is dissimilar to all A 's with respect to being an A. Clearly, it does not matter whether one's language is about private or about public objects; one's uses of words simply fail to be capable of being incorrect. They would seize reality so well that each "would have to be at once a statement and a definition" (Rhees p. 82).
Thus, the following question arises. If, on assumption (D ), language is a matter of rules and rules are the sort of thing that can be misapplied or not, how, then, is language possible after all? At this stage, obviously, we are not interested in proving anything but are anxious to find an explanation. Wittgenstein seems to suggest one: A man's uses of words can be incorrect only if they are compared with those of his fellow speakers. His "calling" something "A " is correct if his cospeakers now also call it "A." Then it is A and a fortiori similar to the things he and his cospeakers previously called "A." That is why obeying a rule of language is a practice (Wittgenstein, op. cit., Sec. 202). It is not necessary that the speakers of the language should call the thing in question "A " or that they call it "A " afterward. Nor is it necessary that they call it "A " or anything at all, or that they call it the same thing. It is just a contingent fact that they coincide in calling it "A." But this coincidence (or agreement) is an empirical fact that is necessary for the existence of language.
Such is the underlying argument of Wittgenstein's remarks (Secs. 146, 149, 185–190, 208–223; for a discussion of the role of Wittgenstein's extreme nominalism in his views about necessary truth, see Michael Dummett's "Wittgenstein's Philosophy of Mathematics"). He builds a Heraclitean picture of language as something living only in our actual use of it and changing according to our needs. But is this a true picture of the connection between language and reality?
Here we cannot discuss the whole issue of nominalism, but to this writer it seems indefensible. We could doubtless have classified objects in entirely different ways from the ways we in fact do. For instance, we might have had no color words, no terms for species of plants or animals, and instead have used, say, "sha" for some elephants and white roses and reddish sand, and "sho" for female elephants, eggs, and rivers. But even so, we should have had to find features of similarity in the things so classified, and these features would have provided tests for the correct application of our words. At any rate, the view that things are the same because we "call" them "the same" or because we refer to them with the same words can get off the ground only by postulating our recognition of the samenesses of words, that is, the similarities of noises whose application to things constitutes the similarities of the latter. A serious infinite regress would ensue if we also hold that our words are similar only because we "call" them so.
The several propositions that are often debated as the claim that private languages are impossible can be linked to each other only under the assumption of extreme nominalism. None of the arguments given for the claim appear to be successful. There may be no conclusive way of either proving or refuting this claim. Perhaps the only course is to build detailed and rigorous philosophical views on each alternative and assess the adequacy of such views by their consequences. This topic continues to be widely discussed in the literature, and many philosophers adopt a position different from that advocated in the present article.
the problem before wittgenstein
Carnap, Rudolf. "Psychology in Physical Language." In Logical Positivism, edited by A. J. Ayer. Glencoe, IL: Free Press, 1959. Ch. 8. Translated by Frederic (there called George) Schick from the German version, which appeared in Erkenntnis 3 (1932–1933).
Weinberg, J. R. An Examination of Logical Positivism. New York and London, 1936. Ch. 11.
Albritton, Rogers. "On Wittgenstein's Use of the Term 'Criterion.'" Journal of Philosophy 56 (1959): 845–857. An excellent collation and exegesis of the passages in which Wittgenstein discusses criteria; brings out certain internal tensions in Wittgenstein's conception.
Dummett, Michael. "Wittgenstein's Philosophy of Mathematics." Philosophical Review 68 (1959): 324–348.
Malcolm, Norman. "Wittgenstein's Philosophical Investigations." Philosophical Review 63 (1954): 530–559. An extended review and discussion that presents Wittgenstein's arguments against the PLT, with emphasis on the private language speaker's lack of a criterion of correctness.
Rhees, Rush. "Can There Be a Private Language?" PAS, Supp., 28 (1954): 77–94. A symposium with A. J. Ayer containing an exposition of Wittgenstein's views on private language understood mainly as PLT-2; conveys Wittgenstein's nominalistic rationale very well, without discussing it as such.
Wittgenstein, Ludwig. Philosophical Investigations. German with facing translation by G. E. M. Anscombe. Oxford: Blackwell, 1953.
the problem after wittgenstein
Ayer, A. J. "Can There Be a Private Language?" PAS, Supp., 28 (1954): 63–76. A symposium with Rush Rhees in which Ayer criticizes the PLT and emphasizes the need for criterionless identification at some point.
Carney, James. "Private Language." Mind 64 (1960): 560–565. Favors the PLT.
Castañeda, Héctor-Neri. "Criteria, Analogy, and Knowledge of Other Minds." Journal of Philosophy 59 (1962): 533–546. Discusses some problems of Malcolm's views on criteria.
Castañeda, Héctor-Neri. "The Private-Language Argument." In Knowledge and Experience, edited by C. D. Rollins. 1962 Oberlin Philosophy Colloquium. Pittsburgh: University of Pittsburgh Press, 1963. A symposium with V. C. Chappell and J. F. Thomson; surveys critically the arguments for the PLT and also replies to the other symposiasts.
Chappell, V. C. "Comments." In Knowledge and Experience, edited by C. D. Rollins. Vigorously defends the PLT from criticisms raised by Castañeda's discussion.
Garver, Newton. "Wittgenstein on Private Language." Philosophy and Phenomenological Research 20 (1960): 389–396. Favors the PLT.
Hardin, Clyde L. "Wittgenstein on Private Language." Journal of Philosophy 56 (1959): 517–528. Attacks the PLT.
Hervey, Helen. "The Private Language Problem." Philosophical Quarterly 7 (1957): 63–79. Accepts PLT* but claims that the real problem is whether or not one can recognize and classify one's sensations apart from their outward expression.
Tanburn, N. P. "Private Languages Again." Mind 72 (1963): 88–102. Attacks PLT-2 and the thesis that all entities referred to in conversation must be publicly available.
Thomson, J. F. "Comments." In Knowledge and Experience, edited by C. D. Rollins. In his part of the symposium with Castañeda and Chappell, Thomson contends that the PLT is too unclear for decision.
Todd, W. "Private Languages." Philosophical Quarterly 12 (1962): 206–217. Argues that "a private language, in the sense of a personal sensation language, is logically possible," refutes certain arguments of Malcolm, and attempts to show that a private language need not lack criteria of correctness.
Wellman, Carl. "Wittgenstein and the Egocentric Predicament." Mind 68 (1959): 223–233. Discusses critically some of Wittgenstein's arguments for the PLT in connection with the view that all knowledge is based on private experiences.
Héctor-Neri Castañeda (1967)