THE PRIVATE LANGAUGE ARGUMENT: LITERATURE

P.M.S.HACKER: OSTENSIVE DEFINITION & THE PRIVATE LANGUAGE ARGUMENT

(p.316) "Being a sample ... is a role conferred upon an object... a use to which we put the object, ... A sample represents that of which it is a sample, and hence must be typical of its kind. ...
Qua sample, the object belongs to the means of representation, and is properly conceived as belonging to grammar(x), in an extended sense of the term....
Far from samples being 'entities in reality' to which indefinables are linked by ostensive definition, they themselves belong to the means of representation. In that sense, there is no 'link between language and reality', for explanations of meaning, including ostensive definitions, remain within language."

Comment: This is intended to be a different argument, based on Wittgenstein's writings, for the impossibility of private ostensive definition leading to a private language.
The conclusion is that one of the four criteria for a successful ostensive-type definition is lacking for a private attempt to associate a sensation with 'S'. Further claims such as 'This is an S' are not contentful claims, they are no more than repeated ostensive definitions (see also the next section, from Candlish).

Hacker agrees that things (objects; reality) are involved in making propositions true and false. He also agrees that they are involved, in a different way, in making meanings for words.
What I think he means is that qua sample, the objectness of the object is not relevant; all that matters is its redness. The thing qua particular entity is not involved - its essence is not involved - in the process that gives words meaning. When I point at a thing, using it as part of ostensive definition, I am not making a link between words and things; the establishing of meaning does not involve the whole reality of a thing. The link between words and things (if there is one at all) comes later; we establish meaning, and then, for meaningful propositions, reality establishes truth. (Sounds like a slogan)
But warning bells should ring as Hacker honestly writes that in making the sample part of grammar(x), he is extending the usual sense of the term 'grammar(x)'. We can make anything anything if we are doing this.

(i) Is this true?
(ii) If it is, does it make any difference to the PLA?

To establish the meaning of 'red', I need, amongst other events, to experience a red object. If I didn't, a community of blind people could establish meanings for 'red' and 'yellow' which were the same as ours (Ayer makes this commonsense point). So experience of a property of things is necessary (if not sufficient). After we have the experience of the particular thing, we extract from the experience the property to which we intend 'red' to refer. So real experience of things is necessary for defining 'red' - it may not be sufficient, but it is necessary.
The redness, not the thingness, of the thing is the sample; the property is a representative sample, a typical example, of the property to which 'redness' refers. So the property, and hence the object which possesses the property, is involved with language. Suppose we choose to use 'reality(1)' in such a way that the object, as a particular (thing), is in reality(1), and 'language(1)' in such a way that the symbol and the property are in language(1). We deduce, trivially, that there is "no 'link between language and reality(1)', for explanations of meaning, including ostensive definitions, remain within language(1)". I propose that this is what Hacker is doing, obtaining impressive-sounding conclusions by equivocation.
After all, I could just as plausibly define 'reality(2)' to include particular objects, and their properties, such as redness - and 'language(2)' just to be the set of symbols, in which case I deduce, equally trivially, that there is a link between language(2) and reality(2), because ostensive definitions move outside language(2).

So, I label a particular aspect - a property, like redness - of a platinum bar in a vault in Paris as "1 metre"; it becomes the standard. Since that bar functions as the sample, it establishes the meaning of 'one metre'. The length property of that object is henceforth deeply entangled in language. To say, "The length of this bar is equal to one metre" is therefore to say "The length of this bar is equal to ... the length of this bar". (p.316): "The Standard Metre bar cannot be said to be (or not to be) one metre long". It is a tautology - true but contentless; true by definition; analytic. This is the 'meaning' stage.
This is correct; Physicists are taught it.
But of any other bar, including ones which are apparently identical, we can say, truly or falsely, "The length of that bar is equal to one metre". This is the 'truth' stage. Reality now intrudes, to establish truth-values.

Compare sensations: I label my sensation, when it first occurred, as 'S'; it becomes the sample. "This sensation is S" is then true by definition, because it is the one that we choose to use to establish the meaning of 'S'. This is the 'meaning' stage. But this does not imply that *all* subsequent sensations are tarred by the same brush. The other sensations are like the other bars; some are like the sample, and some aren't. This is the 'truth' stage.

I am arguing that no difference has been proved between the relationship between language and public objects, and that between language and private sensations; in both cases I can aim, firstly (i) to establish the meaning of 'one metre' or 'S', somehow using a real object in a quasi-grammatical role as typically possessing a property that we want to abstract as a universal (one-metre-ness; S-ness, and then, secondly (ii) to use 'one metre' or 'S' in sentences to express new propositions about the properties of other entities or events in reality (either in public or in private).

I note that Hacker also writes (p.316) that "it is noteworthy that one and the same object may function now as a sample in an explanation of meaning or evaluation of correct application, and now as an item described as having the defined property {now as establishing meaning, now as establishing truth}. But these roles are exclusive in as much as what functions as a norm for description cannot *simultaneously* {my emphasis} be described as falling under that norm".
Comment: He thus accepts that an aspect of a sensation could be used as the sample to establish the meaning of 'S', and then the same sensation, as an item, could establish the truth of a later proposition containing 'S'. If this is right, then a later proposition should have no problem, as an item, in establishing truth (even though this cannot be checked, of course).

Hacker's description of the PLA is familiar:
(p.317): "There can be no such thing as a rule for the use of a word which cannot logically be understood or followed by more than one person".
Comment: I have already argued against this.
Interestingly, Hacker describes the PLA as criticising Pears' 'C-crude', according to which the entire meaning of 'pain' is determined by reference to the sensation - summarised thus(p.369): "One knows what psychological predicate such as 'pain' means if one knows, is acquainted with, what it stands for - a sensation one has."
Comment: We have agreed that this crude view is a straw man, and clearly indefensible; the meaning of 'pain' is agreed to be significantly tied up with its teaching links.

(p.373): "(Wittgenstein's) arguments do not rely upon a form of verificationism...they merely remind us that if we are to talk of a rule for the use of a word, then there must be an operative distinction between the correct and the incorrect application of the rule."
Comment: I disagree. Why is an operative way of verifying whether an association is holding or not, essential before we are allowed to propose (talk of) that 'an association is holding'. Can we not 'talk of' it, because it is meaningless? But if it is meaningless, then it is so because the difference between its truth and its falsity cannot be operationally distinguished - which is to say, verified. So, in brief, it is meaningless because it cannot be verified - which looks like "a form of verificationism" to me.
To avoid this, Hacker would have to propose that "we cannot talk of it", not because it is meaningless, but because we are not following ordinary linguistic practice; following Pears, he would have to say that Wittgenstein is just telling us that the ordinary use of the word 'rule' includes the idea that we can provide a public operation for distinguishing between cases where the association has held, and cases where it has not.
Without linguistic essentialism, this is of little interest.

Hacker reckons that none of the four criteria for an ostensive definition, even if adjusted for the private case, can be satisfied. To respond to this, I need to establish what are the minimum, most abstracted, criteria, which would be sufficient to establish an association (meaning) between a signifier and a signified, and then judge whether these abstracted criteria are satisfied. As usual, I will strip out from his arguments unsupported positivism and verificationism, and Ordinary Languagism combined with linguistic essentialism.
Hacker's four criteria, initially expressed in the concrete form suitable for publicly observable qualities, are (p.370):

(Ci) An indication of the type of the word eg. that it is a colour word ('stage setting')
(Cii) A method of establishing the original association between the standard and the word (an 'ostensive gesture').
(Ciii) A standard for comparison (a 'sample').
(Civ) Some way of laying a sample alongside a new instance and confirming match or mismatch (a 'method of projection') - some independent check on the correctness of the assignment of the word to new instances.

Hacker makes no attempt to adjust them. He proposes that they are necessary conditions for any intelligible ostensive definition, in any kind of symbolic system, and that private ostensive definition does not satisfy them.
The reasonable-seeming criteria derive from public symbolic systems, which are devices for communicating between non-philosophical sensing beings, living in an agreed public world of things. The observed fact that these describe ascription of meaning in a public symbolic system is not a proof that they must hold for all such symbolic systems. We need a proof - or at least an argument - if the description is to become a prescription. There is otherwise a clear danger that we start merely with features (not 'conditions') of public ostensive definitions, and then tut-tut if they are not displayed by private analogs.
So, before extending the conditions(x) so that they apply to symbolic systems operating in all circumstances, we need to go from describing to justifying - from 'is' to 'must be'. What in general, in the abstract, is the aim that the criteria are attempting to achieve?
Our aim is to attach a meaning to a symbol - to give a symbol meaning. 'Meaning' is not easily defined, but we will suppose that:
(a) it involves an association between a signifier and a signified.
(b)
We assume that our memory is totally reliable - as previously discussed. We reject unproven positivism with regard to mental entities. We take seriously warnings about the baleful effects of incautious models of the workings of the mind. We presume that sensations and memories of sensations are aspects of private experience, since to deny this at the start would make the argument pointless.
What features are provably necessary, as opposed to desirable or present in public, for establishing meaning in a symbolic system:

ON (i): Since a word is going to play a particular role in a symbolic system, we need to establish this role. In the particular public case, we need to be able to communicate which aspect of the sample the word is referring to; does 'sepia' refer to the colour, the shape, the number, the size, or the degree of interest? In general, we are establishing the type or category to which the aspect of the instance belongs - which also establishes its grammatical role.
So a justifiable general form of the first condition is:
C'(i): There must be some way of communicating and establishing the type, the aspect, of the identified instance to which the word refers. This is satisfied by the person's ability to communicate with himself, and decide to which aspect of the sensation the word refers.

ON (ii): a particular, typical method of establishing associations in public is an ostensive gesture. This is impossible in private. What the gesture does is to draw a person's attention to the object to which the word refers.
In general, what is necessary is the drawing of attention. In the individual case this is easy, since the person has his own attention. He attends to the sensation, and which associates 'S' with it.
So a justifiable general form of the second condition is:
C'(ii): There must be some way of drawing attention to the instance which becomes the reference standard. This is satisfied by the person's ability to control his own attention.
How can we give an independent criterion to establish that the attention has been drawn? It is meaningless to talk of 'attention being drawn' if there is no operational distinction between 'attending' and 'thinking that one is attending'.
Comment: We cannot give, and do not need to give, an independent criterion. To insist on one is verificationist.
(p.371): "Concentrating one's attention on a sensation...does not determine a criterion of identity for subsequent uses of 'pain', and emphatically saying 'This' does not either, for what is the 'this' one concentrates on? (One cannot reply 'a pain', since that presuppose the very concept one is trying to determine. Nor can one reply 'a certain sensation', or even 'a something', for these are words in our public language with a determinate grammar of their own)"
Comment: I am hardly expecting the isolated person, establishing his private putative language, to be chatting away in normal public language, saying to himself, or to the watching philosopher: "I am going to take 'S' to mean THIS sensation". The act of drawing the attention is non-verbal.

ON (iii): The role of the sample is played by remembering the sensation. Wittgenstein is right that we should be very wary of referring to 'memory of the sensation' as though it is a thing in a mental world. Expresses in ordinary language, 'We remember the original sensation, and realise that the new sensation is kind we remember'.
The general purpose of condition (ii) is to ensure that there is some system of reference for the correct use of a word (even if first-person judgements of correctness cannot be corroborated - cannot be established independently).
So a justifiable general form of the third condition is:
C'(iii): There must be some system of reference for putative checking of new cases of S. (I have avoided the word 'sample' because of possible public object associations) This is satisfied by the memory of the original sensation.
(p.371): "Remembering that pain is THIS (conjuring up a mental image) is no substitute, for...there can be no criterion for reproducing the *right* sample or mental surrogate. Here there would be no distinction between remembering correctly the connection between 'S' and the paradigm that defines it, and merely thinking that one remembers, and there is no independent court of appeal to fix that distinction."
Comment: I accept that there is no criterion - no independent court of appeal. To move from this to there being "no distinction" is positivism or verificationism.

ON (iv): If only one person is involved, it is ordinary-languageism or verificationism to insist that an independent check on projections is required. Independent checking of linguistic ascriptions is very nice, but if it is not available, then we do the best we can without it. 'Right' will just mean 'is judged by me to be right'. I may be mistaken. But a symbolic system in which individual mistakes cannot be corrected independently, is not therefore unintelligible - it is just ... prone to unidentifiable mistakes. To suggest that the possibility of such mistakes is unacceptable is a form of positivism.
A sensation cannot, of course, be laid alongside a memory of a sensation - nor indeed alongside an object - in the way that a metre rule can be laid alongside a string.
So a justifiable general form of the fourth condition is:
C'(iv): There must be some way of comparing a new instance with a sample. This is satisfied by the private ascription of 'S' to the sensation.
'One cannot perceive it (a mental image)...Nor can one lay it alongside reality for match or mismatch'
Comment: This is Ordinary Languageism, loading the scales in favour of the particular form of ascription of meaning possible in public.

We conclude that *starting* from four features of public ostensive definition, we can obtain four generalised, justifiable, features describable as 'conditions'.
But these four conditions are satisfied by private definitions. So a private symbolic system is possible.

Any comments?