Perfect Speaker Theory

AC Grayling

One thing we all agree is that meanings do not just fall out of the sky; they were and are at least in important part generated by what users of signs do with them. This suggests that the semantics of natural language is the product of the history of its pragmatics. It does not automatically follow, of course, that therefore the right way to go about giving a theory of meaning is in some sense to do it from the pragmatic end: but in the larger project I wish to suggest a reason for doing so, and I suggest that what increases the plausibility of this thought is evidence that certain problem cases in our understanding of language are amenable to an approach in which pragmatic considerations are given full emphasis.

I do this in this paper by giving a brief sketch of part of a theory I call Perfect Speaker Theory (PST), which, despite its grand name, is a modest attempt to provide a methodological perspective from which to comment on some familiar debates in the philosophy of language and mind. For present purposes, it is the simple and natural way that PST deals with certain well-worn problems in these debates that prompts the suggestion that pragmatic considerations should be treated as irreducibly central to meaning.

At this juncture I should state that by 'pragmatics' I mean what, over and above what is standardly assigned to theories of structure, reference, sense and satisfaction conditions, relates to the use of expressions of the language on given occasions, where what the speaker intends to convey, and the means the speaker employs in conveying it on the occasion, are especially in focus. I want to say that among pragmatic considerations questions of force and point (for my purposes especially point) are central. These facets of linguistic behaviour include, essentially, considerations about interaction between speakers and their audiences.

I found, in working at these ideas, that they afford independent reason for agreeing with some of what Grice says. I can only claim some agreement, because Grice is professedly inclined to accept two things which PST is not, namely, the distinctness of meaning and use, and the privileged role of a notion of truth in any account of natural discourse. PST's commitment to arguing for the dominance of pragmatic considerations is allied to a view that notions of truth (which in my view is not one insubstantial thing but a number of different substantial things for which the expression 'truth' is a homophonic dummy) play roles of a different kind (I argue for this view elsewhere.) A major tension between PST and Grice's theory is that because the former has it that the crux in meaning is point, which is to be explained in terms of speakers' intentions to mean something on an occasion, conventional meaning is to be characterised as the dry residue of speakers' meanings, agreed in the language community under constraints of publicity and stability (another view I argue for elsewhere). Now Grice has given considerable attention to showing that speakers' intentions and conversational maxims are insufficient for a full account. PST aims to suggest that from these resources a satisfying account of meaning can be brought into view without having to import considerations from outside the reach of pragmatics, suitably construed.

There are many debates about language which can be exploited to show how PST pushes us in the indicated direction, but I will advert to just four very familiar examples, and a fifth less familiar one. The problem about using familiar examples is that the literature on them is large, so with no space to review it I have to be unceremonious; and also, everyone has a favourite view about these cases, so doubtless what follows will seem too swift. The cases are (1) the proper understanding of natural language analogues of logical constants, (2) presupposition-failures for certain uses of verbs of doing and trying, where the appropriateness of certain locutions comes into question because some implied condition for their appropriate utterance fails to obtain; (3) some questions about referential uses of definite descriptions, and (4) a certain application of the Twin Earth story. This looks like a heterogeneous collection, but I submit it as a virtue of PST that it identifies some patterns in the aetiology of problems arising in relation to them.

And finally (5) I suggest that PST provides a way of approaching the question whether, for any given natural language, there is something that counts as 'the language', or whether a natural language is a vaguely bounded family of idiolects, perhaps as numerous as their speakers. But first it is necessary to define a Perfect Speaker (PS), and to explain why he is so called. A PS of his language is one who so uses it that whenever he makes an assertion (and mutatis mutandis for other kinds of utterance) he:

(1) expresses his intended meaning as fully as, if not more fully than, his audience needs in the circumstances;

(2) expresses his intended meaning as exactly as, if not more exactly than, his audience etc; and

(3) is as epistemically cautious as the circumstances do or might require, if not more so, with respect to the claims made or presupposed in or by what he says.

Together these conditions make the PS determining or overdetermining in respect of the point and epistemology of his utterances. Note that they do not cover the same ground as Grice's maxims – although there is some overlap – and they do not pretend to be regulative for ordinary communication. Together they make a PS of his language one who is in practice absolved a certain duty, namely, the responsible speaker's duty to stand ready to clarify, qualify or defend what he says if called upon to do so by hearers who have and are exercising anything recognisable as normal competence with the language in use. The PS is absolved the duty of restatement because he is by definition never in a position to have to fulfil it, as long as his audience is as specified.

The PS is of course an idealisation, and one which immediately prompts questions. So it is important to be clear about what kind of idealisation he is. Let us distinguish him from two other possible kinds of idealised speaker whom I shall call the Ideal Speaker and the Optimal Ordinary Speaker. (Strictly, it is only the former who is an idealisation, because there might in fact be Optimal Ordinary Speakers). The Ideal Speaker (IS) satisfies the first two conditions, but for a third has 'is omniscient'. Thus the IS is god or relevantly godlike. One immediate difference between an IS and a PS as I define the latter is that the language of an omniscient speaker would have to be apt for the expression of everything, whatever that means, whereas the language of a speaker who suffers the finitary predicament (is finite in knowledge and powers) cannot be guaranteed to be apt for the expression of everything. So not only does a PS differ from an IS in being subject to finitary constraints, but his language (which is an ordinary natural language after all) carries the mark of that finitude also. This might make it seem that the PS should not be so described, but rather that he should be thought of as an Optimal Ordinary Speaker (OOS), i.e., an ordinary speaker who simply is as careful and precise as he can be, and that is all. We might be trying to be such speakers when we do philosophy or law; so OOS–or 'optimal speaking' by ordinary speakers – might be relatively commonplace. But there are significant differences between OOSs so considered, and the PS I require for my model. One is that an OOS, an ordinary speaker doing his best, must be allowed at times to be in states where the beliefs and intentions that determine the content of what he says are not transparent to himself. This means that an OOS may at times fail to satisfy (1) and (2), and at times all of (1) to (3). But by stipulation a PS is one who always satisfies all three. So his meaning, intentions and

....2/