Perfect Speaker Theory - page 2

Suppose, however improbably, that there are no other aspects which in the circumstances could direct the audience's attention to the man whom the speaker says is happy tonight. It would remain enough that it is a plausible belief for the speaker and audience to hold about him, and which the speaker can exploit for referential purposes, that he is drinking champagne.

It seems entirely natural to say that what the PS does in these cases is what all speakers are enthymematically doing anyway, as an entirely natural extrapolation of the cases shows: if the audience responds, 'he is not her husband, he is her lover', 'it is not champagne, it is Perrier water', the speaker replies, 'Oh well I thought he or it was such-and-such–but anyway you know what I mean'.

This does not defend in any way Russell's allegedly pure attributive theory, in which the form of the example is represented as 'the uniquest x to satisfy F, Gs', because the epistemic restraint built into specification of F imports something which is additional to the purely conventional aspect of the other expressions constituting F. We can best see what is at work by reminding ourselves of Kripke's suggestion. For Kripke, the distinction lies between semantic and speaker reference. For some idiolect, the semantic referent of an indexical-free designator is fixed by general intentions of speakers to pick out a given object by its means. Speakers' referents are given by specific intentions on specific occasions to refer to a given object. And again familiarly, Kripke identifies simple and complex cases; in the former, a speaker's specific intention just is his general intention. In the latter, his specific and general intentions diverge, but as a matter of fact he believes that his specific intention determines the same object as his general intention. He might be right or wrong but still succeed in referring. On this view, D's 'attributive' case is nothing but the simple case, his 'referential' case nothing but the complex case.

On the PS view, however, all cases are to be understood as complex cases. Reflecting on the oddity of Kripke's distinction suggests why they have to be so. It is surely never the case, intentions to deceive apart, that speakers believe that their specific and general referential intentions diverge. Their choice of designating expression is governed by a desire to succeed in making reference. So as a matter of fact speakers are always in Kripke's complex case: they believe that their specific and general intentions coincide. Indeed, the distinction between the intentions is not one that speakers make from inside their referring practice: it is one that retrospectively offers in theoretical reflection on the fact that many beliefs are defeasible.

But the PS builds recognition of this fact of life into his practice. So his is expressly the complex case: his specific intention to refer is effected by articulably cautious attribution of certain properties to the referent by means of expressions whose conventional meaning is apt for the task.

Among the points worth noting here are the following. A familiar contrast between conventional meaning and speaker meaning lies in the offing of those remarks: it is just such a Gricean distinction from which Kripke takes his start, but this way with the outcome hints that any account of conventional meaning is, pace Grice, parasitic upon an account of speaker meaning: that the former is, in a sense worth specifying, the dry residue of agreements forced among speakers seeking success for their intentions to mean. This is a point which lies just off-stage in this paper.

The second is that what gave rise to the debate about a reference-attribute distinction is ellipsis: the thought is that making speakers express too little of what a model speaker–the PS–would say is exactly what generates the problem. And there is nothing philosophically problematic about ellipsis.

But now we need to bring in another feature of PS-hood, implied by the conditions but not so far exploited. This feature comes out in the final case I consider: a certain employment of Putnam's Twin Earth argument.

Putnam challenged two received assumptions, the first being that to know the meaning of an expression is to be in a certain sort of psychological state, and the second being that the meaning of an expression determines its extension. These assumptions entail that psychological states determine the extension of terms. Putnam's twin-earth case purports to show that this cannot be so, giving us his celebrated conclusion that meanings are not in the head. Of course this applies just when the psychological states in question are narrow, that is, understood concordantly with 'methodological solipsism', the view that psychological states supervene on intrinsic states of an individual considered independently of anything besides, in particular without relation to environmental factors causing or being effected by those states. Now on PS theory the crux for meaning is point: explanation of what an utterance means has to make essential reference to what speakers intend to convey. This means that a psychological state privileged by the theory, viz. 'intending to mean p', determines the extensions of expressions used. Putnam's twin-earth considerations seem to block this. Do they? I suggest not. The claim I think we justify by appeal to PS theory is this–and here we introduce the additional feature of PS-hood required by the conditions–: that 'intending to mean p' does indeed determine extension, under the two constraints (a) that the words used are used according to what I shall label 'the Best Dictionary', which reports the agreed, relatively stable senses of expressions employed as tokens of communicative exchange in the linguistic community, and which does so according to the best current theories of the linguistic community, and (b) that the PS's audience of normally competent ordinary speakers will so take it.

We get at this thought by noting something about how the twin-earth case is set up. The first thing to note is an apparently tangential point, namely that rejection of methodological solipsism does not by itself entail some kind of realism about the domain of application of the terms whose meanings are to be understood broadly. For so far all that we have identified as required in addition to a speaker's knowledge of meaning is the existence of other speakers whose interactions with him and one another constrain that knowledge. If this is not enough to serve as the broad context, then we need to find the reason in Putnam's twin-earth argument. Does his argument show that unless there is H2O and XYZ out there on earth and twin-earth respectively, Oscar and his twin could not respectively have meant 'water' and 'water'? That of course is not what Putnam sought to conclude: his argument was not aimed at proving the existence of the external world, but at showing that meanings cannot be individuated narrowly. So there are three assumptions at work in the twin earth case. The existence of the different waters–water H2O and water XYZ–is assumed, as is the qualitative identity of the twins' narrow states, as in the twins' woeful ignorance of views current about water in their worlds. Thus richly equipped, the argument has no difficulty in delivering its celebrated conclusion. But before we accept these premisses in our welcome of the conclusion, we should enquire whether the twin-earth tale could not be as well or better told, because more parsimoniously told, using just the idea that the meanings of expressions in a language are the agreed dry residue of speakers' meanings.

Any individual's problem is that he does not know everything that all other speaker's jointly know about the meaning of the expressions in his language. This is unsurprising: if some best and latest dictionary pooled a community's knowledge of meanings, it would be a rare individual whose linguistic knowledge came close. Such a dictionary would report the knowledge possessed by the completest speakers–OOSs–including, for example, chemists and hydroengineers. Now let us bring in the PS, who knows everything in the best dictionary. Note that the PS is not omniscient, like the IS: he just knows the Best Dictionary, which is finite and fallible, but reports at their completest the linguistic community's dry residue agreements over what expressions can be used to do. In knowing the meaning of 'water' the PS therefore knows that, in the latest state of chemical theory on earth, it is stuff of molecular structure H2O. And let us also introduce the twin-earth PS, who is by definition equally up to date. Then he knows that 'water' according to twin-earth chemistry is XYZ. In this idealised state of knowledge of meanings- that is, where a PS has at his disposal the linguistic community's best joint knowledge–we find that when he says 'water' he intends to refer to water, that is H2O, or in the case of the twin-earth PS, then to water XYZ; and so in either case the PS's grasp of the expression's meaning determines its extension, and the psychological state in which his grasp of the meaning consists is broad. But this is not because it is related, causally or in some other way, to water, but rather to theories about water, because he is speaking in conformity with the Best Dictionary, that is, with the fullest shared knowledge of meanings, in accordance with the best current theories held by the linguistic community.

The trick in Putnam's thought experiment is that the people talking about water are ignorant, in the way people are apt to be, as to the best current theories about the stuff. So we who know something more about H2O than, ex hypothesi, they do, can see the point as to what else is needed for them to achieve a successful reference: namely, to intend to refer to just that stuff, and not something that cannot be distinguished from it when one's level of knowledge about it is suitably impoverished. Putnam's thought experiment does its work because it premisses that the speakers on earth and twin-earth should be identically ignorant in respect of what their referential intentions would be if they were ideally, or just more, knowledgeable: that is, if they were, or at least approximated the status of, PSs. Now the third condition in PS-hood was epistemic caution. The suggestion here is that it is a corollary of being epistemically cautious that one be as well informed as one can be for the purposes of satisfying (1)-(2). The PS, in obedience to (3), knows (or carries around and consults) the Best Dictionary. So when he refers, he does not do so under Putnamiam epistemic privation of the kind suited to making the twin-earth case plausible.

I conclude now by drawing a couple of morals. The point of the PS model is claimed to be that by applying it to cases like the ones just canvassed, we get perspicuous accounts of what is being said and done, and they show that the problematic character of the canvassed cases is an artefact of failing to give full weight to considerations made salient by the pragmatics of Perfect Speaking. What helps with the problem cases is appeal to considerations of point and epistemic aptness; the PS is 'perfect' in his practice with respect to both–and in being so is such that the problems do not arise for him. This, my inference is, suggests that we should look to beliefs and intentions for the basis of a general account of meaning.

In the first two cases the problems were generated by incomplete determination of point in the formulation adopted by a putative ordinary speaker. In the second two cases the problems were generated by epistemic underdetermination in the cases; the putative ordinary speakers failed in achieving their intended targets of utterance precisely because of it.

A PS perspective on the cases brings a salutary reminder to our attention. It is that no non-idly employed sentence of natural language exists outside a pragmatic frame. For every non-idle use of a sentence the particularities–how things are in respect of the utterer, his intentions, his audience, the current state of the language, and the circumstances of its use – determine the meaning of what is said on that occasion.

It is these two thoughts together that suggest the third – at this juncture I do not claim they do more – namely that the conventional meanings of expressions in natural language are the precipitates of the linguistic community's tacit agreements to place the use made of certain signs under publicity and (relative) stability constraints, so that the ends of communication can be realised. These agreements are the conventions which dictionaries report. Any account of meaning so understood would have on this view to make essential reference to the pragmatic considerations–and central among them, point–which figure thus in its genesis.

Now it is widely held that–and I quote Searle–"meaning is more than a matter of intention, it is also a matter of convention" and in illustration he quotes Wittgenstein's remark in Philosophical Investigations, 'Say "it's cold here" and mean 'it's warm here"'. There is no inconsistency between agreeing with this and accepting what the PS theory says is central. For one thing, conventional meanings are, so the theory seeks to suggest, the precipitate of intentions anyway: what an expression means is what , in effect, it is intended to mean by the linguistic community. For another, nothing in PS theory excludes the obvious, which is that the point of a speaker's utterances, even a PS's utterances, cannot be individuated independently of facts about the conventional meanings of the expressions he uses, the context, the speaker's attitudes, and whatever other pragmatic features are required besides. But what it does do is to say that point figures centrally.

By way of conclusion, and even more briefly than in the familiar problem cases mentioned, I suggest that among the other things PST can offer is a simple solution to the question whether there is such a thing as 'the language' for any natural language. The thought that there is such a thing is held by some to play the role that naive realism does in theories of perception; it is the dumbo view, to be replaced by more sophisticated views such as that there are as many personal (so to speak) paroles as speakers, and that what we too loosely call communication is in fact a form of translation or, more accurately, interpretation. A natural desire to respond that at least a notional 'the language' is required to provide a norm–which among other things can be invoked to explain such phenomena as (say) the differing divergences of idiolects from majority intelligibility–might be met with the riposte that such norms are in fact constituted not by something which is genuinely 'the language' but by the family of idiolects of an historically favoured class of speakers (the currently rich, the currently powerful, the people currently in charge of the culture. There could be–after the revolution, say, there might be – quite different folk in these roles, speaking a different family of idiolects). And so the debate might proceed. But one thing that would help to give it shape would be to offer something that could count as a criterion of identity for 'the language' if there is such a thing. This is where PST comes in: for such a criterion is offered by saying: the language is what the PS speaks. And this genuinely does offer a normative conception across which mappings must fall if the very idea of an idiolect of some language is itself to make sense.