I said that the distinction between semantics and pragmatics was supposed to be that the former deals with the acontextual mea

I said that the distinction between semantics and pragmatics was supposed to be that the former deals with the acontextual meanings of sentence types, while the latter addresses the social uses of linguistic expressions in context.But there are two reasons why that characterization is too simple.The first reason is there is an important sense in which most sentence types do not have acontextual meanings.The second is that as we shall see later on, social-use factors interpenetrate in certain special ways with what we would otherwise think of as propositional meaning.

Here is the sense in which most sentence types lack acontextual meanings.Recall the phenomenon of deixis, brought up in in Objection 4 against the Truth-Condition theory, and consider a heavily deictic sentence.Suppose you and I come into an empty classroom and find the following words written on the blackboard: "I have never been to a karaoke bar, but you and I will visit one tomorrow morning."Unless we can find out who had written those words when and to whom, we do not know what has been said (even though we know something about what has been said).In terms of the possible-worlds theory, we do not know the sentence's intension.In fact, if the sentence had been scribbled on the board merely as a linguistic example and no referents had even tacitly been assigned to its deictic elements, it would not even have an intension.

The common moral of the original Objection 4 and of this last argument is that a sentence's complete truth condition depends on contextual factors.And even if one does not accept the Truth-Condition theory of meaning, one can see that a sentence's meaning, in the sense of its propositional content, depends on context in just the same way.

Cresswell (1973) distinguished between two kinds of pragmatics: "semantic pragmatics" and "pragmatic pragmatics."Semantic pragmatics deals with those elements of meaning in the sense of propositional content that simply do depend on context; it is the discipline that tells us how propositional content is determined by contextual features.But before we say more about it, let us deal with Objection 4.

The problem of deixis

Returning to Davidson's problem:He needs to find a way of formulating T-sentences that accommodates deictic or indexical elements without getting truth conditions wrong.I mentioned Davidson's own proposal for doing this.Other notable attempts have been made by Weinstein (1974) and especially Burge (1974), but here I shall present a simple idea suggested by Harman (1975).[1]

We saw that one drawback of Davidson's proposal was its limiting the potentially relevant contextual factors to speaker and time.There are many others.To take a slightly exotic example, hemisphere.[2]"It is autumn" is true as I write this in North Carolina, USA, but it would not be true were it uttered simultaneously in Sydney or in Buenos Aires.(Nor is the relevant hemisphere determined by the speaker's location; it depends on the audience and on conversational purposes as well.)So we need an approach to deictic target sentences that does not presuppose a fixed number of contextual variables.

Let us get the whole job done in one stroke.We can relativize "true" to contexts -- since we already know that truth does depend on context -- and introduce a function, a, that will look at a deictic element occurring in a context and tell us what that element contributes in that context to propositional content.Then we can write the right-hand sides of Davidson's T-sentences in terms of what a assigns in the context C to each deictic element in the target sentence.Thus:

"I am sick now" is true in C if and only if a ("I",C) is sick at a ("now",C).

"I have never been to a karaoke bar, but you and I will visit one tomorrow morning" is true in C if and only if a ("I",C) does not go to a karaoke bar duringa (perfect tense, C)[3] but a ("you",C) and a ("I",C) visit a karaoke bar during the morning of a ("tomorrow",C).

Problem solved.The technical problem of formulating T-sentences, that is; further philosophical questions can and will be raised about a.

The intensional logicians dealt with deixis by relativizing truth to an "index," which was a fixed set of contextual variables.Montague (1968) and Scott (1970) took an index to be an octuple consisting of a possible world, a time, a place, a speaker, an audience, a sequence of indicated or demonstrated objects, a "discourse-segment," and a "sequence of assignments to free variables" (never mind what those last two mean).In this system, a truth condition assignment would look like this:

"I am sick now" is true at <w,t,l,p,a,i,d,s> if and only if in w, p is sick at t.

But this shares the drawback of Davidson's method, though not as severely, in that it arbitrarily restricts the number of contextual features that can be cited.[4]We have no way of foreseeing what further such features might become relevant to the truth of an utterance.

For example, we have already introduced one unexpected variable, hemisphere (as between southern and northern).There are plenty of others, seemingly without limit.The truth of “It is 5:00 p.m.” depends on time zone, an entirely conventional construct.(As Wittgenstein once noted, time zones are bound to our planet; “It is 5:00 p.m. on the sun” has no truth-value.)And some locutions presuppose a kind of vantage point, often distinct from the place of actual utterance, that can shift even within a single sentence (Fillmore 1975; Taylor 1988).Take

(1)a.Zonker went to Uncle Duke’s party.

b. Zonker came to Uncle Duke’s party.

(2)I’m going out to clean the pasture spring;…. –You come too.(Robert Frost, “The Pasture”)

(1a) and (1b) can have the same truth condition, but (1b can be said properly only by a speaker whose assumed vantage point is the party location itself.(Note too that what counts is the vantage point at the time of the party under discussion, not at the time of utterance; this is yet another context variable, usually called the reference time.)In (2), the vantage point shifts fluidly from the place of utterance to the pasture spring, or at least to somewhere along the journey at which the speaker is pictured as being ahead of the hearer.

Arriving at Princeton to give a talk, I am met by a former colleague whom I last saw teaching at Wellesley.I ask, “Are you here now?,” meaning not whether she is physically located in Princeton (duhh) but whether she is now employed in the Princeton philosophy department (Nunberg 1993: 28); thus truth-value can vary with employing institution.Or take

(3)Tomorrow is always the biggest party night of the year.

Uttered on the Friday before classes begin (Nunberg 1993: 29; Nunberg credits Dick Oehrle with the example).“Tomorrow” in (3) cannot refer, as it usually would, to the day or night following the date of utterance; it refers to a type of date on the students’ academic calendar, namely to the annual Saturday before classes begin.

I could go on and on.The moral is that we cannot ever be sure we have anticipated all the context variables that can affect truth-value.So I would advise the intension theorists to avail themselves, instead, of my mighty assignment function a.

The work of semantic pragmatics

The trick is to find out howa is computed, that is, what rules we use in particular contexts to fill in the missing chunks of propositional content corresponding to deictic elements.Presumably each such element in the language is governed by an appropriate rule.

For example, we might look at the pronoun "I" and suggest that in a context, "I" always denotes the speaker.Turning to "now," it seems reasonable to say that "now" always refers in context to the time of the utterance.In fact, these first attempts are too simple.("I" can be used as a device of deferred reference, as in "I'm in the parking lot" meaning that that is where my car is.Sometimes "I" is used as a bound variable, as in "If I'm a music department, I'm a snake pit."The temporal reference of "now" can be deferred also, as when we are looking at a time-line representation of the evolution of life and, pointing, I say, "Now the dinosaurs appear," or when you leave a message on your answering machine that says "I am not home now.""Now" is sometimes spatial rather than temporal at all -- "Now Hillsborough Road crosses Airport Road and becomes Umstead Drive" -- and sometimes not even spatiotemporal -- "Now comes the first prime number whose square is greater than 1,000.")But one job of semantic pragmatics is to refine such rules until they are adequate to the data.

The intensional logician David Kaplan (1977) thinks of such rules as functions.As an intension is a function from worlds to extensions, a semantic-pragmatic rule is a function from contexts to intensions.At the level of the sentence, the intension is a function from worlds to truth-values.Kaplan calls that the sentence's "content," and as before, it corresponds to the traditional notion of a proposition.The composite semantic-pragmatic rule is a function from contexts to contents; Kaplan calls that "character."Content is what is left undetermined by the deictic sentences in our examples; character is what does determine content given all the relevant contextual features of a context of utterance.

I said that when we encounter the karaoke sentence unprepared, we do not know (in full) what it says.And I was right.But there is another perfectly good sense in which we understand the sentence itself, and virtually any English speaker understands "I am sick now" entirely out of context.Kaplan argues that the "m"-word should be reserved for character rather than for content, on the entirely reasonable ground that ordinary English speakers surely know the meanings of everyday deictic sentences even when they do not know the contextual parameters that would fix those sentences' contents.Yet content in his sense is also still a perfectly good thing to mean by "meaning."It is hardly a matter for heated dispute.

Computinga and/or characterizing character is not the only task of semantic pragmatics.Another and very vexed one is disambiguation.Many sentences, like "Visiting philosophers can be boring," "Ted is lying about meditating" and (Paul Ziff’s example) "The mouse tore up the street," are obviously ambiguous.And in fact, almost every sentence we ever encounter in life is technically ambiguous, in the sense that it has one or more possible if farfetched meanings in addition to the one that would normally be intended by an utterer.Yet we rarely pause to think, or even notice that we are choosing from among a range of possible meanings (not merely filling gaps in an otherwise unique propositional content).How we do this is a deep question, much deeper than that of how we compute a.Certainly too deep for this book, though some hints will be furnished in Chapter 13.