Pragmatics and Semantics

ly specify a certain language L in order to characterize the speech habits of a community P by saying that they are using that language and conforming to conventions involving it. 6 The uses of linguistic forms on which their semantics depends, and which therefore constitute the pragmatic basis for their semantics, are their past uses: what an expression means at time t in a given community depends upon the history of its uses before t in the community. But of course, pragmatics is not merely concerned with past uses. Beside the past uses of words (and constructions) that determine the conventional meaning of a given sentence, there is another type of use that is of primary concern to pragmatics: the current use of the sentence by the speaker who actually utters it. That use cannot affect what the sentence conventionally means, but it determines another form of meaning which clearly falls within the province of pragmatics: what the speaker means when he says what he says, in the context at hand. That is something that can easily be separated from the (conventional) meaning of the sentence. To determine “what the speaker means” is to answer questions such as: Was John‟s utterance intended as a piece of advice or as a threat? By saying that it was late, did Mary mean that I should have left earlier? Like the pragmatic basis of semantics, dimensions of language use such as illocutionary force (Austin, Searle) and conversational implicature (Grice) can be dealt with in pragmatics without interfering with the properly semantic study of the relations between words and their designata. So the story goes. There are two major difficulties with this approach to the semantics/pragmatics distinction — the Carnapian approach, as I will henceforth call it. The first one is due to the fact that the conventional meaning of linguistic forms is not exhausted by their relation to designata. Some linguistic forms (e.g. goodbye, or the imperative mood) have a “pragmatic” rather than a “semantic” meaning: they have use-conditions but do not “represent” anything and hence do not contribute to the utterance‟s truth conditions. Because there are such expressions — and because arguably there are many of them and every sentence contains at least one — we have to choose: either semantics is defined as the study of conventional meaning, or it is defined as the study of wordsworld relations. We can‟t have it both ways. If, sticking to Carnap‟s definition, we opt for the latter option, we shall have to acknowledge that “semantics”, in the sense of 7 Carnap, does not provide a complete (descriptive) account of the conventional significance of linguistic forms. The second difficulty is more devastating. It was emphasized by Yehoshua Bar-Hillel, a follower of Carnap who wanted to apply his ideas to natural language. Carnap explicitly said he was dealing “only with languages which contain no expressions dependent upon extra-linguistic factors” (Carnap 1937: 168). Bar-Hillel lamented that this “restricts highly the immediate applicability” of Carnap‟s views to natural languages since “the overwhelming majority of the sentences in these languages are indexical, i.e. dependent upon extra-linguistic factors” (BarHillel 1970: 123). In particular, Carnap‟s view that words-world relations can be studied in abstraction from use is no longer tenable once we turn to indexical languages; for the relations between words and their designata are mediated by the (current) context of use in such languages. The abstraction from the pragmatic context, which is precisely the step taken from descriptive pragmatics to descriptive semantics, is legitimate only when the pragmatic context is (more or less) irrelevant and defensible as a tentative step only when this context can be assumed to be irrelevant. (Bar-Hillel 1970: 70) Since most natural language sentences are indexical, the abstraction is illegitimate. This leaves us with a number of (more or less equivalent) options: 1. We can make the denotation relation irreducibly triadic. Instead of saying that words denote things, we will say that they denote things “with respect to” contexts of use. 2. We can maintain that the denotation relation is dyadic, but change the first relatum— the denotans, as we might say — so that it is no longer an expression-type, but a particular occurrence of an expression, i.e. an ordered pair consisting of an expression and a context of use. 3. We can change the second relatum of the dyadic relation: instead of pairing expressions of the language with worldy entities denoted by them, we can pair them with functions from contexts to denotata. 8 Whichever option we choose — and, again, they amount to more or less the same thing — we are no longer doing “semantics” in Carnap‟s restricted sense. Rather, we are doing pragmatics, since we take account of the context of use. Formal work on the extension of the Tarskian truth-definition to indexical languages has thus been called (FORMAL) PRAGMATICS, following Carnap‟s usage (Montague 1968). As Gazdar (1979: 2-3) pointed out, a drawback of that usage is that there no longer is a contrast between “semantics” and “pragmatics”, as far as natural language is concerned: there is no “semantics” for natural language (and for indexical languages more generally), but only two fields of study: syntax and pragmatics. 2. Meaning and speech acts Can we save the semantics/pragmatics distinction? Yes: we can give up Carnap‟s definition of semantics as the study of words-world relations, and define it instead as the study of the conventional, linguistic meaning of expression-types. According to Jerrold Katz, who made that view explicit, “Pragmatic phenomena [are] those in which knowledge of the setting or context of an utterance plays a role in how utterances are understood”; in contrast, semantics deals with “what an ideal speaker would know about the meaning of a sentence when no information is available about its context” (Katz 1977: 14). This view has been, and still is, very influential. Semantics thus understood does not (fully) determine words-world relation, but it constrains them (Katz 1975: 115-16). Because of indexicality and related phenomena, purely linguistic knowledge is insufficient to determine the truth conditions of an utterance. That much is commonly accepted. What semantics assigns to expression-types, independent of context, is not a fully-fledged content but a linguistic meaning or CHARACTER that can be formally represented as a function from contexts to contents (Kaplan 1989a: Stalnaker 1999, part 1). Thus the meaning of the pronoun I is the rule that, in context, an occurrence of I refers to the producer of that occurrence. 9 Insofar as their character or linguistic meaning can be described as a rule of use, indexical expressions are not as different as we may have thought from those expressions whose meaning is purely “pragmatic”. What is the meaning of, say, the imperative mood? Arguably, the sentences “You will go to the store tomorrow at 8”, “Will you go to the store tomorrow at 8?”, and “Go to the store tomorrow at 8” all have the same descriptive content. The difference between them is pragmatic: it relates to the type of illocutionary act performed by the utterance. Thus the imperative mood indicates that the speaker, in uttering the sentence, performs an illocutionary act of a “directive” type. To account for this non-truth-conditional indication we can posit a rule to the effect that the imperative mood is to be used only if one is performing a directive type of illocutionary act. This rule gives conditions of use for the imperative mood. By virtue of this rule, a particular token of the imperative mood in an utterance u “indicates” that a directive type of speech act is being performed by u. This reflexive indication conveyed by the token follows from the conditions of use which govern the type. The same sort of USE-CONDITIONAL analysis can be provided for e.g. discourse particles such as well, still, after all, anyway, therefore, alas, oh, and so forth, whose meaning is pragmatic rather than truth-conditional. There still is a difference between indexical expressions and fully pragmatic expressions such as the imperative mood. In both cases the meaning of the expressiontype is best construed as a rule of use. Thus I is to be used to refer to the speaker, just as the imperative mood is to be used to perform a certain type of speech act. By virtue of the rule in question, a use u of I reflexively indicates that it refers to the speaker of u, just as a use u of the imperative mood indicates that the utterer of u is performing a directive type of speech act. But in the case of I the token does not merely convey that reflexive indication: it also contributes its referent to the utterance‟s truth-conditional content. In contrast, the imperative mood does not contribute to the truth-conditional (or, more generally, descriptive) content of the utterances in which it occurs. In general, pragmatic expressions do not contribute to the determination of the content of the 10 utterance, but to the determination of its “force” or of other aspects of utterance meaning external to descriptive content. It turns out that there are (at least) three different types of expression. Some expressions have a purely denotative meaning: their meaning is a worldly entity which they denote. For example, square denotes the property of being square. Other expressions, such as the imperative mood, have a purely pragmatic meaning. They have conditions of use but make no contribution to content. Finally, there are expressions which, like indexicals, have conditions of use but contribute to truth conditions nevertheless. (The expression-type has conditions of use; the expression-token contributes to truth conditions.) This diversity can be overcome and some unification achieved. First, we can generalize the content/character distinction even to nonindexical expressions like square. We can say that every linguistic expression is endowed with a character that contextually determines its content. Nonindexical expressions will be handled as a particular case: the case in which the character is “stable” and determines the same content in every context. Second, every expression, whether or not it contributes to truth-conditional content, can be construed as doing basically the same thing—namely, helping the hearer to understand which speech act is performed by an utterance of the sentence. A speech act typically consists of two major components: a content and a force (Searle 1969). Some elements in the sentence indicate the force of the speech act which the sentence can be used to perform, while other elements give indications concerning the content of the speech act. Unification of the two sorts of elements is therefore achieved by equating the meaning of a sentence with its speech act potential. On the view we end up with — the speech-act theoretic view — semantics deals with the conventional meaning of expressions, the conventional meaning of expressions is their contribution to the meaning of the sentences in which they occur, and the meaning of sentences is their speech act potential. Pragmatics studies speech acts, and semantics maps sentences onto the type of speech act they are designed to perform. It follows that there are two basic disciplines in the study of language: syntax and 11 pragmatics. Semantics connects them by assigning speech act potentials to well-formed sentences, hence it presupposes both syntax and pragmatics. In contrast to the Carnapian view, according to which semantics presupposes only syntax, on the speechact theoretic view semantics is not autonomous with respect to pragmatics: There is no way to account for the meaning of a sentence without considering its role in communication, since the two are essentially connected... Syntax can be studied as a formal system independent of its use (...), but as soon as we attempt to account for meaning, for semantic competence, such a purely formalistic approach breaks down, because it cannot account for the fact that semantic competence is mostly a matter of knowing how to talk, i.e. how to perform speech acts. (Searle, “Chomsky‟s Revolution in Linguistics”, cited in Katz 1977: 26) There are other possible views, however. We can construe semantics as an autonomous discipline which maps sentences to the type of thought they express or the type of state of affairs they describe. That mapping is independent from the fact that sentences are used to perform speech acts. Note that communication is not the only possible use we can make of language; we can also use language in reasoning, for example. Be that as it may, whoever utters a given sentence — for whatever purposes — expresses a thought or describes a state of affairs, in virtue of the semantics of the sentence (and the context).5 Let us assume that the sentence is uttered in a situation of communication. Depending on the audience-directed intentions that motivate the overt expression of the thought or the overt description of a state of affairs in that situation, different speech acts will be performed. Those audience-directed intentions determine the force of the speech act, while the thought expressed by the sentence or the state of affairs it describes determines its content. In this framework, pragmatic indicators like the imperative mood can be construed as conventional ways of making the relevant 5 Because of indexicality, a sentence expresses a complete thought or describes a complete state of affairs only with respect to a particular context; but that determination is independent from issues concerning illocutionary force. 12 audience-directed intentions manifest. Their meaning, in contrast to the meaning of ordinary words like square, will be inseparable from the speech act the sentence can be used to perform. It is therefore not the overall meaning of the sentence which must be equated with its speech act potential. The major part of linguistic meaning maps linguistic forms to conceptual representations in the mind or to things in the world in total independence from communication. It is only a small subset of linguistic expressions, namely the pragmatic indicators and other expressions (including indexicals) endowed with use-conditional meaning, whose semantics is essentially connected with their communicative function. Whichever theory we accept, semantics (the study of linguistic meaning) and pragmatics (the study of language use) overlap to some extent (see Figure 1). That overlap is limited for the theories of the second type: the meaning of a restricted class of expressions consists in conditions of use and therefore must be dealt with both in the theory of use and the theory of meaning. According to the first type of theory, there is more than partial overlap: every expression has a use-conditional meaning. Since, for that type of theory, the meaning of a sentence is its speech act potential, semantics is best construed as a sub-part of speech act theory. pragmat ics (t heory of use) semantics (t heory of use potential) semantics (t heory of meaning) pragmat ics (t heory of use) pragmat ic meaning of indicators