[in J. Keim Campbell, M. O'Rourke, and D. Shier, eds., Meaning and Truth, New York: Seven Bridges Press (2002), pp. 284-292]


Kent Bach


The distinction between semantics and pragmatics has received a lot of bad press in recent years. It has been claimed to be faulty, confused, or even nonexistent. However, these claims are based on misconceptions of what the distinction is and of what it takes to show there to be something wrong with it. As I see it, the semantic-pragmatic distinction fundamentally concerns two types of information associated with an utterance of a sentence. Semantic information is encoded in the sentence; pragmatic information is generated by, or at least made relevant by, the act of uttering the sentence. This explains the oddity of such pragmatic contradictions as "I am not speaking" and "It is raining but I don't believe it." In "The Semantics-Pragmatics Distinction: What It Is and Why It Matters" (Bach 1999a), I develop this conception of the distinction and contrast it with alternatives. Here I will to try clarify that conception by showing how it avoids certain objections. Space will not permit going into much detail on the various linguistic data and theoretical considerations that have been thought to undermine the semantic-pragmatic distinction in one way or another.

        Historically, this distinction has been formulated in various ways. These formulations have fallen into three main types, depending on which other distinction the semantic-pragmatic distinction was thought to correspond to:

        +     linguistic (conventional) meaning vs. use

        +     truth-conditional vs. non-truth-conditional meaning

        +     context independence vs. context dependence

None of these distinctions does the job. The trouble with the first one is that there are expressions whose literal meanings are related to use. The second distinction is unhelpful because some expressions have meanings that do not contribute to truth-conditional contents. And the third distinction overlooks the fact that there are two kinds of context. This last point deserves elaboration.

        It is a platitude that what a sentence means generally doesn't determine what a speaker means in uttering it. The gap between linguistic meaning and speaker meaning is said to be filled by "context": what the speaker means somehow "depends on context," or at least "context makes it clear" what the speaker means. But there are two quite different sorts of context, and they play quite different roles. What might be called "wide context" concerns any contextual information that is relevant to determining (in the sense of ascertaining) the speaker's intention. "Narrow context" concerns information specifically relevant to determining (in the sense of providing) the semantic values of context-sensitive expressions (and morphemes of tense and aspect). Wide context does not literally determine anything. It is the body of mutually evident information that the speaker exploits to make his communicative intention evident and that his audience relies upon, taking him to intend them to do so, to identify that intention.

        Another source of confusion is the phrase "utterance interpretation." Strictly speaking, sentences (and subsentential expressions), i.e. types not tokens, have semantic properties. Utterances of sentences have pragmatic properties. Also, the term "interpretation" is ambiguous. It can mean either the formal, compositional determination by the grammar of a language of the meaning of a sentence or the psychological process whereby a person understands a sentence or an utterance of a sentence. Using the phrase "utterance interpretation" indiscriminately for both tends to confound the issues.

        My conception of the semantic-pragmatic distinction involves certain asssumptions about semantics and a certain view of communication. I take the semantics of a sentence to be a projection of its syntax. That is, semantic structure is interpreted syntactic structure. Contents of sentences are determined compositionally; they are a function of the contents of the sentence's constituents and their syntactic relations. This leaves open the possibility that some sentences do not express complete propositions and that some sentences are typically used to convey something more specific than what is predictable from their compositionally determined contents. Also, insofar as sentences are tensed and contain indexicals, their semantic contents are relative to contexts (in the narrow sense). Accordingly, the following distinctions should be recognized:

+         between a sentence and an utterance of a sentence

+         between what a sentence means and what it is used to communicate

+         between what a sentence expresses relative to a context and what a speaker expresses (communicates) by uttering the sentence in a context

+         between the grammatical determination of what a sentence means and the speaker's inferential determination of what a speaker means (in uttering the sentence)

As for communication, when a speaker utters a sentence in order to convey something, the content of the sentence provides the basis for his audience's inference to what he is conveying and what attitudes he is expressing, e.g., belief in the case of assertion and desire in the case of requesting. In fact, as Bach and Harnish (1979, ch. 3) argue, because types of communicative speech acts may be individuated by the types of attitudes they express, their contents are simply the contents of the attitudes they express. That is one reason why the notion of the content of an utterance of a sentence has no independent theoretical significance. There is just the content of the sentence the speaker is uttering, which, being semantic, is independent of the speaker's communicative intention, and the content of the speaker's communicative intention. When one hears an utterance, one needs to understand the sentence the speaker is uttering in order to figure out the communicative intention with which he is uttering it, but understanding the sentence is independent of context except insofar as there are elements in the sentence whose semantic value are context-relative. Recognizing the speaker's communicative intention is a matter of figuring out the content of that intention on the basis of contextual information in the broad sense.

        This information does not literally determine that content. In no case does the semantic content of the uttered sentence determine what the speaker is communicating or, indeed, that he is communicating anything. That he is attempting to communicate something, and what that is, is a matter of his communicative intention, if he has one. If he is speaking literally and means precisely what his words mean, even that is a matter of his communicative intention. Communicative intentions are reflexive in the sense discovered by Grice: a communicative intention is one whose fulfillment consists in its recognition by the audience, partly on the basis that it is intended to be recognized. The role of Grice's maxims, or presumptions as they might better be regarded (Bach and Harnish 1979, pp. 62-65), is to provide inference routes across any gap between what the sentence means and what the speaker aims to be communicating in uttering it.

        This Gricean view of linguistic communication (it is developed in detail in Bach and Harnish 1979) lends itself to a certain conception of the semantic-pragmatic distinction. This distinction can be drawn with respect to various items, such as ambiguities, contradictions, implications, presuppositions, interpretations, knowledge, processes, rules, and principles, and, of course, "semantics" and "pragmatics" are also names for the study of these phenomena. For me the distinction applies fundamentally to types of information. Semantic information is information encoded in what is utteredóstable linguistic features of the sentenceótogether with any extralinguistic information that contributes to the determination of the references of context-sensitive expressions. Pragmatic information is (extralinguistic) information that arises from an actual act of utterance, and is relevant to the hearer's determination of what the speaker is communicating.

        This way of characterizing pragmatic information generalizes Grice's point that what a speaker implicates in saying what he says is carried not by what he says but by his saying it and sometimes by his saying it in a certain way (1989, p. 39). The act of producing the utterance exploits the information encoded but by its very performance creates new and otherwise invokes extralinguistic information. This extralinguistic information includes the fact that the speaker uttered that sentence and did so under certain mutually evident circumstances. This is context in the broad sense. Importantly, nonsemantic information is relevant to the hearer's inference to the speaker's intention only insofar as it can reasonably be taken as intended to be taken into account, and that requires the supposition that the speaker is producing the utterance with the intention that it be taken into account. There is no such constraint on contextual information of the semantic kind, which plays its role independently of the speaker's communicative intention. Contextual information in the narrow, semantic sense is limited to a short list of parameters associated with indexicals and tense, such as the identity of the speaker and the hearer and the time of an utterance. I may think I'm Babe Ruth and be convinced that it's 1928, but if I say, "I hit 60 home runs last year," I am still using "I" to refer to myself and "last year" to refer to the year 1999.

        Now let us consider some reasons that might be suggested for rejecting the semantic-pragmatic distinction. To the extent that the debate about it isn't entirely terminological (e.g., many years ago "pragmatics" was the name for indexical semantics), the main substantive matter of dispute is whether there is such a thing as "pragmatic intrusion," whereby pragmatic factors allegedly contribute to semantic interpretation. Here is an assortment of objections that are based on supposed pragmatic intrusion of one sort or another. Each of these objections is predicated on some misconception, as the responses indicate.

1. Semantic phenomena are context-independent, whereas pragmatic phenomena are context-sensitive. But the meanings of certain expressions are context-sensitive. Therefore, their meanings are not exclusively semantic.

This objection assumes that anything pertaining to the use of an expression is automatically not semantic. However, the fact that the contents of certain expressions, notably indexicals and demonstratives, are context-sensitive does not show that their meanings vary with context. How their contents vary with context is determined by their fixed meanings, and that is a semantic matter. These variable contents are their semantic values.

2. There are aspects of linguistic meaning that concern how a sentence is used, not its truth-conditional content. So linguistic meaning is not merely a semantic matter.

This objection alludes to the fact that the meanings of certain expressions, what I call "utterance modifiers," such as "to conclude," "frankly," and "to be precise" (for a catalog of them see Bach 1999b, sec. 5), as well as grammatical mood, concern how a sentence is being used. However, all this shows is that semantics is not limited to what is relevant to truth-conditional content. There is no reason to assume that the linguistic meaning of a sentence cannot include information pertaining to how the sentence is used.

3. Since language is rife with semantic underdetermination and vagueness, there is no such thing as literal meaning: sentence "semantics" is adulterated with pragmatics.

These phenomena show only that sometimes the literal meaning of a sentence does not determine a complete proposition or a precise proposition. They do not show that there is no purely linguistic information on which language users rely. Take the case of semantically underdeterminate sentences, which do not express complete propositions, even modulo ambiguity and indexicality. Even though the following sentences do not express complete propositions,

                (1)        Muggsy is too short/isn't tall enough.

                (2)     a. Kurt finished the picture.

                          b. Kurt finished the book draft.

                          c. Kurt finished the newspaper.

they still have determinate semantic contents. However, these are not complete propositions. The semantics of (1) does not specify what Muggsy is too short or not tall enough for, and the semantics of the sentences in (2) do not specify whether Kurt finished painting, writing, reading or, for that matter, eating. However, what the speaker means must include some such thing. So the completion of what the speaker means involves the insertion of something that does not correspond to any constituent of the sentence. This does not show that there is something wrong with the semantic-pragmatic distinction but only that utterancs of semantically incomplete sentences require pragmatic supplementation.

4. There are many sentences whose typical use is not what, according to the compositional semantics of the sentence, the sentence means. Therefore, pragmatic information somehow blends into semantic information.

This objection is illustrated by likely utterances of (3) and (4).

                (3)        Jack and Jill went up the hill.

                (4)        Jack and Jill are married.

(3) is likely to be used to assert that Jack and Jill went up the hill together and (4) that they are married to each other, even though this is not predictable from the meanings of the sentences. Nothing adverse to the semantic-pragmatic distinction follows from this, however. These examples show merely that some sentences are typically not used to mean what the sentences themselves mean. This is clear from the fact that the analogous uses are not typical for sentences like (3') and (4'),

                (3')        Jack and Jill went up the hill separately/on different days.

                (4')        Jack and and his sister Jill are married.

5. There are certain expressions that are generally not used strictly and literally, such as "empty," "everybody," and "circular". Therefore, their semantics does not determine how they are standardly used and pragmatics enters in.

This is also true, but the existence of a distinction between semantics and pragmatics does not imply or even suggest that expressions must standardly be used literally. There may be a presumption of literality, but this presumption can easily be overridden, especially with words like the ones above.

6. There are certain expressions that have a range of related meanings but are neither clearly ambiguous nor clearly unambiguous. What such an expression can be used to mean is always partly a pragmatic matter.

This objection is based on examples like these:

                (5)     a. Gus went from Natchez to New Orleans.

                          b. The road went from Natchez to New Orleans.

                          c. The show went from 7 to ll.

                          d. Gus went from irritated to outraged.

                          e. The house went from Gus to his wife.

The idea behind the objection is that as they occur in these sentences the words "go," "from," and "to," though semantically univocal, have distinct but related meanings; that is, rather than being ambiguous their unitary linguistic meanings underdetermine what they are used to mean "in context," hence that their pragmatics intrudes on their semantics.

        This idea is a definite improvement over the view that the words "go," "from," and "to" are used literally only in (5a), which involves movement from one place to another, and that their uses in the other sentences are in various ways "extended," hence nonliteral uses. However, it does not follow that pragmatics intrudes on semantics. The existence of these various uses shows merely that the meanings of such polysemous terms are more abstract than the movement model would suggest (for further discussion see Bach 1994, sec. 7).

        Nunberg (1979) offers a related objection, based on the multiplicity of uses of terms like "chair" and "newspaper." "Chair" can refer to particular chairs (chair tokens) or to chair types. But it is far from clear why this instance of the general type-token ambiguity poses a problem for the semantic-pragmatic distinction. The case of "newspaper" is more interesting, because that term can refer either to particular copies of a newspaper, to specific issues or editions of a newspaper, e.g., the final edition of today's New York Times,  or to the publishing company. Nunberg claims that there is no basis for singling out one use as the conventional one and treating the others as derived from that. But surely the last use is a derived use, since the publishing company wouldn't be referred to as the newspaper (e.g., the San Francisco Chronicle, which was recently bought by the Hearst Corporation). Indeed, it is arguable that this use of the term is elliptical for "newspaper publishing company." In any case, how to explain polysemy has no particular bearing on the semantic-pragmatic distinction, but is rather a problem in lexical semantics.

7. There are certain complex expressions whose meanings are not predictable from the meanings of their constituents. Therefore, pragmatics impinges upon semantics.

It is true that the meanings of phrases and compounds like the following are not predictable, or at least not obviously predictable, from the meanings of their constituents:

                (6)     a. sad girl

                          b. sad face

                          c. sad day

                          d. sad music

                (7)     a. child abuse

                          b. drug abuse

                (8)     a. election nullification

                          b. jury nullification

                (9)     a. slalom skiing

                          b. snow skiing

                          c. heliocopter skiing

                (10)   a. jellyfish, goldfish, catfish

                          b. goldfish

                          c. catfish

                (11)   a. clipboard

                          b. diving board

                          c. bread board

                          d. game board

                          e. headboard

As interesting as these examples are, they do not undermine the semantic-pragmatic distinction . All they show is that phrasal semantics is not straightforward. They suggest that compositionality is not a simple as it might seem, for there are different ways in which the meanings of words can combine. But this has nothing to do with pragmatics.

8. A well-defined semantic-pragmatic distinction requires that semantics determine what is said. But intuitions about what is said indicate that it includes pragmatically determined elements.

These "intuitions" are not necessarily about what is said (see Bach 1994 and "Seemingly Semantic Intuitions," this volume). They arise when one ignores the distinction between locutionary and illocutionary acts or assumes that the distinction between what is explicit in an utterance and what is implicated by it is exhaustive. Intuitions that are adverse to the semantic-pragmatic distinction are insensitive to the in-between category of what is implicit in what is said (more accurately, in the saying of it), and mistakenly include that in what is said. The same mistake is implicit in relevance theorists' use of the term "explicature" for aspects of utterance content that are not explicit and in the description of the entire content as "explicit."

9. The strict and literal semantic content of a sentence is often not calculated prior to the hearer's determination of the content of a speaker's utterance. Therefore, even if there is a theoretical role to the notion of semantic content, it has no psychological import, hence no empirical significance.

Facts about "pragmatic processing" are not relevant to the semantic-pragmatic distinction. They are often cited in support of claims about what is and what isn't said and even used to argue that there are pragmatic elements in what is said. However, nothing follows from such facts about what is or isn't said, since that's a matter of what a speaker does in uttering a sentence, not what his listeners do in understanding it. Moreover, the psychologically relevant category, if there is one, is information that is available to pragmatic processing, not what goes on in the processing itself.

        None of the above considerations or phenomena poses a serious objection to the viability of the semantic-pragmatic distinction. What they do show is that it is important to avoid a simplistic approach to that distinction. Semantics and pragmatics are both complex, but this doesn't mean that they should be mixed up, much less that they overlap.




Bach, Kent (1994), "Conversational Impliciture," Mind & Language 9: 124-62.

_____ (1999a), "The Semantics-Pragmatics Distinction: What It Is and Why It Matters," in Ken Turner (ed.), The Semantics-Pragmatics Interface from Different Points of View, Oxford: Elsevier, pp. 65-84; on line at http://userwww.sfsu.edu/~kbach/spd.htm.

_____ (1999b), "The Myth of Conventional Implicature," Linguistics and Philosophy 22: 327-366.

_____ and Robert M. Harnish (1979), Linguistic Communication and Speech Acts, Cambridge, MA: MIT Press.

Grice, Paul (1989), Studies in the Way of Words, Cambridge, MA: Harvard University Press.

Nunberg, Geoffrey (1979), "The Non-uniqueness of Semantic Solutions: Polysemy," Linguistics and Philosophy 3: 143-184.