Conventional Implicature and Expressive Content
Total Page:16
File Type:pdf, Size:1020Kb
To appear in Claudia Maienborn, Klaus von Heusinger, & Paul Portner, eds., Semantics: An International Handbook of Natural Language Meaning November 12, 2008 106. Conventional implicature and expressive content Christopher Potts UMass Amherst Abstract 1 Introduction This article presents evidence that individ- Natural language meanings are multifaceted. ual words and phrases can contribute mul- Even the simplest words, phrases, and sentences tiple independent pieces of meaning si- can, when uttered, convey a variety of distinct multaneously. Such multidimensionality is messages. Some derive solely from the con- a unifying theme of the literature on con- ventions of language, others from rich interac- ventional implicatures and expressives. I tions between language and context. Important use phenomena from discourse, semantic examples include presuppositions, conversational composition, and morphosyntax to detect implicatures, conventional implicatures, connota- and explore various dimensions of mean- tions, and at-issue (truth-conditional, entailed) ing. I also argue that, while the mean- content, as well as blends of these. Many of the ings involved are semantically indepen- central issues of semantic and pragmatic theory dent, they interact pragmatically to reduce revolve around how to manage this complex net- underspecification and fuel pragmatic en- work of interdependent meanings. richment. In this article, the central case The present article focuses on secondary mean- studies are appositives like Falk, the CEO, ings that (i) derive from the conventions of lan- and the taboo intensive damn, though dis- guage, albeit with extreme context dependency course particles and connectives like but, in many cases, and (ii) are semantically separate even, and still play supporting roles. The from the at-issue content but interact with it prag- primary evidence, both quantitative and matically. Appositives and expressives typify this qualitative, is drawn from large interview multidimensionality: and product-review corpora, which harbor (1) a. Charlie, an infamous axe murderer, is at a wealth of information about the impor- the door! tance of these items to discourse. b. Charlie is at the door. 1 Introduction1 (2) a. The damn dog is on the couch. b. The dog is on the couch. 2 Dimensions of meaning2 2.1 Dimensions in discourse.......4 These sentences are information-rich even with- 2.2 Dimensions in semantic composition5 out contextualization; if uttered, they convey even 2.3 Morphosyntactic parochialism...8 more. My focus is on the meanings that we can 2.4 Summary of findings.........9 trace, in whole or in part, to the highlighted (ital- icized) elements. 3 Pragmatic enrichment 10 For example, both (1a) and (1b) convey that 3.1 Nominal appositives......... 10 Charlie is at the door. However, the appositive in 3.2 The taboo intensive damn ...... 11 (1a) contributes a second meaning, by ascribing 3.3 Unifying themes............ 13 the property of being an infamous axe murderer to Charlie. These two meanings are, in a sense to 4 Conclusion 14 be made clear below, independent of one another, 5 References 14 but they interact pragmatically. In this case, each 1 Christopher Potts 2 supports the other’s relevance to guide us towards (5) Even Bart passed the test. the speaker’s intended message. The interaction is a. At-issue = Bart passed the test quite different if we replace this appositive clause b.CI Bart was among the least likely to with the pizza delivery guy. Thus, we’d like a se- pass≈ mantic theory that allows this sentence to denote two propositions, and we’d like a pragmatic the- The CI paraphrases are very rough, as indicated ory that explains how those propositions interact by the approximation signs. A recurrent theme to produce a pragmatically-enriched message. of CI meanings is that they are hard to specify in Something similar happens in (2). Whereas propositional terms. I return to this in section3. (2b) can be a neutral report, (2a) encodes In the next section, I take a closer look at the charged information about the speaker’s emo- semantic multidimensionality of these examples, tional state. The nature of this contribution is con- providing diagnostics for identifying secondary di- text dependent and challenging to specify, but it mensions of meaning and isolating them compo- nonetheless leaps out, helping us to understand sitionally and pragmatically. Following that, I ad- why the speaker is offering the content of (2b) dress how CIs feed pragmatic enrichment. The at this point in the conversation. Once again, we overall picture reconciles the lexical and construc- have semantic independence — we can identify tional origins of CIs with the usual assumption (2b) in (2a), both semantically and morphosyn- that they belong, in some sense, to pragmatics. tactically — and once again we have rich prag- matic interactions between the two meanings. 2 Dimensions of meaning Grice(1975) sketched the notion of conven- Grice(1975) calls upon multiple dimensions of tional implicature (CI) for roughly this class of meaning to resolve conflicting intuitions about phenomena, and Bach(1999), Neale(1999), and speaker commitments. The definition proceeds by Horn(2007) find the seeds of that classification way of example: in Frege’s writings. Both Frege and Grice used ex- If I say (smugly), He is an Englishman; he pressions like these to probe the limits of their the- is, therefore, brave, I have certainly com- ories of meaning. Ever since, the study of CIs has mitted myself, by virtue of the mean- branched off in numerous directions. The result- ing of my words, to its being the case ing picture appears fragmented; as Horn(2007) that his being brave is a consequence of says, CIs have had “a long and sometimes difficult (follows from) his being an Englishman. history” (p. 39). I’ve argued, though, that multidi- But while I have said that he is an En- mensionality of the sort seen in (1)–(2) unites this glishman and said that he is brave, I do research (Potts 2007b). Here, I argue for a unify- not want to say that I have said (in the ing pragmatic concept as well: CI items are pri- favored sense) that it follows from his marily devices for situating the main clause in the being an Englishman that he is brave, web of information that comprises the discourse. though I have certainly indicated, and so This seems a fitting characterization not only of implicated, that this is so. the above examples, but also of items more stan- dardly regarded as contributing CIs, as in (3)–(5). On Grice’s proposal, the conventional implicature (3) Alfie is a baby, but he is quiet. is the proposition denoted by ‘its being the case that his being brave is a consequence of (fol- a. At-issue = Alfie is a baby, and he is lows from) his being an Englishman’, and the at- quiet issue content (‘what is said’) is the proposition b.CI Babies are not usually quiet denoted by the conjunction ‘he is an Englishman ≈ (4) Isak is still swimming. and brave’. One sentence, two propositions. One might dispute whether Grice’s analysis of there- a. At-issue = Isak is swimming fore is correct, but the logical and linguistic idea b.CI Isak was swimming earlier is compelling. ≈ Christopher Potts 3 Karttunen & Peters(1979) brought this idea to the rest corresponding to secondary meanings of life by fitting it into a standard model-theoretic whatever kind we discover. package. The essence of their idea is captured by However, the phenomena in question are com- the four truth-value combinations in (6). positional: they trace to particular words and con- structions. This is central to Karttunen & Peters’ (6) T, T F, T theory. For them, not only sentences, but also h i h i T, F F, F individual words and phrases, can have multidi- h i h i mensional meanings. Pursuing this idea, we can Suppose we treat the first value in each pair as identify Falk, the CEO of Acme Products as a phrase modeling at-issue content and the second as mod- that has two meaning components. Its first dimen- eling CI content. Then we have a nuanced system sion picks out Falk, and is thus indistinguishable that includes absolute truth (upper left), absolute from the unadorned proper name Falk. Its sec- falsity (lower right) and blends of the two with an ond dimension is the proposition that Falk is the intermediate status. Appositives provide an easy CEO of Acme Products. Since appositives can af- illustration of the promise of this idea: fix to sentences (It’s raining, which is unexpected), (7) Falk, the CEO of Acme Products, gave the verb phrases (Joan jogs, which Jed does too), and keynote address. a host of other constituents, it looks like the space n of meanings is at least as broad as M }(W) , Here, we have two propositions expressed. Let’s the set of all pairs in which the first coordinate× is connect them with the meaning tuples in (6) by a member of the set M of all meanings (whatever assuming that the at-issue dimension (that Falk that space is like) and the rest of the coordinates gave the keynote address) is the first coordinate, are propositional. with the appositive content (that Falk is the CEO I think we want to generalize even more than of Acme) given by the second coordinate. If both that to deal with expressive content. Here is propositions are true, the value is T, T . If Falk Kaplan(1999) drawing a distinction that bears merely consults for Acme, but heh didi give the a family resemblance to Grice’s above, but that keynote, then the value is T, F . And so forth. more directly links semantics and pragmatics: This seems to be very close toh Gricei ’s (1975) orig- When I think about my own understand- inal proposal; the quotation at the start of this sec- ing of the words and phrases of my na- tion continues with its analysis of therefore by say- tive language, I find that in some cases ing, “I do not want to say that my utterance of this I am inclined to say that I know what sentence would be, strictly speaking, false should they mean, and in other cases it seems the consequence in question fail to hold.” Presum- more natural to say that I know how to ably, it wouldn’t be, strictly speaking, true in this use them.