LINGUIST List 12.1678

Wed Jun 27 2001

Review: Levinson, Presumptive Meanings (2nd rev.)

Editor for this issue: Terence Langendoen <terrylinguistlist.org>

What follows is another discussion note contributed to our Book Discussion Forum. We expect these discussions to be informal and interactive; and the author of the book discussed is cordially invited to join in.

If you are interested in leading a book discussion, look for books announced on LINGUIST as "available for discussion." (This means that the publisher has sent us a review copy.) Then contact Simin Karimi at siminlinguistlist.org or Terry Langendoen at terrylinguistlist.org.

Directory

  • Mayumi Masuko, Review of Levinson, Presumptive Meanings

    Message 1: Review of Levinson, Presumptive Meanings

     Date: Wed, 27 Jun 2001 10:00:05 +0900
    From: Mayumi Masuko <mayumimn.waseda.ac.jp>
    Subject: Review of Levinson, Presumptive Meanings


    Levinson, Stephen C. (2000) Presumptive Meanings: The Theory of Generalized Conversational Implicature, MIT Press, paperback, xxiii+480 pp., A Bradford Book

    Mayumi Masuko, School of Commerce, Waseda University

    Since the first reviewer of this book, by Bert Bultinck (http://linguistlist.org/issues/12/12-305), provided an excellent summary, I do not include one here. My conclusion is practically the same as that of Bultinck, although my reasons naturally differ from his. In what follows, I shall try to present them as succinctly as possible.

    This is a fairly comprehensive survey of Gricean pragmatics. Being an unorthodox, Levinson provides summaries of alternative accounts, and thus the book functions as a sort of encyclopaedia which can be used as an extensive source for researchers (including postgraduate students) who wish to do further work in this area. While such an effort is admirable especially in pragmatics, the area I suspect has not progressed as much in the last twenty years or so as in syntax or formal semantics, there are some problems. I shall address three.

    One problem concerns Levinson's choice of examples. Firstly, as pointed out by Bultinck, he spends too much time on the numerals as the prime example of Generalized Conversational Implicatures (GCIs), even though he notes that they are "NOT the correct testbed for the whole theory of scalar implicature" (p.90: the emphasis italics in the original). Personally, I find the argument rather counter-intuitive, the examples rather convoluted and their repetition rather otiose. Even more unfortunate, however, is the inclusion of the ordered pair <bitch, dog> whose behaviour, according to Levinson, is analogous to that of Horn scales. "<thumb, finger> and <bitch, dog> form ordered pairs where the first member entails the second in a suitable sentence frame, and the use of the second member thus implicates that the speaker is not in a position to assert that the stronger expression holds (hence "I cut my finger" suggests the speaker knows he did not cut a thumb)" (p.103). It is not clear to me at all what this "suitable sentence frame" might be, but I would have thought that many people will avoid "bitch" and opt for "dog" in many utterances. Bultinck blames Levinson's theory for its "logic-inspired meanings"; as long as it only considers prepositional meanings, it will fail to provide a satisfactory account of either pragmatic inference or pronominal coreference.

    Another drawback is the age-old problem: the semantics- pragmatics distinction. Bultinck points out that although Levinson argues against the Fodorian modular approach (cf. Fodor 1983), he still maintains that "semantics and pragmatics are distinct domains, which operate on specific, domain-exclusive principles, even though they are "interleaved" (p.187)." Levinson claims at one point that pragmatics concerns itself with nonmonotonic reasoning whereas semantics concerns itself with monotonic; in other places, however, he states that "[p]ragmatics too encompasses more than one process: on the one hand, default inferences can be calculated on the fly, given fragments of semantic representation; on the other hand, further pragmatic inferences (of both a specific and default character) can be calculated given the results of semantic interpretation" (p.168). By a "specific character" he may mean particularized as opposed to generalized conversational implicature, but his phrasing does not make it clear. If he in fact means non-default, then the distinction becomes vacuous.

    This relates to another, potentially more critical, issue: lack of precision. Levinson criticizes Hirschberg's (1985) account of scalar implicatures for overgeneration: there do not seem to be any constraints in her theory and any groups of words can potentially form scales. His own account, however, seems also liable to overgenerate, for the lexicalization constraint that was originally proposed by Atlas and Levinson (1981) and is repeatedly mentioned in this book wouldn't do as the restrictive condition as he admits in a footnote (n.49, p.392).

    One recurring criticism of Relevance Theory, Levinson's main adversary , is that it has never been formalized and thus is not clear what it actually maintains. It can be inferred that Levinson will have formalized his theory, but I am not sure if that is the case. Despite his claim that GCIs are default interpretations, he rejects default logics as an appropriate formal tool, saying that "they may seem to have little to offer a general theory of implicature because they provide a restricted set of limited inference rules, incapable of modeling the open-ended, creative, and indefinite set of inferences that come under Grice's theory" (p.46). But then he dithers: "[h]owever, as an account of the inference of GCIs, just those implicature that seem to have the default property, they may indeed have something to offer" (p.46). So do not all implicatures "have the default property", then? Which ones do and which ones do not, we are not told, though it can be inferred that GCIs (and presumably conversational implicatures, though that is not clear as Levinson does not discuss them at all) do and particularized conversational implicatures (PCIs) don't. But then are PCIs non-default and hence monotonic inferences?

    Does that mean a theory of GCIs can be formalized by using one (or more) of default logics? Well, maybe but maybe not. Since Levinson announces that "perhaps this account of pragmatic inference as based on stable heuristics is all we need to give an account of this species of defeasible inferences. There is no need for a nonmonotonic logic to model GCIs, it might be claimed, because GCIs are simply instantiations of communicational heuristics and indeed prototype default inferences" (p. 54). He then cites Johnson-Laird's (1983) argument that formal deduction is not necessary to model properties of human deduction, and suggests that the same holds true with nonmonotonic reasoning by humans. Perhaps, but what exactly are "stable heuristics"? Are they so "stable" that they apply regularly? If so, how do they differ from plain rules or principles? He even alludes to the supposedly formal characteristic of these heuristics: they are "mutually assumed by sender and receiver, that can serve to multiply the coded information by a factor of, say, three, by licensing inferential enrichment of what is actually encoded by choice of a specific signal" (p.30). But how could he assume that the "heuristics" are shared and justify "a factor of, say, three"? As he provides no evidence for neither, I see no reason to accept either of his claims.

    Levinson suggests further that "rather than seek understanding of implicature in theories of default inference, perhaps we should instead think of generalized implicatures as the prototype of default inferences" (p.49). Perhaps, but are they? Since no clear evidence is provided to prove that indeed is the case, I see no reason to think that is so. And since he goes on to claim that "�c because the inferences are defeated by recognition of a goal and because THAT kind of defeasibility is precisely what practical reasoning systems are set up to capture perhaps these examples are indications that the plan-generation and plan- recognition types of nonmonotonic inferences are what are really involved in implicature generation" (pp.52-53). Perhaps, but perhaps not: we never know where he stands. This echoes Bultinck's observation that Grice's original account is falsifiable: Levinson's theory as it stands appears falsifiable because we do not really know exactly what claims it makes.

    The title of the book itself begs the question: 'presumptive meanings' on the basis of what? Levinson also talks about 'preferred interpretations' without explicating 'by whom'. 'Default interpretation' is the term also used interchangeably with these two, and again it is not clear when default does not hold. I suspect that the ceteris paribus part, which is inherent in any discussion of default and probably implicature in Levinson's theory, is unproblematic in morphology but not so in pragmatics. That is precisely why the AI local pragmatics (cf. Hobbs et al. 1990), which Levinson disputes, explicitly declares in the system what assumptions are made and shared between the speaker and the addressee: what counts as mutual knowledge or common ground depends at least on the identity of the speaker and the addressee and context.

    To recap, this book is an encyclopedic survey of a certain kind of pragmatic inference that the author argues shows clear evidence for the existence of utterance-type meanings: it is an excellent source book from which the interested reader can obtain data and references to work on. Nevertheless, in view of issues that remain to be solved, it may be appropriate to end this review by citing Wilks (1986:276), which was an argument against Relevance Theory. This is because although Levinson (pp.26-27) denies the charge, since he never explicates the ceteris paribus part of the theory and just assumes that it is shared between the speaker and the addressee, I suspect it applies to his theory as well. I hope it has become clear by now that though my conclusion is virtually the same as Bultinck's, I do not accept his claim that "meanings can be more or less conventional, or more or less context-independent, not semantic or pragmatic". Linguistic meanings are conventionalized in the sense that they are conveyed by certain grammatical constructions or lexical items but they are largely context-dependent. And that is precisely why accounts of linguistic meaning that are free of any discoursal and/or contextual issues will be deemed unsatisfactory.

    "[Sperber and Wilson's (1982)] errors all stem from their conviction that there is an objectively "right" context set of propositions, one that can be assumed independently of how it is located, and independently of what individuals may in fact believe."

    A bibliographical note

    I would not recommend this book to undergraduates who do not know much about the background. Levinson (1983) provides the minimum required to understand the issues discussed, though it would be desirable to read Grice (1989) and Gazdar (1979), which inspired Levinson (2000). I heard, however, the latter is out of print, and if it cannot be found in libraries, Levinson (1983) or Soames (1982) provides the necessary background information.

    For those who find 480 pages too long, Levinson (1987a and 1987b) will give the gist of the discussion and enough examples to understand his arguments. As for its main antagonist, Relevance Theory, Sperber and Wilson (1986), or the second edition of the book, which I have not read, still remains the main source. For those who prefer shorter accounts, read Sperber and Wilson (1982), which I believe gives the gist of the argument. For its critical analysis, there's Gazdar and Good (1982), to whose argument, as far as I am aware, neither Sperber nor Wilson has answered. If neither of these proves inaccessible, read Wilks (1986).

    References

    Atlas, Jay and Stephen C. Levinson (1981) "It-clefts, informativeness, and logical form: Radical pragmatics (revised standard version)", in P. Cole (ed.), Radical Pragmatics, Academic Press, 1-61.

    Fodor, Jerry (1983) The Modularity of the Mind, MIT Press.

    Gazdar, Gerald (1979) Pragmatics: Implicature, Presupposition, and Logical Form, Academic Press.

    Gazdar, Gerald and David Good (1982) "On a notion of relevance: Comments on Sperber and Wilson's paper". In Smith (1982:88-100).

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

    Hirschberg, Julia (1985) A Theory of Scalar Implicature. PhD dissertation, University of Pennsylvania.

    Hobbs, Jerry, M. Stickel, D. Appelt and P. Martin (1990) Interpretations as Abduction. Technical Note 499, SRI International.

    Johnson-Laird, P. N. (1983) Mental Models: Towards a Cognitive Science of Language, Inference and Consciousness, Cambridge University Press.

    Levinson, Stephen C. (1983) Pragmatics, Cambridge University Press.

    Levinson, Stephen C. (1987a) "Minimization and conversational inference". In J. Verschueren and M. Bertuccelli-Papi (eds.), The Pragmatic Perspective, John Benjamins, 61-129.

    Levinson, Stephen C. (1987b) "Pragmatics and the grammar of anaphora", Journal of Linguistics 23:379-434.

    Smith, Neil (ed.) (1982) Mutual Knowledge, Academic Press.

    Soames, Scott (1982) "How presuppositions are inherited: A solution to the projection problem", Linguistic Inquiry 13:483-545.

    Sperber, Dan and Deirdre Wilson (1982) "Mutual relevance and knowledge in theories of comprehension". In Smith (1982:).

    Sperber, Dan and Deirdre Wilson (1986) Relevance, Blackwell.

    Wilks, Yorick (1986) "Relevance and beliefs". In Myers, Terry, Keith Brown and Brendan McGonigle (eds.), Reasoning and Discourse Processes, Academic Press, 265-289.

    About the reviewer: Mayumi Masuko did her postgraduate studies at the University of Cambridge, where she received an MPhil and a PhD in linguistics. She is an Associated Professor of English at Waseda University, where she teaches English and linguistics. Her main research interest lies in the interaction between semantics (broadly conceived) and morphosyntax.