A cognitive approach to inferencing in conversation

ELSEVIER Journal of Pragmatics 30 (1998) 755-769 A cognitive approach to inferencing in conversation Klaus -Uwe Panther ~,*, L inda Thornburg b " Englisches Seminar, Universitiit Hamburg, Von-Melle-Park 6, D-20146 Hamburg, Germany b Department of American Studies, E6tv6s Lor6nd University, Ajtfsi Diirer sol 19-21, H-1146 Budapest, Hungary Received 13 January 1997; revised version 14 January 1998 Abstract Discourse analysts using a speech-act theoretic or Gricean approach to conversation assume that the identification of the communicative intention of linguistic acts requires infer- encing on the part of the hearer. However, this theoretical approach does not satisfactorily account for the fact that conversational participants usually draw the inferences necessary to arrive at intended interpretations without any noticeable effort. Furthermore, it does not sys- tematically specify the kinds of inference patterns that are needed for utterance interpretation. We show that a cognitive approach incorporating the notion of scenario structure can be fruit- fully applied to the analysis of conversation. We argue that within in this framework it is nec- essary to distinguish between 'stand for' relations (i.e. metonymic relationships proper) and weaker 'point to' relations (i.e. indexical relationships). Conceptual relationships such as part-whole, cause-effect, ability-action, etc., have metonymic and indexical function and facilitate the inferential work of conversational interactants. These general relationships within scenario structure thus constitute natural inference schemata. 1. Introduction The problem of understanding discourse in terms of inferences about a speak- er 's /author 's intentions has a long-standing tradition in classical rhetoric and hermeneutics. Even today, discourse analysts who use speech-act theoretic, Gricean, or relevance-theoretic approaches to the study of conversation assume that the iden- tif ication of the communicat ive intention of l inguistic acts requires a certain amount of inferencing on the part of the hearer (for good summaries see Schiffrin, 1994; An earlier version of this paper was presented at the 5th IPrA Conference, Mexico City, July 13-17, 1997. We have benefited from the helpful comments of two anonymous referees in preparing this revi- sion. * E-mail Klaus-Uwe Panther: panther@rrz.uni-hamburg.de; e-mail Linda Thornburg: lthornburg@isis,elte.hu 037 8-2166/98/$ - see front matter © 1998 Elsevier Science B.V. All rights reserved. PII: S0378-2166(98)00028-9 756 K.U. Panther, L. Thornburg /.lournal of Pragmatics 30 (1998) 755-769 Sperber and Wilson, 1995). For example, Searle (1975), in his classical analysis of indirect speech acts, claims that the intended force of an indirect speech act has to be inferred on the basis of what is literally conveyed, the linguistic and extralinguistic context of the utterance, and Gricean principles (see Grice, 1975) of what constitutes rational and cooperative behavior. Others (e.g. Bach and Harnish, 1979; Leech, 1983) have claimed that even direct speech acts cannot be interpreted without recourse to inferencing strategies. Sperber and Wilson (1995) present a similar argu- ment with regard to what they call 'explicatures', i.e. devices for enriching the semantic form of an utterance. In contrast to the inferential approach it is sometimes claimed that conventional- ized indirect speech acts do not involve any kind of inferential work but can be understood on the basis of "simple rules of interpretation" and are "not likely to include a literal phase" (Ervin-Tripp, 1976: 25). We place ourselves within the long-standing tradition outlined above; i.e., we take the inferential approach to be a necessary component of an adequate theory of utter- ance meaning. However, the inferential theory alone is incomplete for two reasons: (1) it does not satisfactorily account for the fact that conversational participants often very quickly draw the inferences necessary to arrive at intended interpretations with- out any noticeable effort, and (2) there is no systematic account of the nature of the inference patterns themselves and their cognitive grounding. The Gricean approach fails on both accounts, since it states only a very general framework comprising a Cooperative Principle and various conversational maxims. Morgan (1978: 274), attempting to come to terms with problem (1) above, has pro- posed that 'short-circuited implicatures', facilitate rapid utterance comprehension (and may eventually lead to usage conventions). With regard to problem (2) above, Gordon and Lakoff (1975) have assumed general inference patterns called conversa- tional postulates; Fraser (1975) and Searle (1975) have postulated the existence of specific inference patterns for hedged performatives and for indirect speech acts, respectively. We hope to demonstrate that these earlier insights about conversational postulates and other inference patterns are based on conceptual metonymies operating within cognitive domains that we call 'scenarios'. In particular, we will show that such metonymic processes are at work in speech act scenarios and contribute to a better understanding of how conversation works. 2. A cognitive approach to metonymy In accordance with recent developments in cognitive semantics, we assume that meaning, including discourse meaning, is fruitfully represented in terms of e.g. frames (Fillmore, 1982), Idealized Cognitive Models (Lakoff, 1987), or what we will call scenarios. Scenarios consist of parts which can bear metonymic relations to each other and the whole of the scenario. It is important to emphasize at this point that we use the term 'metonymy' to include more than its common function of indi- rect referring (cf. Lakoff, 1987; Gibbs, 1994). Moreover, we consider metonymy to K.U. Panther, L. Thornburg / Journal of Pragmatics 30 (1998) 755-769 757 be a property of conceptual structure, i.e., it is a relation among concepts and not merely among words (see Lakoff and Johnson, 1980; Gibbs, 1994). Best-known cases of the traditional conception of metonymy are noun phrases used referentially, as in, for example: (1) the White House This is a noun phrase that is conventionally used to refer indirectly to the executive branch of the government of the United States, a spokesperson for that branch, or even the President himself, but is not synonymous with any of these. We claim, however, that the concept of metonymy applies not only to the refer- ential but to the predicational part of a proposition as well. Consider the utterance in (2a): (2) a. The Chicago Bulls were able to nail down their fifth NBA championship ... [CNN Sportscaster, June 1997] Utterance (2a) very strongly implicates the statement given in (2b): b. The Chicago Bulls nailed down their fifth NBA championship ... In (2a) the speaker literally predicates the past ability of the Chicago Bulls to nail down their fifth championship, whereas in (2b) the speaker would be predicating the past actuality of the Bulls winning the championship. We assume a metonymic rela- tion between the utterance in (2a) and its implicature in (2b). In the specific context given, (2a) expresses the same state of affairs as (2b) although they differ in their literal meanings and pragmatic effects. In this sense we take the predication in (2a) to stand for the predication in (2b) (just like the referential term the ham sandwich may stand for a customer in the jargon of restaurant staff). This metonymic relation is so highly conventionalized that it is easily overlooked, but it is nevertheless an instance of a very general metonymic principle that we call the POTENTIALITY FOR ACTUALITY metonymy (see Thornburg and Panther, 1997; Panther and Thornburg, forthcoming). Perhaps even less apparent is the operation of metonymic principles beyond the propositional level, i.e. at the level of illocutionary force. We call this third type of metonymy speech act (or illocutionary) metonymies. In this case one illocutionary act stands for, i.e. has the same illocutionary force as, another illocutionary act. Examples (3a-b) illustrate this concept: (3) a. I don't know where the bath soap is. (heard by the native English-speaking author and interpreted as (3b)) b. Where is the bath soap? l ] Note that in this case the propositional content also changes. 758 K.U. Panther, L. Thornburg / Jouinal of Pragmatics 30 (1998) 755-769 Utterance (3a) has the di rect i l locutionary force of an assertion about what the speaker does not know, but in many instances, as in the context above, it is used with the ind i rect i l locutionary force of a question. That is to say, the asser t ion in (3a) may stand for the quest ion expressed in (3b). The concept of 'speech act metonymy' requires the notion of 'speech act scenario' , which will be developed below in Section 3. To summarize, then, we have distinguished two basic types of metonymic rela- tionships: 2 (4) a. proposit ional metonymies: (i) referential 3 (ii) predicational b. i l locutionary metonymies We have already carried out more basic work in speech act and predicational metonymies in Thornburg and Panther (1997), using data from the Lancaster-Oslo- Bergen textual corpus of English. That research led to a cross-l inguistic study, Pan- ther and Thornburg (forthcoming), which contrasts the exploitation of a particular metonymy, the POTENTIALITY FOR ACTUALITY metonymy, in English and Hungarian. In our present paper we will i l lustrate our approach w i thdata from the The Bergen Corpus o f London Teenage Language (COLT) focusing on i i locutionary metonymies as they occur and guide participants in English conversations. 3. Speech act scenarios and metonymy We assume that speech acts and their felicity condit ions are best described as sce- narios, more specif ical ly as action scenarios. 4 An action scenario consists of at least the fol lowing components. First, a component that states the conditions that must be 2 These metonymies can also be combined as in: (i) I don't know whether the first violin was able to pass her driver's test. which contains all of the metonymic types mentioned above: - referential: the noun phrase thefirst violin may stand for the person in an orchestra playing the first violin; - predicational: the propositional content that The speaker does not know whether the first violin was able to pass her driver's test may stand for the propositional content that The speaker does not know whether the first violin passed her driver's test; - illocutionary: the assertion 1 don't know whether thefirst violin was able to pass her driver's test may stand for the question Was the first violin able to pass her driver's test? 3 In distinguishing between the referential and the predicational parts of propositional contents we fol- low Searle (1969). 4 Our approach is in contrast to Sperber and Wilson (1995: 243ff.), who argue against a speech act account of utterance meaning in favor of a relevance theoretic account, which recognizes only very gen- eral communicative types like saying, telling and asking plus the retrieval of propositional attitudes like believing and desiring. We assume that conversationalists activate fairly specific speech act concepts in orienting themselves in a discourse. K.U. Panther, L. Thornburg / Journal of Pragmatics 30 (1998) 755-769 759 fulfilled before the action proper can take place. We call this part the BEFORE com- ponent. Second, there is a CORE that describes the essential features of the action itself, and the immediate RESULT that obtains if the action is felicitously per- formed. Finally, there is an AFTER component, which describes the intended conse- quences of the action. These components correspond more or less to the felicity con- ditions which have been posited for speech acts. Furthermore, we assume that the components can, to varying degrees, be metonymically linked to the speech act sce- nario as a whole. Thus, the scenario is structured in such a way that the arrangement of the components reflects their conceptual distance from the CORE component. In (5) we give a simplified request scenario to illustrate the notion of 'conceptual distance': (5) Simplied scenario for requests (i) The BEFORE: H can do A. S wants H to do A. (ii) The CORE: S puts H under a (more or less strong) obligation to do A. The RESULT: H is under an obligation to do A (H must/should/ought to do A). H will do A. (iii) The AFTER: As is well-known (cf. Searle, 1975), the question (6) Will you close the door? is conventionally used in English to perform a request to close the door. In this sense a component of the illocutionary scenario request, (a question about) the future action of the addressee to close the door, stands for the request itself to close the door. In other words, utterance (6) is a pragmatic substitute for the explicit request Close the door. The reason for (6) being able to stand for a request is, in our frame- work, that the component H will perform A is part of the request scenario and con- ceptually close to the CORE of that scenario. On the other hand, an utterance like (7) What's that smell? in a situation where two people are sitting in a room with an open door is not conven- tionally used as (i.e., does not stand for) a request to close the door. In our terminol- ogy, the utterance could only point to a request. The conceptual content of utterance (7) is located outside the boundaries of the request scenario. However, (7) could, with some conceptual effort, be interpreted as a motivation or reason for the hearer to per- form some action, here to prevent the smell from entering the room by closing the door. In this sense, utterance (7) does not stand for but merely points to a request. Further support for our distinction between stand for and point to relations comes from distributional evidence: The request marker please can freely co-occur with (6) but not with (7): 760 K.U. Panther, L. Thornburg /Journal of Pragmatics 30 (1998) 755-769 (6) a. b. C. (7) a. b. C. Please, will you close the door? Will you please close the door? Will you close the door, please? #Please, what's that smell? ##What please is that smell? #What's that smell, please? We provide another and more elaborate representation of our scenario approach in Fig. 1 for a specific type of request that will be relevant to our ensuing analysis of conversational data. Distance Scale and Strength of Metonymic Link Presuppositional Branch BEFORE CORE RESULT AFTER Motivational Branch X exists/is available I ~ S's reasons for wanting X H has X ( " " "~._ . I S wants X A~ H can give X to S ¢ . ~ "l" ~ N ~ ~ s H to give X to S q~ S puts H under an obligation to give X to S $ H is under an obligation to give X to S I H will give X to S / S will have X I Other consequences Distance Scale and Stren2th of Metonvmic Link Realization Branch Fig. 1. Directive speech act scenario for the transfer of objects (Type: S requests of H that H give some X to S). K.U. Panther, L. Thornburg / Journal of Pragmatics 30 (1998) 755-769 761 In Fig. 1 the components of the scenario designated BEFORE, CORE, RESULT, and AFTER are ordered vertically along the left axis. The BEFORE component consists of the subcomponent given in its row as well as those situated above its row. Likewise, the AFTER component consists of the subcomponents given in its row as well as those situated below. All the individual components are arranged to reflect relative conceptual distance from the CORE component; we indicate this with the vertical axis on the left labelled Distance Scale and Strength of Metonymic Link. All subcomponents are linked to the CORE along three branches: the Presuppositional Branch, the Motivational Branch, and the Realiza- tion Branch. The Presuppositional Branch contains existential presuppositions and what Searle calls 'preparatory conditions'. The Motivational Branch contains the sincerity condition for directives and other reasons for performing the speech act. Subcomponents in the Motivational Branch may also be linked to those in the Pre- suppositional Branch. The Realization Branch contains the satisfaction condition of the speech act and other consequences. As stated above, any subcomponent in the scenario is a potential metonym for the whole scenario, whose essence is expressed through the CORE component. Moreover, components can be metonymically linked to other components in the scenario. For example, the com- ponent instantiated by the utteranceYou can lend me some money can be metonymically linked to You will lend me some money and either can be linked to the CORE of the scenario (the obligation of the hearer to lend some money to the speaker). 4. Speech act metonymies in conversation Given the scenario in Fig. 1, we can now formulate our first hypothesis. Hypothesis 1 The more distant a speech act scenario component is from the CORE, the weaker is its ability to evoke the scenario metonymically. In other words, the more conceptu- ally removed a component is from the CORE, the less likely that component will be in a stand for (metonymic) relation to the scenario as a whole. Given Hypothesis 1, we would expect that the basic existential presupposition X exists~is available, located at the periphery of the Directive Scenario, would be less likely to stand metonymically for a request than, for example, the more central I want X component. To illustrate this hypothesis, we have analyzed some excerpts of conversations from The Bergen Corpus of London Teenage Language (COLT). Conversations 1-3 involve points of contention about the want of an object and accompanying uncooperative behavior on the part of some interlocutors. The prag- matic function of the statements that convey the want of an object in Conversations 1-3 is constrasted with the pragmatic function of existential statements in Conver- sation 4. Bold type is used to highlight the relevant parts of the conversational frag- ments. 762 K.U. Panther. L. Thornburg / Journal of Pragmatics 30 (1998) 755-769 4.1. Metonymic Function of I want X Conversation 1 Iwl Osmond, I want my tape! ... No thank you, you haven't got one big enough. Osmond, I want my tape! ... You fucking cunt. Claire [tell him] Iw43 [] Iwl Claire, tell him I want my tape. Iw43 She wants her tape ... Iw47 Iwl Osmond, you're out of order you know. Iw43 Iw47 Iwl I apo= Iw47 Iwl I apo= What did you say? Did you say fuck Saira? ... Oh, Osmond, come on, give me the tape don't be out of order. Iw47 Shut the fuck, woman! Conversation 1 provides strong evidence that the statements of the type I want X are metonyms for urgent requests to give the desired object X to the speaker. At the beginning of the conversational fragment participant wl (Saira) states that she wants her tape back. She explicitly addresses participant w47 (Osmond) and reacts to a non- verbal (?) response of the latter by saying No thank you .... which might be inter- preted as the rejection of some inappropriate offer that does not correspond to her wishes. This seems to support the interpretation of the want statement as an (indirect) request. Saira apparently intended her utterance as a request that should be complied with by Osmond; Osmond seems to have interpreted it as a directive speech act as well but does not comply appropriately. As a result of this noncompliance with her implied request she resorts to an imprecati