Prefinal copy The Development of Pragmatics: Learning to Use Language Appropriately* Anat Ninio Catherine E. Snow Department of Psychology Graduate School of Education Hebrew University Harvard University Jerusalem 91905, Israel Cambridge, MA 02138, U.S.A. Reference: Ninio, A. and Snow, C. (1999). The development of pragmatics: Learning to use language appropriately. Invited chapter, in T. K. Bhatia & W. C. Ritchie (Eds.), Handbook of language acquisition (pp. 347-383). New York: Academic Press. *Work on this paper has been supported by the Israel Foundation Trustees through a grant to Anat Ninio, and by the National Institute for Child Health and Human Development through HD 23388 to Catherine Snow. 1 Defining the domain of pragmatics Pragmatics versus syntax and semantics in formal models. Broadly speaking, pragmatics is the study of the use of language in context for the purpose of communication. Thus, developmental pragmatics is concerned with the question of how a child acquires the competencies underlying the rule-governed employment of speech in interpersonal situations. The criteria applied in assessing the pragmatic success of an utterance have to do, not with grammaticality (the criterion applied within the domain of syntax) or truthfulness and interpretability (the semantic criteria), but with appropriateness and communicative effectiveness. This demarcation of the field of developmental pragmatics is, however, far from straightforward or noncontroversial. The status of language use in the study of language is a highly vexed issue within linguistics and the philosophy of language. Indeed, the very distinction between the domain of pragmatics and those of syntax and semantics is itself theory-dependent. The trichotomy of syntax, semantics and pragmatics originated with Morris (1938), who defined syntax as the study of the formal relations of signs to one another, semantics as the study of the relations of signs to the objects they denote, and pragmatics as the study of the relation between signs and their interpreters, i.e., the speaker and the addressee(s). This distinction between the concerns of pragmatics on the one hand and those of semantics and syntax on the other appears in various forms in the writings of many linguists and philosophers of language. It is echoed (with theory-specific distinctions of meaning) in de Saussure's (1922/1983) distinction between language and speech (la langue and la parole); in Chomsky's (1965) distinction between performance and competence; in Austin's (1962) differentiation of illocutionary and locutionary acts; in Searle's (1975) distinction between illocutionary force and propositional content and between literal and indirect speech acts (1969); in Grice's (1968; 1969) discrimination of utterer's-meaning and sentence-meaning; in Lemmon's (1966) statements vs. sentences, and so forth. Such a principled distinction between pragmatics and semantics and syntax is the central tenet of formal linguistics. This approach is closely associated with, and mutually reinforcing of, a particular philosophical theory of linguistic meaningfulness, the so-called 2 truth-conditional theory of meaning, identified with such authors as Carnap (1956), Davidson (1979), or Lewis (1972). According to this theory, the sense of sentences is determined by their truth-conditions; to know the conditions under which a sentence is true is to know the meaning of the sentence. Linguistic meaning or meaningfulness is a property of sentences, largely independent of actual contexts of use (Davidson, 1979; Wiggins, 1971). The conception of language this theory supports is one of an autonomous system of fixed symbols and of abstract rules for their lawful combination, all defined independently of their possible contexts of use. The uses people make of the meaningful sentences generated by this system are properly speaking extra-linguistic; and if these uses are rule-governed, the description of such rules belongs to a separate discipline, such as pragmatics or sociolinguistics. Within this conception, then, the proper concern of pragmatics is the description of phenomena related to the use of meaningful linguistic forms for communicative purposes. Chief among these is the production and comprehension of speech acts--making statements, requesting, promising and the like. Other phenomena include the regulation of conversational exchange; politeness rules and other culturally conventionalized variations in speech register that convey social meaning and determine appropriateness; the control of presuppositions; and the creation of connected discourse. Linguistics proper, namely syntax and semantics, is confined to concern with the generation of meaningful utterances independent of their potential or actual context of use (cf. Gazdar, 1979; Kempson, 1975; Levinson, 1983). As Chomsky (1990) put it, 'there is no reason to single out communication among the many uses to which language is put' (p. 57). It is however obvious even to the most extremely formalist of linguists that language contains expressions which point to contextual information rather than symbolizing some context-independent abstract concept. These expressions are the deictic or indexical elements of language: pronouns, deictic locatives such as here and there, deictic temporal terms such as today, tomorrow, verbs like come, go, bring, and the like. These expressions receive their value from the context of their use; thus, the sentences in which they are used do not have truth conditions (hence, meaning) except as a function of their situated use. Since they are bona fide lexical items, it is impossible to exclude the rules governing the use of such deictic expressions 3 from linguistics proper. However, the treatment of such expressions does not fit well into linguistic theories developed to deal with linguistic entities of apparently context-independent nature. It has therefore been proposed (cf. Bar Hillel, 1954; Kalish, 1967; Montague, 1968) that linguistic theory proper be complemented by a separate pragmatic component dealing with indexical or deictic expressions. From the formalist point of view, then, pragmatics is already an ill- defined domain, best characterized by means of a list enumerating the various phenomena of inquiry (cf. Levinson, 1983). These use-related phenomena are thought by the formalist to possess varying degrees of affinity with language phenomena proper. Formalists would place on the very boundaries between linguistics and language use deictic expressions, as well as other in-between phenomena such as the grammar of units larger than a single sentence. Speech act production is only slightly more peripheral, since it involves such purely grammatical features as the control of grammatical mood, performative sentence-structure, a propositional core and so on (cf. Lyons, 1981; Searle, 1975). Further away from linguistics proper are the control of presuppositions, politeness rules, and the management of conversational exchanges, in that order. Many would claim these fall in the province of sociolinguistics rather than linguistics proper, and thus even beyond the concerns of pragmatics. The place of pragmatics in functionalist models. However, the distinction between language and its use underlying the above demarcation of the field is not the only conceptualization possible. An alternative philosophical approach to linguistic meaningfulness with its attendant linguistic metatheory is the so-called use-conditional theory of meaning. This approach ties the concept of linguistic meaningfulness to the communicative use people make of language. The most well-known expression of this viewpoint is the later Wittgenstein's dictum that "for many expressions... to know their use is to know their meaning" (1953). According to this approach, the primitive concept from which meaning is derived is one of communication rather than one of truth: The meaning of an utterance is what a speaker has intended overtly to communicate to an addressee (Alston, 1964; Grice, 1957, 1968; Strawson, 1970). This approach gives logical priority to utterance-meaning over sentence-meaning. The meaningfulness of utterances derives from their use in performing overtly communicative acts. 4 The meaning of sentences is a kind of abstraction from their occasions of use; it is their potential to be used communicatively (Allwood, 1981; Alston, 1964, 1967, 1968; Fillmore, 1971; Grice, 1969). The literal meaning of a sentence may be the meaning element common to all contexts of use, the union of all possible uses or, most probably, the most representative or prototypical use the sentence can be put to (cf. Allwood, 1981; Gibbs, 1984). Similarly, word meaning may be defined as the word's contribution to the use-potential of a sentence (Alston, 1968), or its prototypical conditions of use (Allwood, 1981). The linguistic theory associated with this philosophical tradition emphasizes the indexical, context bound nature of all language. Semanticists working in this tradition have shown that the depiction of word meaning as a context-free conglomeration of semantic features is unable to account for such varied semantic phenomena as polysemy, contextual expressions, labeling, and misusage (eg., Anderson & Ortony, 1975; Clark & Gerrig, 1983). Adjustments designed to rescue the formal approach, such as two-stage models for the interpretation of speech-acts (e.g., Searle, 1975; Gordon & Lakoff, 1971) and of ironical and metaphorical utterances (Kintsch, 1974; Searle, 1979), have not stood up to theoretical analysis or psycholinguistic experimentation (eg., Gibbs, 1984; Dore & MacDermott, 1982; Gazdar, 1981; Lakoff, 1982; Rumelhart, 1979; Sperber & Wilson, 1981, Streek, 1980). Linguistic theories associated with the use-conditional viewpoint are called functional rather than formal (e.g., Dik, 1978; Halliday, 1985). As a generalization, such theories see linguistic forms (words, grammar) as tools for the purpose of communication, namely, sets of subordinate units to be made use of in the construction of decipherable messages. The implicit, internalized linguistic system is thought to consist of "production rules", namely, of procedures for verbally expressing various kinds of communicative intents. Formal rules of language (syntax, morphology, lexicon) appear in this view merely to be components of such a system of production rules (e.g., Bates and MacWhinney, 1982). In the functionalist approach, the boundaries between syntax, semantics and pragmatics are blurred. What the formalists call linguistic competence proper is but a subset of the "communicative competence" needed to use language meaningfully (Hymes, 1971). According to this viewpoint, no difference exists between pragmatics and the rest of linguistics. The goal of linguistics is to achieve an understanding of how speakers create 5 interpretable utterances in communicative situations. All boundaries between formal and contextual aspects of language are seen as artificial and ill-conceived; the system as a whole is completely contextual and does not possess autonomous components. If it is agreed that the task of pragmatics is the study of language use in context, and if all language is inherently contextualized, then pragmatics is the most general discipline encompassing all aspects of language. Pragmatics in developmental theories. Developmental psycholinguists studying the acquisition of pragmatic competence take any of a variety of standpoints -- whether explicitly or implicitly -- on the general theoretical issue of the nature of the linguistic system. Much work in this field accepts implicitly the 'separateness' of pragmatics, and studies pragmatic development by looking at pragmatic skills more or less in isolation. Other research, though, embodies the assumption that children's initial language is best described in use-conditional terms, whatever the form and nature of the endstate linguistic system. According to this approach, a substratum of social meanings or communicative intents underlies the regularities observed in early single and multiword speech, and the young child's linguistic system consists of realization rules that connect form and function in their speech (e.g., Bates, 1976; Bates, Camaioni, & Volterra, 1975; Bateson, 1975; Benedict, 1979; Braunwald, 1978; Bruner, 1975a; Clark & Clark, 1977; Greenfield & Smith, 1976; Halliday, 1975; Nelson, 1985; Ninio, 1983; Ryan, 1974; Rogdon, Jankowsky, & Alenskas, 1977; Snow, 1979). Among these studies taking a theoretically intermediate position one could also include studies concerned with the identification of pragmatic factors influencing language acquisition, whether or not such studies operate with an integrated pragmatically-based acquisition model. These include investigations of the interactive context of language use in early childhood (Bruner, 1983; Cazden, 1970; Chapman, 1981; Lieven, 1978a, 1978b; Snow, 1977a, 1979); studies concerned with the role of maternal input and scaffolding behavior on the acquisition of linguistic forms (eg., Mervis & Mervis, 1988; Moerk, 1976; Nelson, 1977; Nelson, Denninger, Bonvillian, Kaplan, & Baker, 1984; Ninio, 1985; Ninio & Bruner, 1978; Sachs, Brown, & Salerno, 1976; Snow, 1972, 1977b; Tomasello & Todd, 1983); and, more marginally, studies in which an appeal is made to putative pragmatic reasons for the explanation of acquisition biases (e.g., Benedict, 1979; Stephany, 1986; Wales, 1986). Studies 6 focussing primarily on the interactive context of language acquisition will not be reviewed systematically here (but see Snow, 1995in press, for a review). A theoretically radical group of investigators do not consider the early use-conditional language as a transitory stage but explicitly adopt a use-conditional framework for the characterization of endstate language itself. On this approach, the use-based linguistic system postulated for young children is assumed to be in principle, if not in detail, isomorphic and continuous with the full-blown adult system. In the latter, as in the former, meaning is assumed to be use-conditional; the latter, as the former, is thought to consist of rules for the realization of (and comprehension of) communicative messages in verbal form. In these studies the assumption is made that the goal of developmental psycholinguistics is to explain the entirety of language acquisition as the acquisition of functionally effective tools for communication (Bates & MacWhinney, 1979; Halliday, 1975; Ninio, 1994ain press a; Tomasello, 1992; Tough, 1977; van Valin, 1991). These studies adopt as their theoretical framework linguistic theories that do not acknowledge a rigid demarcation of syntax, semantics and pragmatics, such as Halliday's (1985) Systemic-Functional Grammar; Givon's (1989) Functional Linguistics; Cognitive Linguistics (eg., Lakoff, 1990; Langacker, 1987); or Hudson's (1990) Word Grammar, a communicatively oriented dependency grammar. The boundaries between this group of studies and the previous one are fuzzy. As long as the target population whose language development is studied are beginning speakers, the theoretical stance taken regarding the ultimate course of development seldom makes a difference in terms of research goals or methodology. There are however exceptions to this generalization; one focal controversy in the field exemplifies the pervasive presence of high-theoretical considerations in the study of early pragmatic competencies. The controversy arose from Bates et al.'s (1975) claim that, prior to the emergence of language, children perform communicative acts such as requesting by nonverbal means, and that the status of these acts is essentially identical to those of illocutionary acts. The later use of verbal means for the expression of the same kind of illocutionary intents is merely the substitution of verbalizations for prior nonverbal behaviors. Dore (1978) criticized this analysis and, relying on Searle's (1969) Speech Act Theory, argued that illocutionary acts incorporate a grammatical component which preverbal communicative acts obviously lack. Therefore, he claimed, 7 prelinguistic communicative skills are necessary but not sufficient for the acquisition of linguistic communication. By themselves, developments in the pragmatic domain cannot provide an explanation for the emergence of speech; substitution of means is not the vehicle of development. Thus, for Dore, children's gradual mastery of types of communicative intents is a separate line of development from the development of language proper, the one not feeding into the other. Indeed, according to Dore, it is not only preverbal communication which is irrelevant to true illocutionary act production but also most of single-word speech; in his publications on the subject (e.g., 1975) he claims that even in true verbal utterances carrying illocutionary force, the illocutionary intent is conveyed by nonlinguistic means such as intonation, whereas the words themselves serve only as rudimentary referring expressions naming some random element of the communicative situation. Thus, Dore's studies of the development of speech act production, although superficially similar in their subject matter to Bates' investigations of early pragmatics, actually represent a different viewpoint, one that that clearly differentiates between these pragmatic developments and language acquisition proper. A relatively complete review of empirical work in the domain of pragmatics must abstract from the theoretical differences among the authors on the issues identified above. In the following, we shall review work on the phenomena that, in our view, constitute the major achievements of language learners within the domain of pragmatics, organizing our review under the following four topics: 1. The acquisition of communicative intents and the development of their linguistic expression, seen as a separate issue from the acquisition of linguistic forms and grammar in the abstract. 2. The development of conversational skills and the acquisition of rules that govern turn-taking, interruptions, back-channeling, signalling topic-relevance or topic switch, and so on. 3. The acquisition of rules of politeness and other culturally determined speech use rules. 4. The development of control over the linguistic devices that govern connected discourse, such as ways to generate cohesion across utterances within a longer discourse, to signal presuppositions, as well as procedures for organizing information into familiar genre-specific forms. We will not review those studies dealing with the acquisition of deictic forms such as 8 pronouns and deictic locatives which just refer to the pragmatic underpinnings of this kind of knowledge without studying directly the pragmatic aspects of its development (e.g., Charney, 1980; Chiat, 1981, 1986; Clark & Sengul, 1978; de Villiers & de Villiers, 1973; Karmiloff-Smith, 1979; Wales, 1986; Webb & Abramson, 1976). Developing Communicative Intents and Appropriate Ways of Expressing Them In this section we shall be concerned with two questions: first, what type of communicative intents can children express, and second, how are these intents expressed, throughout the course of development. We focus, thus, on children's entry into language via their development of control over rules of expression for speech acts. Prelinguistic communicative abilities. Among investigators of early social behavior, there exists wide agreement that infants possess a precocious social understanding which enables them both to emit intentional communicative signals and to interpret the significance of social signals, actions and events in their environment, prior to understanding speech (Acredolo & Goodwyn, 1988; Antinucci & Parisi, 1975; Bates, 1976; Bronson, 1981; Bruner, 1975a, 1975b, 1978; Carpenter, Mastergeorge, & Coggins, 1983; Carter, 1975a, 1979; Chapman, 1981; Collis & Schaffer, 1975; de Villiers & de Villiers, 1978; Dore, 1974, 1975, 1976, 1978; Eckerman, Whatley, & Kutz, 1975; Edwards, 1978; Golinkoff, 1983; Golinkoff & Gordon, 1983; Harding, 1983; Harding & Golinkoff, 1979; Lewis et al., 1975; Lock, 1980; Moerk, 1975; Rheingold, Hay, & West, 1976; Ross & Kay, 1980; Scoville, 1984; Snow, de Blauw, & Dubber, 1982; Snyder, 1978; Sugarman-Bell, 1978; Vandell, 1977; Zinober & Martlew, 1985). Although this assumption is occasionally attacked by skeptics (e.g., Shatz, 1982), it is supported by a long line of meticulous research into young children's interactive abilities (cf. papers in Feagans, Garvey, & Golinkoff, 1984; Lamb & Sherrod, 1981; Rubin & Ross, 1982). According to some authorities, the possession of communicative intents and the interpretation of others' intents is such a basic human capacity that there is reason to believe that it is a biological disposition or language universal (Dore, 1985; Miller, 1970; Trevarthen, 1977, 1979; Trevarthen & Hubley, 1979). Communicative intents expressed at the preverbal stage. Even though it is widely agreed that preverbal infants use intentional vocalizations and gestures such as pointing, gazing and 9 giving for communicative purposes, there is a relative dearth of systematic information on the types of social meanings they express by such means. The studies that have been conducted fall into three categories. First, there are investigations documenting preverbal infants' gradually evolving participation in specific interactive contexts such as peek-a-boo (Bruner, 1975a; Bruner & Sherwood, 1976; Ratner & Bruner, 1978) or book-reading (Ninio & Bruner, 1978). Second, there have been several studies of infants' nonverbal expression of illocutionary-type speech acts in dyadic interaction with their adult caretakers. In one of the earliest studies of this kind, Bates et al. (1975) described two broad categories of preverbal precursors to verbally performed speech acts: proto-declaratives, used to direct the other's attention to a focus, the precursors of statements; and proto-imperatives, used to bring the other to do something, the antecedents of requests and other directives. Coggins and Carpenter (1981), using a more detailed typology, found the following types of pragmatic behaviors in preverbal children: requesting (object requests, action requests, information requests); greeting; transferring objects; showing off; acknowledging; answering. Carpenter et al. (1983) used 7 categories to characterize preverbal communication, including, in addition to the above-mentioned, protesting, comment on action, and comment on object. Interestingly, none of the studies in this group mentioned game-embedded meaningful behaviors of the type documented by Bruner and his collaborators among the preverbal communicative acts infants were able to control. A third category of studies focussed on preverbal and early nonverbal communicative behaviors exhibited by infants in interaction with a peer (Bronson, 1981; Eckerman, Whatley, & Kutz, 1975; Dunn & Kendrick, 1982; Lewis, Young, Brooks, & Michalson, 1975). These studies expand the list of early social meanings communicated to include expressions of feelings including sympathy; offers and sharing; compliance and non-compliance or resistance; stage-setting, and pretend-games. In all, infants appear to be able to control an impressive array of social meanings and to convey them to others by nonverbal means. These results provide converging evidence on the precocity of the social capacities underlying communication, and indicate that communicative exchanges may indeed serve a functional role in children's entry into language. The continuity hypothesis. To what extent early language use is built upon the preverbal communicative system is an as yet unsettled issue. Some investigators have claimed that