Introduction: a functional approach to ellipsis(1).

BAVIN, EDITH L.

The papers in this issue discuss children's ellipsis of arguments, focusing on the interaction of discourse (or usage) with grammar and typology. Over the past years, many of the discussions on argument ellipsis in children's language were motivated by theoretical claims about universal grammar, and much of the discussion focused on subject ellipsis in an examination of the so-called pro-drop parameter. Research on the pro-drop parameter considered the implications of language differences on the acquisition process. Languages considered included Italian, Portugese, French, Chinese, English, and German (e.g. Hyams 1986; Pierce 1992; Valian 1991; Valian and Eisenberg 1996; Huang 1989). The research focused on empirical support for the initial setting of the parameter, or the factors that influenced the setting of the parameter appropriately.

However, while such cross-linguistic research has been informative, drawing out typological features and pointing out variations across languages that must be accounted for in any theory of acquisition, there are other issues. To explain ellipsis in children's productions we need to consider a range of factors that may influence it, as well as changes in patterns of ellipsis over the course of acquisition. The adult GRAMMAR of a language has generally been the target in discussions of argument ellipsis, not adult USAGE patterns. The discourse contexts in which adults include subjects or the contexts in which they are null have not been considered in grammatically motivated accounts of null arguments. Yet it is important to examine whether ellipsis is totally variable in different contexts, or if predictions can be made about when ellipsis is likely to occur in those contexts in which it is not obligatory. A focus on adult grammar or on when a child has acquired the adult grammar does not reveal the extent to which the use of arguments in children's utterances matches adult usage patterns.

There are differences across languages in which arguments may be null. In some languages subjects will be (or may be) null only if they are pronominal; in other languages there may be ellipsis of lexical subjects as well as lexical objects (e.g. Wang et al. 1992). In languages that licence pro-drop, speakers have the choice of adding pronominal subjects for certain discourse functions, including contrast or change of topic. In non-pro-drop languages the subject may be null in specific grammatical contexts, such as in English subordinate nontensed clauses with the same subject as the main clause (e.g. `He wants to run'). In addition, the subject may also be null across adjacency pairs in conversation; that is, if the linguistic context has already provided an argument, it is omitted in the following utterance.

There are different views about what counts as ellipsis. In Tomasello's (1992) verb-island hypothesis, the assumption is that a child acquires a verb in the structure in which it is heard. Some verbs may be heard without their full complement of arguments in discourse. Thus if the child hears a transitive verb with a null object and then produces the verb in an utterance without an object, this is not an example of object ellipsis, since the child does not assign the same argument structure to the verb as does the adult. It is necessary, therefore, to consider both the language input and the child's developing grammar. As illustrated by Theakston et al. (2000), it is the statistical properties of the input that determine children's usage of arguments with different verbs.

A number of explanations have been proposed as to why young children miss out arguments in their speech; some of these are discussed in more detail in the papers in this issue. In addition to explanations based on universal grammar and grammatical constructions (e.g. Hyams 1986; Huang 1984; Radford 1990), several processing accounts have been proposed. An assumption made in the processing account proposed by Bloom (1990) is that a child cannot handle a sentence with a full set of arguments because of limited processing resources. He argues that the subject is more likely to be dropped if there is a long verb phrase included in a child's utterance, so that if an object is overt with a transitive verb, the subject is less likely to be. Gerken's (1991) processing explanation is compatible with a large amount of research on children's sensitivity to the prosodic features of the language being acquired. Arguing within a metric theory framework, she proposes that children omit weakly stressed syllables in certain positions. Because pronominal subjects are likely to be weakly stressed, they are susceptible to omission. A third type of explanation is discourse-based, influenced by the notion of preferred argument structure (Du Bois 1985, 1987). In this view, ellipsis is related to informativeness. It is proposed that new information is introduced in an intransitive subject position (S) or object position (O), but not in transitive subject position (A). Because "old" information is less likely to be included in discourse than "new," it is predicted that more As will be null than Ss or Os. Support for this view is provided by Clancy (1993), who discusses null arguments in young Korean children's utterances. Several of the papers in this issue make reference to preferred argument structure in explaining the patterns of ellipsis reported.

Acquisition takes place in a discourse context and is discourse-driven. In many cultures, an adult prompts a child to produce utterances and the child responds with information that complements the adult's, leaving out information that may be assumed from the context. From older speakers, null arguments in narratives may relate to topic continuity. By examining narrative data across different age groups, we can draw a distinction between subject and topic ellipsis and identify developmental patterns. By adopting an integrated approach to the study of null arguments, that is, an approach that considers the grammar, typology, and use of language, we are in a better position to identify developmental trends in the ellipsis of core arguments, and to understand when a child shows knowledge of language-specific patterns.

The papers in this issue draw together data from a range of typologically different languages (Hebrew, Inuktitut, Swedish, Icelandic, Sesotho, Warlpiri, and French), and each paper draws out particular features of the language under investigation. The patterns of ellipsis reported are not restricted to early word combinations from young children but are extended to include productions from children at later stages in their language development. The papers on Warlpiri and French focus on data from children's narratives.

Overall, it is evident from the studies reported that different factors may influence whether arguments are overt or null at different stages of language development. For example, in their paper on Hebrew, Uziel-Karl and Berman argue for three stages: ellipsis is first guided by pragmatic factors and then by morphosyntactic factors, and at a third stage the two are integrated. This finding is compatible with other discussions in the acquisition literature showing that language forms are not necessarily used by young children with the same function or range of function as by older speakers (e.g. Karmiloff-Smith 1981, 1985; Slobin 1985). There is often reorganization of knowledge as the child moves from child speaker to mature language user, and such reorganization is frequently represented as U-shaped learning. It is also compatible with reports that infants' early combinations may show lexical specificity; for example, only certain verbs may appear in specific constructions, leading to the view that infants may not be operating with categories of the mature system but may be in the process of building up categories, as argued, for example, by Pine et al. (1998). Because of limitations in the use of a linguistic form by a child, Maratsos (1998) is hesitant to attribute an adult-like category; he adopts the term prospective category to describe a category that might or might not be adult-like for the child. We need to keep this clearly in mind when we refer to null "subjects" or null "objects" in early child language.

The first paper in this issue, by Uziel-Karl and Berman, draws on a model that views acquisition as a stepwise process (Uziel-Karl 1997). The authors propose two distinct developmental criteria: one to determine when a child has knowledge of a linguistic item or construction, and the other to determine when a child has adult-like knowledge. They argue that grammatical and discourse factors play a role in accounting for null arguments in Hebrew children's language. Three factors that influence argument ellipsis are discussed: permissibility, recoverability, and syntactic function. Basing their arguments on longitudinal data collected from four children between the ages of 17 and 28 months, supplemented by other data, Uziel-Karl and Berman argue for different explanations of argument ellipsis at different ages. Pragmatic factors can account for the subject ellipsis in the data from the younger children, but subject ellipsis at later periods is conditioned by morphosyntactic rules of the language. In contrast, object ellipsis is conditioned by pragmatic or semantic factors. Hebrew is of particular typological interest in discussions of ellipsis since its pro-drop property is conditioned: first and second person pronominal subjects can be omitted, not third, and only in the past and future tense.

Allen's paper illustrates that even young children are sensitive to the dynamics of information flow. Her paper shows that grammar-based accounts of ellipsis are not adequate to account for null versus overt arguments in child language, because some contributing factors will be overlooked. Allen argues for an integrated approach, which draws upon theories of grammar and of discourse-pragmatics. She examines the factors that contribute to the inclusion or omission of arguments in the speech of four Inuktitut children aged 2-3;6 years. The data are longitudinal and from spontaneous productions. For the analysis, Allen adopts a detailed coding scheme; eight features are used, relating to informativeness, (joint) knowledge, (potential) confusion, and search space. A logistic regression analysis is used to determine if certain factors are better than others as predictors of argument ellipsis. The finding that third person arguments are more likely to be overt than first or second can be explained on the basis of the grammatical features of the language. However, other identified factors indicate that even young Inuktitut children are sensitive to discourse-pragmatic features.

Stromqvist and Ragnarsdottir investigate the use of subject arguments by a child acquiring Swedish in the age range 22-24 months, and the distribution of spatial arguments by the same child (24-26 months) and a child acquiring Icelandic (also 24-26 months). The authors argue that constraints govern the use or ellipsis of these arguments and that the constraints are mastered gradually. They illustrate, from longitudinal data, that both input and pragmatic factors influence the distribution of the arguments in the early stages of acquisition. A processing explanation for subject ellipsis cannot be supported; the child utterances with overt subjects are found to be longer than those with missing subjects. However, discourse factors influence the inclusion of subjects since an overt subject tends to be a new referent. An examination of the use of spatial arguments with spatial verbs in the child data shows differences across Swedish and Icelandic. Explanations are proposed, not only as to why differences are evident across the two languages, but also as to why there are differences in usage between adult and child.

The paper by Demuth, Machobane, and Moloi presents an analysis of null objects in Sesotho ditransitive applicative constructions. Spontaneous productions from two two- to three-year-olds are discussed as well as adult speech directed to these children. In addition, results from an experimental task are reported. The task was presented to children aged from three to eight years, and also to a group of adults. A main purpose of the study was to determine at what age Sesotho children show awareness of animacy restrictions on the ordering of double objects. In the child-directed speech only about one-quarter of the ditransitive applicative constructions were found to include both objects; none of these had both arguments overt as postverbal noun phrases. This raises a theoretical issue that is taken up in the paper, the problem of learnability. The authors consider the impact of statistical properties of the input on language acquisition and its implication for theories of acquisition. They propose that a form of discourse bootstrapping may be required for children to learn the argument structure of Sesotho verbs.

The other two papers in the issue focus on ellipsis in narrative contexts. My paper reports on ellipsis in Warlpiri children's narratives. The language allows ellipsis of both lexical subjects and objects. It also has flexible word order, and thus factors other than word order must be used when interpreting sentences; these include morphology, semantics, and discourse-pragmatic factors. Pronominal clitics that register the person, number, and case of subject and object arguments are obligatory with finite verbs; however, third person singular forms are null. The narratives elicited from Warlpiri children aged from four to ten years are compared with adult narratives. A major factor found to influence lexical subject ellipsis is same- versus switch-subject contexts; topic continuity seems to influence the use of null lexical subjects. The use of overt pronominal clitics in relation to lexical argument ellipsis is also analyzed. Overall developmental trends are reported and discussed in terms of reorganization in knowledge of language and language use, drawing upon Karmiloff-Smith's (1985) three-phase model.

Jisa's paper focuses on referential cohesion in French narratives from five- to ten-year-olds, and a group of young adults. As Jisa elaborates in her paper, previous research has concentrated on the use of pronouns to maintain discourse referents with less attention paid to other grammatical options. The author examines a number of the grammatical options available in French and discusses developmental changes in the use of the available structures for the discourse functions of maintaining and reintroducing referents. Her analysis shows a developmental increase in ellipsis. Jisa argues that both ellipsis and nonfinite subordination contribute to tighter syntactic cohesion of information. On the basis of her findings, she proposes two kinds of developmental explanation: one relating to children's developing control over the means for packaging information, and the other to their developing flexibility in adapting to a register appropriate for the task.

The conclusion that can be drawn from the papers in this issue is that different perspectives can be complementary; we need an integration of different approaches to account for ellipsis (see also Hamman and Plunkett 1998). An approach focusing on grammatical explanations for ellipsis in young children's speech overlooks other factors that influence when arguments are overt or null, including statistical properties of use in the input language. Discourse-pragmatic knowledge develops alongside grammatical knowledge; this needs to be remembered when we examine young children's productions in an attempt to determine knowledge of grammar.

La Trobe University

Note

(1.) Correspondence address: School of Psychological Science, La Trobe University, Bundoora, Victoria 3083, Australia. E-mail: e.bavin@latrobe.edu.au.

References

Bloom, Paul (1990). Subjectless sentences in child language. Linguistic Inquiry 21,491-504.

Clancy, Patricia M. (1993). Preferred argument structure in Korean. In Proceedings of the Twenty-Fifth Child Language Research Forum, Eve Clark (ed.), 307-314. Stanford: CSLI.

Du Bois, John W. (1985). Competing motivations. In Iconicity in Syntax, John Haiman (ed.), 343-365. Amsterdam: Benjamins.

--(1987). The discourse basis of ergativity. Language 63, 805-855.

Gerken, LouAnn (1991). The metrical basis for children's subjectless sentences. Journal of Memory and Language 30, 431-451.

Hamman, Cornelia; and Plunkett, Kim (1998). Subjectless sentences in child Danish. Cognition 69, 35-72.

Huang, C.-T. James (1984). On the distribution and reference of empty pronouns. Linguistic Inquiry 15, 531-574.

--(1989). Pro-drop in Chinese: a generalized control theory. In The Null Subject Parameter, Osvaldo Jaeggli and Ken Safir (eds.). Dortrecht: Kluwer.

Hyams, Nina (1986). Language Acquisition and the Theory of Parameters. Dortrecht: Reidel.

--(1987). The theory of parameters and syntactic development. In Parameter Setting, Thomas Roeper and Edwin Williams (eds.), 1-22. Dortrecht: Reidel.

Jaeggli, Osvaldo; and Safir, Ken (1989). The null subject parameter and parametric theory. In The Null-Subject Parameter, Osvaldo Jaeggli and Ken Safir (eds.), 1-44. Dortrecht: Kluwer.

Karmiloff-Smith, Annette (1981). The grammatical marking of thematic structure. In The Child's Construction of Language, W. Deutsch (ed.), 121-147. London: Academic Press.

--(1985). Language and cognitive processes from a developmental perspective. Language and Cognitive Processes 1, 61-85.

Maratsos, Michael (1998). Relations of lexical specificity to general categories. Linguistics 36, 831-846.

Pierce, Amy (1992). Language Aquisition and Syntactic Theory. Dortrecht: Kluwer.

Pine, Julian; Lieven, Elena V. M.; and Rowland, Caroline (1998). Comparing different models of the development of the English verb category. Linguistics 36, 807-830.

Radford, Andrew (1990). Syntactic Theory and the Acquisition of English Syntax. Oxford: Blackwell.

Slobin, Dan I. (1985). Crosslinguistic evidence for the language-making capacity. In The Crosslinguistic Study of Language Acquisition, Dan I. Slobin (ed.), 1157-1256. Hillsdale, NJ: Erlbaum.

Theakston, Anna; Lieven, Elena V. M.; Pine, Julian; and Rowland, Caroline (2000). The role of performance limitations in the acquisition of "mixed" verb-argument structure at stage I. In New Directions in Language Development and Disorders, Michael Perkins and Sarah Howard (eds.), 119-128. New York: Kluwer/Plenum.

Tomasello, Michael (1992). First Verbs: A Case Study of Early Lexical Development. Cambridge: Cambridge University Press.

Uziel-Karl, Sigal (1997). How general is verb argument structure: the developmental history of some early verbs in Hebrew. Proceedings of GALA, 172-177.

Valian, Virginia (1991). Syntactic subjects in the early speech of American and Italian children. Cognition 40, 21-81.

--; and Eisenberg, Zena (1996). The development of syntactic subjects in Portuguese-speaking children. Journal of Child Language 23, 103-128.

Wang, Q.; Lilo-Martin, Diane; Best, Catherine; and Levitt, Andrea (1992). Null subjects versus null objects: some evidence from the acquisition of Chinese and English. Language Acquisition 2, 221-254.