Summary of Proposed Research
The two main syntactic properties of early child language are: a) early alignment with the distributional patterns of the target adult grammar (1a); and, b) constrained patterns of omissions of functional material (1b), and of arguments (1c-d).| (1) | a. Elle roule pas. | ‘It does not roll.’ | (Grégoire 1 year; 11 months) |
| b. | Mary go. (Sarah 2;3) | Papa have it. | (Eve 1;6) |
| c. | Ø want Ø! | ‘I want that’ | (M, 1;11) |
| d. | Mets Ø dedans. | ‘Put in.’ | (Grégoire 1;11) |
Argument omission in child language raises important questions about the initial representation of syntactic categories in clausal structure and the relationship between the various stages of development. Our research concentrates on object omission, an area of research that has not received much attention until fairly recently. Various proposals link early object omission to clitic spell out (Fujino & Sano 2002), parametric discontinuity (Müller & Hulk 2001), pragmatic principles (Schaeffer 2000), computational limitations in the developing child (Wexler, Gavarrò & Torrens 2003). The debate on the status of null objects has developed into multiple studies on object omission in a variety of languages including Swahili, Dutch, German Spanish, Catalan, Romanian and Serbo-Croatian. Paradoxically, no study has yet explored the relationship between the development of transitivity in child grammar and object omission.
Recent theoretical work in syntax (Borer, Bowers, Hale & Keyser, Roberge) has resulted in new approaches to transitivity, which allow us to reconsider the issue of early object omission. Previous studies assume that it is specific properties of a given verb that determine whether an object can be merged with this verb or not. Note that for the subject of a clause, a general grammatical principle is what provides an external subject position while the verb provides the interpretation of the role (if any) of the subject. We adopt a similar view for objects: the direct object position is merged independently of the lexical properties of the verb. Merge is an operation that creates a structure through the association of two elements. Under this approach, the traditional verb classes based on transitivity (transitive, unergative, etc.) are no longer attributable to a verb’s selectional property (i.e., it requires an object or not) but rather to the type of interpretation provided by the verb to the object. This approach considers object expression as the contribution of various modules: lexicon/pragmatics/syntax. Therefore, in principle, any instance of target-deviant object drop as in (1c-d) can originate in a developmental discontinuity in any domain (lexicalknowledge of the status of the entry as intransitive, or morphosyntacticknowledge of conditions on spell out of clitics or agreement or pragmatic and performance factors). How can we reduce the indeterminacy that inherently underlies object omission? We propose an acquisition model that attributes to the child the same fundamental operations available in adult syntax (continuity hypothesis), as well as the ability to recover the syntactic structure of preceding discourse in their treatment of nominal ellipsis (Wijnen, Roeper & van der Meulen 2004). However, parametric variation in identification mechanisms presents an area where learning is required for the child: we propose that systematic deviations from adult target originate in parametric discontinuity (Rizzi 2002). Additionally, we suggest that functional underspecification in the object clitic/agreement domain yields identification via discourse-linking (Borer & Rorhbacher 2002).
The hypothesis we put forward in this research is that systematic cross-linguistic differences in syntactic development arise in matters involving functional identification and not contextual or lexical identification. For instance, we predict specific developmental difficulties with clitic systems. Our studies are designed to target the three mechanisms: functional, lexical and contextual.