Semantic scaffolding in first language acquisition: The acquisition of raising-to-object and object control.

摘要

This dissertation joins the debates on whether language is innate and/or modular, by examining English-speaking children's acquisition of raising-to-object (RO; (1)) and object control (OC; (2)) utterances. (1) RO: Suki wanted/needed Neili [ti to kiss Louise]; (2) OC: Suki asked/told Neili [PRO i to kiss Louise].;While these verbs may appear in the same surface string, they map onto two distinct underlying structures. As a result, they differ in their syntactic and semantic behaviors, including the interpretation of embedded passives, and whether the subject of the embedded clause may be expletive or inanimate.;Several truth-value and sentence judgment tasks yielded the following results: (1) Children have adultlike comprehension of active RO/OC utterances by age 4. (2) Children who fail on tests of matrix passives can interpret passives embedded under RO verbs (despite their greater length and syntactic complexity), but not under OC verbs (which have syntax more like matrix passives). (3) In sentence judgment tasks, children preferentially parse the embedded clause alone.;To account for these patterns, I offer the semantic scaffolding hypothesis, which comprises two major proposals: (a) children assume a canonical alignment of thematic and grammatical roles, resulting in agent-subjects and patient-objects, and (b) children assume a default clausal shape of contiguous subject and predicate. I argue that children use semantic scaffolding as a stepping stone on their way to adultlike syntactic and processing power. In short, movement may be easier than control structures, if these assumptions are not violated.;Moreover, the fact that children do maintain a distinction between the verb classes is evidence for innateness and modularity in language. However, the language module interacts crucially with other cognitive modules (e.g., the conceptual-semantic system) and with domain-general faculties (e.g., attention, memory).;Finally, the results presented here also bear on the following issues: (1) There is no evidence for maturation of A-chains and/or control, contra Wexler (1992, 2004). (2) Children's performance on active RO, passives, and embedded passives suggest that RO utterances should instead be analyzed as instances of "exceptional case marking." (3) The data can neither support nor refute Hornstein's (1999) proposal that RO and OC both be analyzed as instances of movement.