Integrating sentence-structural and event information in early verb learning.

摘要

Children use syntax as well as observations of events to learn verb meanings. This is known as syntactic bootstrapping. This dissertation investigated the origins and mechanisms of syntactic bootstrapping. Prior evidence suggested that two-year-olds, but not younger children, could use aspects of sentence structure to assign different interpretations to novel transitive and intransitive verbs. Persistent negative findings with younger children raised questions about the kind of knowledge young children need about syntax-semantics links in order to begin using sentence-structural information to learn verbs. In this dissertation, I reported evidence (Chapter 2) that 21- and 19-month-old children showed sensitivity to the sentence structure in interpreting novel transitive and intransitive verbs, as long as the number of nouns in the sentence was informative. The results suggest that syntactic bootstrapping might begin with a bias in young children to assign each noun in the sentence to a core participant-role in their conceptual representations of events. Moreover, I found that 2-year-old Mandarin learners also used the number of noun phrases in novel-verb interpretation (Chapter 3), despite frequent argument omission that makes the number of noun phrases per sentence a less reliable indicator of verb argument-structure in Mandarin than in English. This extension to Mandarin learners suggests that syntactic bootstrapping is a fundamental part of verb learning, even for learners of languages (such as Mandarin) that provide much less evidence for a reliable link between noun-phrase number and verb argument-structure in the input. Finally, I demonstrated that 2-year-olds found sentence-structural information independently informative when learning about a new verb (Chapter 4). Specifically, 2-year-olds learned about a new verb's combinatorial privileges---transitivity and participant-role number---from brief dialogues alone, and later retrieved this information to guide their interpretation of the verb when it was presented in a referential context. The results suggest that verb learning is rooted in children's learning about the combinatorial properties of verbs. Given a simple bias toward a one-to-one mapping between nouns in sentences and participant-roles in their construals of events, children can use what they learn about the combinatorial privileges of particular verbs to figure out skeletal aspects of their meanings.