A fundamental task in sentence comprehension involves assigning semantic roles to sentence constituents, determining who does what to whom. Verb knowledge plays a central role in this task. The verb determines what constituents can appear in the sentence, and what participant roles they will convey. In learning a new verb, a child must determine what relationship among participants the verb refers to, without the set of semantic instructions provided by the verb. The syntactic bootstrapping theory proposes that children use precursors of the adult's knowledge of syntax to understand sentences and therefore to learn verbs. This view is supported by evidence that children as young as 2 assign different meanings to verbs presented in different sentence structures. The proposed research asks what syntactic cues are helpful early in acquisition, before many of the complexities of syntax acquisition have been conquered. First, we argue that children treat the number of nouns in the sentence as a cue to its semantic predicate- argument structure. The number of nouns in the sentence is useful because it provides a probabilistic indicator of the verb's number of arguments. Second, early syntactic bootstrapping requires that children represent language experience in an abstract mental vocabulary that permits rapid generalization of syntactic learning to new verbs. Thus, we argue that language-specific grammatical learning, such as detecting the significance of word order in English, should transfer quickly to sentences containing new verbs, permitting progressively finer constraint on sentence interpretation and verb learning. This project explores how syntactic bootstrapping begins, and how it interacts with early progress in syntax acquisition. We take two complementary approaches: (1) Experiments with infants and toddlers will investigate the detection and use of the proposed simple structural cues to sentence interpretation and verb learning. (2) Computational experiments using a system for automatic semantic role labeling will test the main claims of our account using a substantial sample of natural child-directed speech. This combination of experimental and computational studies is intended to advance scientific knowledge about how children learn their native languages, and to guide the development of new, robust learning protocols that will be of use in automatic natural language processing. The proposed research will help us to understand how infants and toddlers learn the words and syntax of their native languages;such research will contribute to the detection and remediation of language delays, and to language pedagogy.
