A preliminary model of pronoun/verb co-occurrences in child-directed speech

Given the selectional restrictions on the kinds of subjects and objects that a verb may take, it seems likely that children learn verbs partly by exploiting statistical regularities in co-occurrences between verbs and nouns. This paper explores the role of pronouns in this process. Although pronouns are semantically "light," they dominate nouns in the speech children hear and systematically partition important classes of verbs. We show that a statistical learner can exploit these regularities to constrain the possible verbs that might fit in a simple syntactic frame.