Acquiring a Grammar with Covert Head Incorporation: The Case of Have as a Complex Verb

This thesis investigates the problem of induction through an examination of a particular set of constructions (those containing the verb have) and a particular syntactic analysis that unifies this set (an analysis according to which have is a complex verb). Two main questions are addressed. One asks what type of grammatical input is necessary to acquire a grammar containing a complex verb. The other asks whether this input is available to children, and what type of information children actually rely on in order to acquire this complex verb. I summarize the syntactic evidence for analyzing have as a complex verb (as the copula be plus an incorporated abstract preposition), which is based on arguments seeking to unify crosslinguistic patterns found in possessive and past participle constructions. Based on these arguments, I use a computational learning algorithm to demonstrate that fairly rich syntactic input is needed to identify a grammar containing complex have. My results show that it is possible to learn this grammar if the input provides the locations of empty heads. Finally, I examine the syntactic patterns that children might exploit in acquiring a grammar containing an element such as have. Here I discuss the necessity of reasoning across utterances, and of relying on have-be alternations to determine that have is related to be. I conclude that in order to analyze have as a complex verb (be+P), children must rely on knowledge of syntactic structure that is not extractable from auditory input: the locations of empty heads.