| In this dissertation, a fairly uncommon relative clause construction is studied: the noun that the relative clause is "about" (the head) occurs only inside the relative clause and not outside of it, hence the name internally headed relative clause (IHRC). Data from many languages is used to support the arguments and conclusions.; IHRCs are defined as nominalized sentences which modify a nominal internal to the sentence. This definition distinguishes IHRCs from externally headed relative clauses and correlative clauses, as well as from other nominalized sentences.; A very general syntactic of IHRCs is argued for. It is then shown how this analysis can be expressed in terms of three current syntactic theories: Government and Binding, Head Driven Phrase Structure Grammar, and Lexical Functional Grammar. The basic analysis is then extended for each theory to cover various Island Constraints as well as lexical restrictions on the verb forms in IHRCs. In every case, it is shown how the mechanisms of the more familiar externally headed relative clauses can be used to account for the facts of IHRCs, demonstrating the similarity between IHRCs and externally headed headed relative clauses.; A general semantic analysis of IHRCs is then argued for, using the general framework proposed in Heim (1982). This analysis is shown to account for an otherwise puzzling definiteness effect on the internal head. Apparent exceptions to the definiteness effect are also accounted for. It is shown how this analysis, coupled with the appropriate analysis of negation put forth here, explains the distribution of negative indefinites with respect to IHRCs in Lakhota.; Finally, the cross-linguistic distribution of IHRCs is considered. A necessary condition for the presence of IHRCs is proposed, accounting for the connection between IHRCs and nominalized sentences. One proposed universal is shown to be false, and its effects and exceptions are shown to be explained by a processing motivated constraint on center embedding. This constraint has a wider application to different constructions in other languages. A second proposed universal is also shown to be false, and it is argued that the generalization is not significant. |
