An information-based theory of topics and grammatical relations.

This dissertation proposes a formal semantic characterization of topichood and an account of the relationship between Topic and core Grammatical Relations. The theoretical framework employed is a form of HPSG (Pollard & Sag (1994)). The notion of Topic has been widely invoked in descriptions both of sentence structure and of intersentential discourse relations. Despite this a formal characterization of this notion is lacking in the literature. It is proposed here that Topics should be seen as predication targets at an underlying semantic level, and that the Topic-Comment relation is analogous to that between possible worlds (situations) and the propositional contents which they support. A Topic is interpreted as a point whose location has to be fixed in some conceptual space formed by the Comments, and this metaphor is extended to the overall Topic of a discourse sequence. Formally, it is suggested that Topics and Comments can be treated as the points and open sets respectively of a topological space. It is claimed that this captures well-known semantic restrictions on which NPs can be made Topics of a sentence. The proposed treatment is also extended to intersentential Topic relations. This account of Topics is made the basis of a revision to the relational hierarchy, which underlies many relational theories of grammar. It is proposed that basic predicates in language are maximally binary and sensitive to topichood, their initial Subject being the default predication target or Topic. Predicates of greater valency are treated as composite, and the effects of the relational hierarchy are derived from rules governing the process of composition. A number of cross-linguistic phenomena are examined which bear on the relationship between Topics and core Grammatical Relations, including the double Subject constructions characteristic of Japanese and other East Asian languages, the clitic doubling of Objects which is an areal phenomenon of the Balkans, and the so called "Object agreement" of Amharic. Finally a chapter is devoted to the nature of Indirect Objects, which are argued (against standard views) to rank above Direct Objects. It is claimed that with this approach an important part of the relational basis of syntax can be derived, without losing descriptive accuracy, from the proposed treatment of predication.