Information, Truth, Structure, and Sound

Abstract

A persistent element of the Principles & Parameters framework is the T-model of grammar. The strong claim of this model is that phonological rules cannot refer to semantic entities, and vice versa. However, apparent interactions between prosody and semantic interpretation provide a prima facie challenge to this model. I propose a theory of Information Structure (IS) and its related interfaces that is sufficient to account for apparent semantic-prosodic interactions while still satisfying the strong requirements of the T-model. This proposal hinges on three inter-connected sub-proposals.First, I defend a representation of IS based on two kinds of primitives, partitions and instructions. The partitions are nested structures derived from sub-trees of LF. The instructions are functions that operate on the partitions to specify their connection with the discourse. Interactions between these primitives account for apparent truth-conditional effects of IS.Second, I propose that IS partitions are generated in the syntax via the mechanism of derivational phase, or cyclic Spell-Out. I claim that phases are determined by interface conditions, and that IS provides some of those interface conditions, such that some phases will be isomorphic with IS partitions. Not only does this theory provide appropriate structures for IS, but it does so without violating the T-model, and it manages to leave previous work on phases intact.Third, my phase-based proposal provides an indirect correlation between phonological stress and IS. Because stress rules apply cyclically at each phase, and because IS dictates some phases, stress and IS will naturally synchronize, without needing to stipulate a connection between emphasis and information. This has the added benefit of deriving so-called "narrow" or "contrastive" focus by the same set of principles as "default" or "informational" focus.Finally, my combined theory accounts for recalcitrant data involving the connection between prosody, information, and Diesing's (1992) Mapping Hypothesis in the interpretation of indefinites, and involving the link between prosody and underlying position in data from Bresnan (1971). The end result is a unified theory of IS and its interfaces, which maintains the T-model architecture and represents a streamlined theoretical and empirical improvement across several domains.