This paper presents a discourse-functional account of English inversion, based on an examination of a large corpus of naturally-occurring tokens. It is argued that inversion serves an information-packaging function, and that felicitous inversion depends on the relative discourse-familiarity of the information represented by the preposed and post-posed constituents. The data moreover indicate that evoked elements and inferrable elements are treated alike with respect to inversion; both are treated as discourse-old information. Finally, it is suggested that discourse-familiarity correlates not with subjecthood, but rather with relative sentence position.

This article reports a nine-month longitudinal study of fifteen children who, at the beginning of the study, ranged in age from 3;10 to 4;11. In the course of the study we traced the development of principles of control in their grammars and tested hypotheses about the linguistic analyses underlying the observed nonadult interpretations of constructions such as Grover tries pro to jump over the fence, Big Bird tells Ernie pro to jump over the fence, and Ernie kisses Cookie Monster before pro jumping over the fence. The study also investigated the children's judgments of pronominal reference in sentences such as Ernie kisses Cookie Monster before he jumps over the fence. The results confirm a developmental sequence that is driven by lexical learning and changing structural analyses. They also provide further information about a period experienced by some children in which pronominal reference is constrained in a manner that is clearly related to control phenomena.

One motivation for nonlinear phonology is the potential for eliminating the devices of linear phonology needed to allow rules to apply to nonadjacent segments. Imposing hierarchical structure on the organization of features within a segment and allowing segments to be unspecified for certain features makes it possible to view apparently longdistance rules as rules operating between segments which are adjacent at a specified level, even though the segments are not adjacent at all levels of representation. This paper presents a theory of phonological adjacency requirements. Locality Theory is defined by a universal Locality Condition, which requires elements to be local within a plane, the Adjacency Parameter, which in turn allows rules to impose further constraints on the maximal distance between interacting segments, and by Transplanar Locality, which bans certain types of relations across featural planes. A survey of phonological processes demonstrates the generality of this theory across feature tiers.