ABSTRACT
This chapter focuses on research questions involving multiple levels of linguistic representation at the interface of morphosyntax and information structure. After introducing key concepts in the paradigm of Probabilistic Syntax, the chapter presents two usage-based case studies on constructional variation: the first study explores variation in word order, passive or active voice and information structure, using corpus data annotated for syntax and information status. Results show a strong, quantifiable interaction of constraints affecting constituent order in English, depending on voice, givenness and genre, as well as ranked constraint violation costs which allow untypical word orders only under infrequent circumstances across multiple layers. The second case study develops a highly accurate, multifactorial, predictive model of agreement variation in notional anaphora, that is, reference to nouns such as ‘committee’ using singular ‘it’ or plural ‘they’. Based on the largest dataset to date with very rich annotations, a large number of factors turn out to influence the alternation, some of which have not been described in the literature to date.
