LingLang Lunch (2/4/2015): Philip Hofmeister (Brown University)

Expectations and linguistic acceptability judgments

A growing and convergent body of evidence points to the role of expectations in online language processing and learning. This evidence includes data which indicate that processing efficiency for various sentential constructions can be improved by making them more expected (viz., more frequent) in a linguistic context (Wells et al 2009; Fine et al 2013). Here, I consider how expectations bear on acceptability judgments and, more specifically, shifts in acceptability judgment patterns. The hypothesis under consideration is that acceptability judgment responses reflect expectations based on previous experience. A prediction of such a hypothesis is that judgments for such constructions are mutable. In a series of acceptability tasks, I illustrate that participants systematically alter their responses over the course of the experiment, such that relatively unacceptable constructional variants improve with repetition. This holds across a range of data including sentences with case errors, resumptive pronouns, island violations, center-embeddings, and more. I will construe this to mean that judgments, like a variety of other response types, are sensitive to probabilistic factors and I will point to the implications of such findings for our understanding of grammatical change.