#### Jon Sprouse · Island effects, gradience, and the role of formal experiments in theoretical syntax

**Linguistics Colloquium**

In any given theoretical syntax paper, there are often two types of informal experimental work on display - exploratory work that gathers a large number of data points together for the creative exercise of theory construction, and confirmatory work that seeks to distinguish between two or more competing theories through well-defined empirical predictions. My impression of the experimental syntax literature (and my own work in particular) is that formal experiments have proven useful for confirmatory work (as expected), but have made relatively little contribution to the greater exploratory enterprise that really drives theoretical syntax. In this talk I'd like to ask what it would look like for experimental syntax to contribute to that enterprise.

One relatively easy answer is to simply re-run earlier informal explorations using more formal methods to see what the empirical landscape looks like from a formal experimental vantage point. It turns out that there are a few theoretically interesting surprises when one does that, so in the first part of this talk I will present the results of such a study for around 40 different configurations of island effects in English (approximating the size of data collection in typical theoretical papers), and discuss the consequences for the theory of island effects.

A potentially more interesting question is whether there is information revealed by formal experiments that is relevant for the big questions that motivate syntactic theory in the first place (What is the nature of linguistic computations? What is the scope of the language acquisition problem?, etc). I think that the quantitative effect size information revealed by formal experiments might be relevant here. Gradient effect sizes provide a potential challenge to the view that linguistic computations and constraints are binary categorical. The question is whether these gradient effect sizes can be captured within a performance system that assumes a binary categorical grammar, or whether these gradient effect sizes necessitate a gradient grammar (with all of the complications that gradient grammars raise for the big picture questions). Therefore in the second part of this talk I will discuss the pattern of effect sizes that we see across the 40 or so configurations of island effects, and explore what it would take to account for that pattern while maintaining a binary categorical syntactic theory.