Wh-dependencies in English and Japanese: a test case for experimental syntax

Jon Sprouse

Department of Cognitive Sciences

University of California, Irvine

While there is a general intuition in the field that formal acceptability judgment experiments (a.k.a. experimental syntax) may be helpful for syntactic theorizing, it is not yet clear what role these experiments will actually play in the development of syntactic theories. The obvious answer is that these experiments should yield new data for syntactic theories; however, informal acceptability judgments have proven to be overwhelmingly correct, and extremely sensitive (Featherston 2008, Myers 2008), such that “new data” are few and far between. In this talk, I would like to pursue the idea that one possible benefit of experimental syntax is that it allows us to probe the inherent source-ambiguity of acceptability judgments: judgment “penalties” can be caused by either a violation of a grammatical constraint or the cost of a parsing process. I’d like to suggest that this source-ambiguity might give us the “new data” that we would like. By asking whether the acceptability penalty could have a processing-cost source, we can begin to figure out whether the similarities/differences that we find between constructions are due to the grammatical operations and constraints necessary to evaluate the representation, or whether they are due to the parsing processes necessary to build the representation in real time.

As a case study, I will present a series of acceptability judgment experiments designed to investigate to what extent wh-dependencies in English and Japanese (wh-displacement, multiple wh-questions, wh-in-situ) may be formed by the same grammatical operations. We find that wh-displacement and multiple wh-questions in English are sensitive to the length of the dependency, while wh-in-situ in Japanese is not. Furthermore, we find that multiple wh-questions in English are sensitive to a subset of the island effects that wh-displacement is sensitive to, while wh-in-situ is sensitive to none. These effects suggest that English wh-displacement and multiple-wh dependencies form a natural class to the exclusion of Japanese wh-in-situ dependencies. To account for these effects, we first propose a mapping between dependency type and grammatical operations; however, this mapping leaves some of the results unexplained. We then propose a second mapping from grammatical operations to parsing processes (and their associated costs) that includes a backward licensor search process for multiple-wh dependencies similar to the forward search for a gap location in overt wh-movement dependencies (the active filler strategy). Overall, our results suggest that formal acceptability judgment experiments can not only further syntactic theories, but may also facilitate a closer interaction between offline grammatical theories and online processing theories.