Ming Xiang (Chicago)
A great amount of sentence processing work has focused on revealing how the parser incrementally integrates each incoming word into the current linguistic representation. It is often explicitly or implicitly assumed that the representation preferred by the parser would determine the ultimate interpretation of the sentence. The current study investigates whether the interpretive bias in sentence comprehension necessarily tracks the parsing bias. Our case study, the Mandarin wh-in-situ scope dependencies, presents a misalignment between the local parsing decisions and the global interpretative decisions.
For Mandarin wh-in-situ constructions that involve scope ambiguity, e.g. two [spec, CP] positions c-command one wh-in-situ phrase, eye movement reading measures and acceptability judgments both showed a locality bias in parsing, such that the local scope dependencies are much less costly to process than the non-local ones. The non-local scope dependency is difficult to obtain when there is an intervening scope position available. The locality bias remains robust even when anticipatory cues are introduced to facilitate the non-local scope. However, when a truth-value judgment task is used to probe which scope, local vs. non-local, is more preferred, there is an anti-locality bias, such that the interpretation compatible with the non-local scope is more preferred.
We propose a Bayesian pragmatic inference model to account for these findings, following the Rational Speech Act framework (RSA, Goodman & Frank, 2016). Under this model, the seeming conflict between parsing and interpretation will ultimately disappear because in the proposed model parsing preferences will be naturally embedded under the pragmatic reasoning process to derive the ultimate interpretation. The currently study therefore makes novel contributions, both empirically and theoretically, to address questions about the relationship between parsing and interpretation. The current study also has implications about the standard methodologies in linguistics research, such as acceptability judgments and truth value judgments.