Decomposing subject retrieval: Diverging interference profiles for agreement and thematic dependencies in English
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
The resolution of long-distance dependencies is mediated by both grammatical knowledge and domain-general working memory systems. Much previous work has investigated how and whether such memory operations are fine-tuned to the grammatical requirements of various linguistic dependencies. The present study recenters this issue in the context of two subject-verb dependencies with distinct grammatical functions: agreement, involving the checking of matching morphosyntactic features, and thematic binding, involving the fulfilment of argument-structure requirements. In one speeded acceptability judgment experiment and three self-paced reading experiments, we examine the online and offline profiles of dependency resolution for agreement and thematic binding in English. Experimental items in each experiment utilized ditransitive alternations to introduce several distractors intervening between the subject-verb dependency. Our results indicate that agreement is highly susceptible to interference from distractors which are syntactically dissimilar to the retrieval target but morphologically compatible with the retrieval probe. However, thematic binding is more restrictive, with interference only arising from distractors maximally similar to the retrieval target. Together, these results motivate a view of subject-verb dependency resolution in which the distinct grammatical functions of agreement and thematic binding yield distinct retrieval operations, despite targeting an identical constituent. We discuss the consequences of these interference profiles for models of cue-based retrieval and interference, whereby subject retrieval is decomposed into multiple retrievals relativized to the features most reliable for executing distinct grammatical functions.