The role of working memory in the processing of reflexives

Abstract

We report results from two eye-movement experiments that examined how differences in working memory (WM) capacity affect readers' application of structural constraints on reflexive anaphor resolution during sentence comprehension. We examined whether binding Principle A, a syntactic constraint on the interpretation of reflexives, is reducible to a memory friendly “recency” strategy, and whether WM capacity influences the degree to which readers create anaphoric dependencies ruled out by binding theory. Our results indicate that low and high WM span readers applied Principle A early during processing. However, contrary to previous findings, low span readers also showed immediate intrusion effects of a linearly closer but structurally inaccessible competitor antecedent. We interpret these findings as indicating that although the relative prominence of potential antecedents in WM can affect online anaphor resolution, Principle A is not reducible to a processing or linear distance based “least effort” constraint.

Keywords:

• Anaphor resolution
• Working memory
• English

Acknowledgments

This research was supported by an Economic and Social Research Council grant (RES-000-22-2508) to the second author. We thank the audience at the Formal vs. Processing Explanations of Syntactic Phenomena workshop (University of York, April 2009) and two anonymous reviewers for their thoughtful and constructive comments.

Notes

¹Gender congruence between the reflexive and the structurally accessible antecedent was not manipulated in this study. This study also compared inaccessible antecedent effects in native compared with nonnative comprehension, and while nonnative English speakers were affected by the gender of the inaccessible antecedent, native English speakers were not.

²Although this seems clear for the processing of co-argument reflexives, we leave open the question of whether or not this also applies to reflexives in non co-argument relations, such as reflexives inside so-called picture noun phrases (see Runner, Sussman, & Tanenhaus, 2003; Kaiser, Runner, Sussman, & Tanenhaus, 2009). Indeed, the question of whether binding Principle A also applies to cases when the reflexive and antecedent are not co-arguments has been the matter of some debate in theoretical linguistics (for discussion, see Pollard & Sag, 1992; Reinhart & Reuland, 1993; Reuland, 2001).