Working Memory Effects in the L2 Processing of Ambiguous Relative Clauses

Abstract

This article investigates whether and how L2 sentence processing is affected by memory constraints that force serial parsing. Monitoring eye movements, we test effects of working memory on L2 relative-clause attachment preferences in a sample of 75 late-adult German learners of English and 25 native English controls. Mixed linear regression analyses find effects of reading span on attachment preferences across tasks in the L2 group. In addition, relative-clause attachment was modulated by slowdowns in lexical processing, which indicates that difficulties in word-level processing affect L2 parsing behavior. Nonnatives who were matched in capacity to native speakers showed target-like syntactic processing. In terms of the interaction of capacity and structural parsing preferences, L2ers pattern similarly to natives, which supports continuity approaches to L2 processing.

ACKNOWLEDGMENTS

I would like to thank Lena Holderer for help with data collection. Thanks are also due to audiences at GALANA 5 as well as the CLS colloquium at Penn State University where earlier versions of the manuscript were presented. Finally, I extend my thanks to the Editors of Language Acquisition and four anonymous reviewers whose constructive criticisms significantly helped improve the paper. All remaining shortcomings are my responsibility.

Notes

1. ¹In Havik et al. (2009), WM effects could be found in a task where L2 participants carried out an additional truth-value judgment task after reading sentences. Similarly, Williams (2006) finds working-memory effects in L2 reading only when L2 readers carry out an additional memory-probe task on top of reading sentences. These findings indicate that L2 comprehension is moderated by capacity effects when L2 readers are provided with an extra task that imposes additional memory loads (for discussion, see Roberts 2012).

2. ²Note that (5) and (6) also differ in the type of disambiguation, i.e., number marking in (5) versus gender marking on reflexives in (6). These differences may lead to processing differences between these sentences on top of locality effects. However, monolingual processing research shows that native speakers evince comparable attachment preferences for both types of disambiguation, so that differences between (5) and (6) likely predominantly reflect effects of locality.

3. ³With respect to the locus of the effects, similar spillovers to the following segment have also been observed in native speakers (e.g., Traxler 2007).