Effects of working memory on structural competition in processing English present tense morphology by L1 Thai learners

Abstract

In the course of second language (L2) acquisition, L2 grammatical processing is often found to be dissimilar to that of native speakers, and nonnative processing difficulties have been attributed to various sources. In the realms of agreement processing, whether both first language (L1) and L2 linguistic systems are co-activated simultaneously within a bilingual mind during grammatical processing remains an active question in language sciences. This study examined individual differences in terms of working memory capacity and distance-based complexity in L2 agreement processing.
 
 In two word-by-word self-paced reading experiments, a total of 80 agreement-lacking Thai learners of L2 English and 80 native speakers of English read sentences involving English subject-verb agreement dependencies in two distance conditions. Distance was manipulated based on the dependency locality theory (Gibson, 1998, 2000), using semantically reversible English relative clauses (e.g., *The guy [that _ knows the driver/that the driver knows _] want to buy a new car). Two explanatory variables were crossed in a 2 x 2 design: distance (short-distance subject-extracted relative clause (SRC) vs. long-distance object-extracted relative clause (ORC)) and grammaticality of the subject-verb agreement (grammatical vs. ungrammatical). Stimuli in Experiment 1 involved singular subjects while those in Experiment 2 contained plural subjects; each consisted of 20 sentences, half grammatical and half ungrammatical, along with 40 distractors. The self-paced reading experiments were designed using the E-Prime 3.0 software package. A complex reading span task was used as a measure of working memory capacity. LexTALE scores showed L2 English proficiency to be upper-intermediate.
 
 In Experiment 1, linear mixed-effects modeling revealed that the native speakers and L2 learners were sensitive to agreement violation in both short-distance SRC ungrammatical and long-distance ORC ungrammatical conditions, shown by reading slowdowns. Their ability to show and maintain their sensitivity was, however, modulated as a function of working memory capacity and distance-based complexity. In Experiment 2, linear mixed-effects modeling showed that unlike the native speakers, who were able to show and maintain their sensitivity in both distance conditions, higher-span L2 learners showed longer reading times which were observed only in the short-distance SRC ungrammatical condition.
 
 These findings indicated that nonnative sensitivity to L2 agreement violation tended to be modulated by individual differences in terms of working memory capacity and distance-based linguistic complexity, particularly in agreement processing that involved multiple unique-to-L2 features, i.e., plural and third-person singular morphology, in long-distance agreement dependencies. To account for L2 learners' agreement processing difficulties, the findings suggested that L2 learners may labor under parallel activation, whereby working memory capacity is insufficient to resolve long-distance agreement dependencies, where the linguistic environment was more complex, thus resulting in reduced sensitivity to L2 agreement violation. The research findings were congruent with the L1-L2 structural competition account (Austin et al., 2015; Trenkic et al., 2014; Trenkic & Pongpairoj, 2013).