A Behavioral and Electrophysiological Investigation of Different Aptitudes for L2 Grammar in Learners Equated for Proficiency Level

Abstract

The declarative/ procedural (DP) model (Ullman, 2004), a neurocognitive model, posits that for late-learned second languages (L2s), both lexicon and grammar initially depend on declarative memory. With increased exposure, the DP model predicts that aspects of grammar become dependent on procedural memory. The DP model differs from modular accounts (Schwartz & Sprouse, 2000), emergentist models (Ellis, 2008; MacWhinney, 2008), and other models that invoke the declarative/procedural distinction (DeKeyser, 2007; Paradis, 2004).

It was hypothesized that internal and external variables would mediate the involvement of declarative and procedural memory in L2. First, it was predicted that learners would display individual differences in declarative and procedural memory abilities (Eichenbaum & Cohen, 2001; Fernandez et al., 2003; Seger & Cincotta, 2005b; Sirrien et al., 2007; Ullman, 2005). In addition, it was predicted that working memory would constrain involvement by the two memory systems (Cochran & colleagues, 1992, 1999; Foerde et al., 2006). Finally, it was predicted that explicit instruction would encourage reliance on declarative memory, while implicit instruction would encourage reliance on procedural memory (Fletcher et al., 2005; Morgan-Short, 2007; Poldrack et al., 2001). Thus, this investigation focused on how moderating variables affect the distinct contributions of declarative and procedural memory to L2 at different stages of proficiency.

Twenty-nine adult participants' declarative, procedural, and working memory abilities were assessed before random assignment to explicit (N=15) or implicit (N=14) instruction in an artificial language. Proficiency was assessed once learners reached low and advanced benchmark levels, ensuring all had obtained a comparable baseline level. L2 performance was assessed with grammaticality judgment tasks, while neurocognitive processes were assessed with event-related potentials (ERPs).

There were several conclusions. Results for declarative memory suggest that CVMT scores predicted L2 performance across learners, but that type and extent of involvement was mediated by learning condition and proficiency level. Behavioral and ERP results for procedural memory suggest that implicit, but not explicit, instruction promoted proceduralization of grammar at advanced proficiency. Learners with high scores on the Weather Prediction Task, which assessed procedural memory abilities, displayed a different pattern of results from learners with low scores. Additional analyses indicated that attention and short-term memory capacity displayed opposite effects on declarative and procedural memory reliance in each condition. Finally, hierarchical linear regression indicated that combining Modern Language Aptitude Test scores (MLAT; Carroll & Sapon, 1959) with a measure of declarative memory better predicted low and advanced L2 proficiency than MLAT scores alone, but only in the explicit condition. Adding a measure of procedural memory did not improve the model in either condition for any structure. Findings suggest learners relied on different cognitive abilities for comparable levels of L2, as mediated by ability differences and instructional conditions.