Processing wh-dependencies in L2 English : the role of L1 and working memory capacity

Abstract

This study investigates online processing of long-distance wh-dependencies in English by a group of Turkish- and Spanish-speakers of second language (L2) English in comparison to a group of native English speakers to explore whether end-state L2 speakers achieve native-like processing in the domain of wh-extractions. The study also examines the role of first language (L1) and working memory capacity (WMC) in on-line processing of whdependencies. To this end, speakers of L1 Turkish as well as speakers of L1 Spanish have been included in the study. Turkish is a wh-in-situ language. Nevertheless, it allows overt wh-movement via scrambling. Therefore, it provides an interesting testing case to verify the influence of scrambling in correctly accepting grammatical wh-extractions and rejecting ungrammatical wh-extractions with island violations. In this vein, the study focuses on the question of whether or not L1 Turkish speakers will be as accurate and as fast as L1 speakers of Spanish, a language with overt wh-movement in processing longdistance wh-extractions. Additional questions investigated in this study are (1) whether there is a subjectobject asymmetry in wh-extractions from finite and nonfinite clauses, and (2) whether there is a relationship between the WMC and sentence processing performance in the L2. An online grammaticality judgment task (OGJT) involved both grammatical whextractions from finite and nonfinite clauses and ungrammatical wh-extractions with island violations and this task was presented in two conditions, namely, the full-sentence condition and the self-paced word-by-word reading in the moving window condition. Both response accuracy and response latency (i.e. reading time) were measured in these conditions. To determine the WMC, all participants were tested on two online working memory (WM) tasks in English: 1) automated reading span (ARSAN) task, and automated operation span (AOSAPN) task. Spanish and Turkish participants also took the ARSPAN task in their respective L1. The accuracy results from the two conditions revealed that Turkish and Spanish speakers were as accurate as native English speakers in correctly accepting grammatical wh-extractions and rejecting ungrammatical wh-extractions with island violations, except for subject extraction from nonfinite clauses, and wh-extractions with that-trace violations. L2 learners were also similar to native speakers in reading patterns. Furthermore, there was no difference between Turkish and Spanish groups in comprehension accuracy and reading time for grammatical and ungrammatical whdependencies in L2 English. This suggests that L2 speakers whose L1 allows overt whmovement (i.e. Spanish speakers of English) do not outperform L2 speakers with a whin- situ L1 (i.e. Turkish speakers of English). The presence of overt wh-movement in scrambled sentences in the L1 (as in the case of Turkish) might be playing a role in accurate processing of wh-extractions in the L2. The results may also suggest that in the end-state L2, speakers achieve native-like processing irrespective of the syntactic properties of their L1.