| Research suggests that adult second language learners have difficulty processing certain argument structures in their L2. In particular, it has been shown that L1 English learners of Spanish are not at first successful in processing preverbal clitic structures in Spanish such as in (a), (e.g. Liceras, 1985; VanPatten, 1984). [special characters omitted] These structures are relatively common in Spanish, and being able to process them properly is essential to becoming proficient in the language. The present dissertation aimed to determine the factors involved in how L2 learners process Spanish structures with preverbal object clitics and postverbal subjects. I considered three areas that possibly contribute: (1) the characteristics of the L1 (i.e. transfer), (2) the proficiency of the learners (previous studies typically tested only low proficiency learners), and (3) individual differences in cognitive abilities such as working memory capacity (WMC) and inhibitory control/executive attention.;Testing (1), I compared the performance of L1 English speakers to L1 speakers of a language that is typologically like Spanish, Romanian, on two sentence processing tasks (one listening and one reading) in Spanish. Romanian has a similar system of pronominal clitics to Spanish, allowing for structures with preverbal clitics and postverbal subjects, as in (b). [special characters omitted] Testing (2), I included participants of a wide range of L2 proficiency levels, which I assessed using an independent measure of proficiency (adapted from the DELE exam). Testing (3), each participant performed the Letter-Number Sequencing test of working memory (Wechsler, 1997) and the Flanker Task, which assesses executive attentional abilities (Luk, 2008). In all, 71 L1 Romanian and 65 L1 English L2 Spanish learners and 36 L1 Spanish monolingual controls participated in the two sentence processing experiments.;With the use of logistic regression with mixed-effects models, I was able to include both discrete (e.g. first language) and continuous (e.g. proficiency level) variables, include all trials and participants with missing trials, and thus, provide a more accurate picture than other more traditional methods (see Jaeger, 2008 for discussion) of the relative contributions of these factors on learners' L2 processing of this challenging structure. Specifically, I found main effects on accuracy for proficiency and condition in both tasks, indicating the strength of the role of these two factors, the importance of testing different proficiency levels, and the difficulty of the clitic structures for all L2 learners. Additionally, there was a main effect for L1 in the listening task, but not in the reading task, and a main effect for working memory in the reading task, but not in the listening task. These results suggest that the two modalities place differing demands on the participant, and that neither L1 nor working memory play a strong determining role in successful processing of the present structure. In addition, there were significant interactions between proficiency and condition in both tasks, such that the more proficient L2 learners were more accurate on the OVS condition, and significant interactions between working memory and condition, such that the L2 learners with higher WMC were less accurate on the OVS condition, a vexing result that requires further research. Most provocative was the result that there was no interaction between L1 and condition, indicating that although the L1 Romanians were more accurate than the L1 English participants on the sentence processing tasks on the whole, this improved accuracy was not specific to clitic structures. Based on these results, transfer appears to play only a small role in L2 processing, and when it occurs, may not be structure-specific. |
