The Integration of Linguistic and Non-Linguistic Information in Second Language Sentence Processing.

Abstract

This dissertation investigates the integration of linguistic and non-linguistic information in the course of second language (L2) sentence processing. Having to process multiple sources of information simultaneously has been claimed to pose the greatest challenges in L2 learning (Clahsen & Felser, 2006; Sorace, 2011; Sorace & Filliaci, 2006), and research on the issue has the potential to shed light on the nature of learning in different populations, which, in turn, could contribute to a better understanding of the role of processing in language acquisition (O’Grady, 2015; Phillips & Ehrenhofer, 2015). The two sources of information, linguistic and non-linguistic knowledge, were operationalized by the definiteness distinction of English articles and real-world knowledge, respectively. To compare how different sources of information are integrated by L1 and L2 speakers, it was first necessary to determine whether both types of information were shared by native (L1) and second language (L2) speakers. The first experiment, implemented via a self-paced reading task, examined whether L1 and L2 speakers are sensitive to the mapping between definite noun phrases (NPs) and unique referents. It was shown that both populations exhibit this pattern of mapping, but that L2 speakers’ sensitivity to the relationship appeared one region later compared to L1 speakers. The second experiment, via a referent prediction task, shows that both L1 and advanced L2 speakers predict a unique referent at the cue of a definite article. What is noteworthy in this experiment is the behavior of intermediate L2 speakers, who predict a unique referent in response to an indefinite article numerically more often than to a definite article. The third experiment was an online norming survey to confirm that both L1 and L2 speakers have the same real-world knowledge. The types of world knowledge studied consisted of associations between two referents—for example, a doctor and a stethoscope, or a basketball player and a basketball. L1 speakers and L2 speakers of all proficiency levels were found to share the same real-world knowledge. The final experiment tested how the two sources of information were integrated incrementally online. A referent identification task measured the reaction time to stimuli when linguistic and non-linguistic information pointed to either the same referent or different referents. The results showed that L1 speakers integrate both linguistic and non-linguistic information incrementally and use both types of information to predict referents yet to be mentioned, but that L2 speakers did not use linguistic information in a native-like manner when non-linguistic information alone was sufficient to predict upcoming linguistic material. The findings suggest that non-linguistic information, operationalized as real-world knowledge in the current research, could be the key to accounting for certain differences between L1 and L2 development. Such findings have important implications for issues in both psycholinguistics and language acquisition research. In particular, L2 speakers’ reliance on realworld knowledge could be interpreted as an effort to minimize processing cost. By focusing their limited cognitive resources on an information source (world knowledge) that is acquired earlier than the relevant components of L2 grammar, is more familiar, and is thus easier-to-process, L2 speakers can maximize their processing efficiency. In contrast, it would not be as efficient to divide attentional resources over multiple sources of information. The dissertation concludes with suggestions for future research that compares adult L2 speakers and L1 children in terms of information integration to better understand what sets L2 acquisition apart from L1 acquisition.