Syntax, interfaces and processing in native language attrition
The linguistic behaviour of bilingual speakers who have had prolonged exposure to a second language (L2) is different from that of monolinguals. This thesis investigates the extent and source of attrition effects by comparing language representation and processing in adult late bilinguals undergoing native language (L1) attrition and their monolingual counterparts. Based on the previous observation that structures that are sensitive to discourse-pragmatic conditions are vulnerable to attrition (Sorace 2011, Sorace & Filiaci 2006), the thesis examines: i) whether the difference between attrited and non-attrited speakers in L1 use is restricted to structures whose distribution is grammatically underspecified; ii) whether the difference is due more to underspecification of mental representation or to on-line processing difficulties; and iii) to what extent the difference is a consequence of transfer from L2. The case investigated in this study is L1 attrition by Korean immigrants who have lived in an L2 (English or Japanese) environment for a period of 6 to 25 years. Two L2 groups and one monolingual control group were tested on two different types of phenomena in Korean: core binding of the reflexive caki whose felicity is determined by grammar (Experiment 1), and the attachment of the plural suffix tul whose felicity is underspecified by grammar (Experiment 2). Experimental data were collected using an on-line methodology (a self-paced reading task) as well as an off-line one (acceptability judgement task) in order to identify the locus of any non-convergence between attrited and non-attrited speakers with respect to the investigated phenomena. Results from the experiments showed that attrition had an impact on both grammatically specified and underspecified structures, but to a different degree. With respect to core binding of caki, attrited Korean speakers diverged from the monolingual norm in the on-line reading task but not in the off-line judgement task, indicating that their representation of caki-binding was intact. With respect to tul-attachment, on the other hand, the attriters displayed divergence in both the off-line and on-line tasks, indicating that their representation of appropriate conditions for tul, as well as their real-time processing of the conditions was affected due to long-term exposure to L2. In both caki-binding and tul-attachment, the attriters’ non-native performance was largely attributable to influence from their L2. However, the attriters’ divergence also seemed to be attributable, at least in part, to inefficient executive control of two languages. Regarding tul-attachment, the results demonstrated that the distribution of tul in unattrited Korean is regulated by several factors, including animacy, number-specificity and distributivity, and thus the acceptability of tul is largely gradient, rather than categorical. The results also provided evidence for an ongoing change in the distribution of tul and suggested that the change is accelerated by attrited speakers living in an L2 English environment.