Use of plausibility information and structural preference in resolving sentence ambiguity in Korean.

摘要

This study investigated online processing of temporarily ambiguous relative clause (RC) constructions in Korean to determine how multiple sources of information, such as syntactic preference and semantic plausibility, were employed in reading comprehension by three groups of Korean speakers with different L1 backgrounds: native speakers of Korean (NSs), heritage speakers of Korean (HSs), and adult second-language (L2) learners of Korean (K2s). HSs in the current study were Korean-English early bilinguals whose dominant language was English while K2s were international students and immigrants living in Korea. The main research focus in this study was to explore how these three groups of Korean speakers, who exhibited different levels of Korean-language proficiency, would integrate information on semantic plausibility and syntactic preferences while resolving temporary ambiguities. Sentence processing data from NSs were used to test existing major theories of sentence processing and comparable data from HSs and K2s were considered to determine the extent to which each group performed native-like sentence processing while resolving temporary ambiguities.;The plausibility of a noun as a direct object of either the embedded RC verbs or the main verbs was manipulated for the two types of RC constructions in Korean: the SR (Subject + Relative clause) construction with two objects and the SOR (Subject + Object + Relative clause) construction with one object and a required structural reanalysis. In addition, individual working memory (WM) capacity was measured through a reading span test (RST) in Korean to determine whether individual variations in WM capacity could influence the use of information on semantic plausibility and syntactic preferences during the resolution of temporary ambiguities in Korean.;The processing load of information on syntactic preferences and semantic plausibility and the final interpretations of the two Korean RC constructions were measured by using a self-paced word-by-word moving-window reading paradigm. The dependent variables were reading times and accuracy of responses to the end-of-sentence comprehension questions. The sample included 40 NSs, 40 HSs, and 42 K2s. In addition, 26 NSs participated in a norming study for the plausibility of the experimental materials. Reading times at each word position in a sentence and responses to comprehension questions were analyzed for further statistical analyses.;The results indicate that both semantic plausibility and syntactic preferences for a simpler analysis had considerable influence on the overall sentence-parsing process in Korean. In general, the results for NSs provide support for both syntax-first serial models and parallel multiple constraint-based models. However, the generally mixed results for the main effects of the plausibility bias and the interaction effect of plausibility and structural preferences on both sentence processing and final interpretations provide support more for parallel multiple constraint-based models than for syntax-first serial models. In addition, these results provide a clear indication of Good-Enough (GE) or shallow parsing in the ambiguity resolution process as well as the final interpretation for both native and non-native speakers of Korean.;HSs showed slower reading times than NSs because of their lower Korean proficiency and provided mixed results for native-like processing, whereas K2s showed less native-like processing patterns than HSs. In terms of the use of plausibility information during the ambiguity resolution process, both HSs and K2s showed native-like sensitivity to plausibility information for both SR and SOR constructions. In general, HSs were more sensitive to plausibility information than K2s, but K2s showed a stronger structural preference for the SR construction than NSs and HSs.;WM had a main effect on the ambiguity resolution process during the online reading process as well as during the interpretation process with some group differences in the effect. There were significant differences between the three groups in terms of their reading proficiency and WM capacity in Korean. WM had stronger main effects on many of the measurements in the current study for HSs than for NSs and K2s.;The results, including those for the use of plausibility information and the syntactic preference in relation to individual WM capacity during sentence processing by these three groups of Korean speakers, provide valuable insights into bilingual sentence processing and present a new way of examine existing models of L1 sentence processing. In addition, these results allow for analyses of competing syntactic and plausibility factors influencing the resolution of temporary ambiguities in Korean, which should provide a better understanding the nature of language-processing mechanisms.