The Availability of Semantic and Syntactic Information and their Use as a Structural Predictor in the Real-Time Processing of L1 and L2 Korean
Eun Young Shin
Committee Members: Robin Morris (Co-Director), Anne Bezuidenhout (Co-Director), Elaine Chun, Mila Tasseva-Kurktchieva
The goal of this dissertation is to investigate the real-time sentence processing of Korean both by native speakers and learners in terms of the availability and the effectiveness of different types of information in the structural prediction. Built on the crosslinguistically contrastive findings between Japanese in Yoshida (2006) and Chinese in Hsu (2007), the parser in Korean is characterized in two aspects. The first aspect is whether the parser is sensitive to different types of information in on-line processing such as the semantic mismatch and the case markers. The second aspect is whether the parser makes a structural prediction based on inherently indirect cues available. In addition, the L2 parser from two language backgrounds is qualitatively compared in the same aspects.
Self-paced reading time experiments were conducted with L1 Korean speakers as well as L1 Japanese and L1 English learners of Korean. The construction under discussion contained a classifier construction and the relative clause, where the semantic match between a classifier and its closest candidate licensing noun as well as the case marker of the noun was manipulated.
The results showed that both L1 and L2 parsers were sensitive to the available semantic information as well as the case marker information in on-line processing. The semantic mismatch had an immediate impact while the case marker information exhibited a latent effect. As for the prediction effect, the semantic mismatch alone did not serve as an unfailing predictor of the relative clause in Korean whereas the semantic mismatch in Japanese (Yoshida, 2006) and the contextual cue in Chinese (Hsu, 2007) made a structural prediction possible. Instead, the combination of the semantic Match and the Genitive marker accelerated the processing of the upcoming relative clause in Korean. The predictive effect was not clearly observed at the expected locus of the effects in the L2 parser. In addition, there was a frequency effect at the classifier itself, which was amplified by the L2 parser.
Based on the findings, this dissertation supports the claim that the parser is universally incremental and predictive, but it further argues that the degree of the predictability and the types of information used for the prediction vary depending on the language.