Internal and External Factors in the Processing of Chinese Relative Clauses

• Main Authors: Pei-fen Du, 杜佩芬
• Other Authors: James H-Y. Tai
• Format: Others
• Language: en_US
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/74432457269927541358

碩士 === 國立中正大學 === 語言學研究所 === 101 === The present study examines the multiple factors that affect the processing of Chinese relative clause (RCs). Two internal factors (extraction types and RC length) involve the syntactic items which locate inside RCs, whereas two external factors involve the syntactic items which locate outside RCs (position of RC in matrix sentence and position of demonstrative + classifier Nage). A self-paced reading task was conducted in two experiments to observe real-time processing of these multiple factors. In Experiment 1, position of RC in matrix sentence and RC length were tested simultaneously. Kemper (1986) and Small et al. (2000) propose that English right-branching structure, a RC branches from object position, is easier to process than left-branching structure, a RC branches from subject position. Their results were attributed to a memory resource capacity theory (Just & Carpenter, 1992; Haarmann, Just, & Carpenter, 1997) that left-branching RC interrupts information of its matrix clause, and this construction increases additional memory demands to storage and computation. However, according to the definition of the branch direction (left vs. right), Chinese left-branching structure which is sentence-initial RC does not interrupt the information of its matrix clause, but right-branching structure does. The corpus-based studies (Pu, 2007; Wu, 2009) also show that the left-branching RCs outnumber the right-branching RCs in Chinese corpus. Thus, we predict that Chinese left-branching RC is easier to process than right-branching RC according to the above reasons. Both results (accuracy rates for comprehension questions and reading time data) in Experiment 1 were consistent with the corpus-based studies. Our findings support the memory resource capacity theory. On the other hand, Kemper (1986) also proposes that the English short-branching RC is easier to process than long-branching RC (i.e. RC + a temporal adverbial + PPs). Thus, Experiment 1 also examined Chinese branching length: short (i.e. a temporal adverbial + RC) vs. long branching (i.e. two temporal adverbials + PPs + RC). However, it shows inconsistent results between off-line (accuracy rates for comprehension questions) and on-line measures (reading time data). The off-line data seem to reflect locality-based hypothesis and the memory resource capacity theory, and reveal that the short- branching RCs were comprehend more accurately than the long-branching RCs. This findings suggest that there were increasing additional memory cost or integration in long branching. By contrast, the online data show totally different results that the participants read long branching faster than short branching. This online findings support the anti-locality hypothesis and memory facilitation hypothesis that long branching with more specific information facilitates RC processing, particularly at the region of the head noun. Experiment 2 focused on the external factor, the position of demonstrative + classifier Nage, to observe whether there is difference of RC processing between Nage in pre- and post-RCs. Sheng & Wu (2012) claim that Nage prefers occurring before subject extracted relative clauses (SRCs) and after object extracted relative clause (ORCs) in their corpus-based study and sentence production experiment. Their findings were interpreted within the Audience Design Model. This model suggests that speakers favor to provide an early cue Nage for their listeners (or audience) in pre-SRCs to make the ease of comprehension. By contrast, the speakers favor to use Nage in post-ORCs to avoid structural ambiguity for Nage in pre-ORCs. Beyond these pragmatic functions, we propose that there are default syntactic and semantic linkings (i.e. the canonicity of thematic order) to affect real-time processing. We predict that Nage_SRCs is easier to process than SRCs_Nage, whereas ORCs_Nage is easier to process than Nage_ORCs, and the results were in line with Shang & Wu (2012). Based on the effect of canonicity of thematic order (Small et al., 2000; Lin, 2012), the structures of Nage_SRCs and ORCs_Nage are compatible with the canonical thematic order like AGENT-ACTION-PATIENT, and thus the participants read faster at the regions of RC and head noun than those with the non-canonical thematic order (i.e. SRCs_Nage and Nage_ORCs). The findings also suggest that there is a delayed thematic role assignment in SRCs_Nage and Nage_ORCs, and this delay leads participants to increase additional processing costs. In this thesis, the integration of the previous corpus-based analyses and the self-paced reading experiments gives an insight into the grammar in language use and process. The distinction of internal and external factors helps shed light on the default syntactic and/or semantic linkings on processing of Chinese RCs.