Classifier Positioning In Chinese Relative Clauses And Its Underlying Reasons: Evidence From Spoken Corpus And Sentence-production Studies

Current models of language production make contrasting predictions regarding thestrategies speakers employ in language production, but existing research found mixedresults. It is not clear whether speakers are audience-designed (Clark&Carlson,1982;Temperley,2003) or speaker-oriented (Ferreira&Dell,2000; Race&MacDonald,2003)in utterance production. Thus, investigations of Chinese, a language typologically differentfrom most other languages, may further our understanding of language productionmechanism. In Chinese, a demonstrative+classifier (DCL) sequence can occur eitherbefore or after the head noun of a relative clause. By investigating the distribution patternof DCL positioning in subject-extracted relative clauses (SRCs) and object-extractedrelative clauses (ORCs) in a spoken corpus and two sentence production experiments, thisstudy aims to test the two language production models.The analyses of Chinese spoken corpus (Lu Yu You Yue) show that pre-RC classifiersprefer to occur in SRCs, and post-RC classifiers in ORCs. This corroborates the findings ofprevious studies based on written corpora. The results are consistent with theaudience-design model, suggesting that even in spontaneous speech, albeit in restrictedface-to-face discourse contexts, speakers are altruistic enough to provide early cues (i.e.,the DCL) for the RC structure, and to avoid potential lexical disruptions induced by theclassifier-noun mismatch to ease comprehenders’parsing difficulty.The behavioral data from two online word-based sentence production experimentsfurther test the two sentence production models. Experiment1manipulated RC types (SRCvs. ORC), with animate head nouns and animate embedded nouns. The results show anasymmetric pattern of DCL positioning and RC extraction types, consistent with the corpusfinding. Experiment2crossed RC types with Animacy (animate head nouns vs. inanimateembedded noun). The results not only support the findings in Experiment1, but also showthat animacy affects DCL positioning in RCs. In conclusion, both corpus and behavioral data show that Chinese speakers employaltruistic production strategies in on-line sentence production. The results are compatiblewith the audience-design model.