汉语辖域关系的在线理解：以全称量化词与否定词为例

Processing scope relations in Chinese:An online study on universal quantifiers and negation

汉语辖域关系遵循"句法同构原则",但对其在线理解的时间进程及认知加工机制,学界则鲜有考察。本文采用自控步阅读及真值判断任务,通过操纵上文语境(偏向表层或逆序辖域),重点考察了主语位置上的全称量化词"每"和否定词"没有"之间辖域关系的在线理解。反应时结果表明,自"都"字起的大部分区域,偏向逆序辖域语境,先前语境偏向逆序辖域比偏向表层辖域的反应时间更短,而真值判断的反应时及正确率显示被试只接受表层辖域。逆序辖域的加工优势与"句法结构说"的预期相悖。该发现表明在线加工汉语辖域关系受句法原则、语义语境、基于工作记忆的认知因素的共同驱动,支持"平行交互说"和"启发策略领先一运算居二说"。

Unlike scope ambiguities present in English(e.g.,Every horse didn't jump over the fence),scope relations in Chinese(e.g.,meipi ma dou meiyou tiaoguo liba' every-CL horse all not jump fence') only allow surface-scope interpretation(i.e.,' None of the horses jumped over the fence').Whereas it has been generally acknowledged that Chinese strictly obeys the Isomorphism Principle(Aoun and Li1989;Huang 1982),it remains to be known how Chinese comprehenders process in real time scope sentences with preceding discourse contexts.The syntactic/structural model predicts that surface-scope interpretation predominates(Fodor 1983;Frazier 1987).In contrast,the Heuristic First,Algorithmic Second account predicts that heuristics-driven shallow processing takes precedence over deep-level processing that involves algorithmic computation(Dwivedi 2013).The parallel interactive account posits that scope processing is dependent on varying weights assigned to syntactic,semantic,pragmatic and cognitive factors(Kurtzman and MacDonald 1993).We tested these predictions by a self-paced reading experiment with truth-value judgment,using target sentences in Chinese with scope relations between the universal quantifier mei' every' and the negation meiyou ' not',as in meaning tongxue dou meiyou ca zhuozi,...' every-CL student all not clean table'.We manipulated preceding contexts that either(i) describe ' none did something'(hence biasing surface-scope reading) or(ii) describe ' not everyone did something'(hence biasing inversescope reading that is disallowed in Chinese).Thirty college students from a university in Shanghai participated in this study.They were asked to read context sentences presented as a whole,then read target sentences word-by-word,before pushing the yes-no button to judge whether the story stated in the target sentence was true or not.Our self-paced reading data showed that across most regions of target sentences(beginning from dou),response times were significantly shorter in inverse scope-biasing contexts than those in surface scope-biasing contexts(p's<.05).But truth-value judgment times and accuracy rates indicated that participants only accepted the surface-scope reading.The novel finding of inverse-scope processing advantage suggests that,contrary to what the syntactic/structural model predicts,online processing of scope relations in Chinese is driven by discourse contexts,event complexity,lexical semantics,as well as syntactic principles.Specifically,participants in the reverse-scope context didn't have to resolve event details,but in the surface-scope context,they were likely subject to similarity-induced interference in retrieving events(Gordon,et al.2004;Lewis,et al.2006).Overall,our results support the parallel processing model and the HeuristicFirst,Algorithmic-Second model.Further research with refined online measures is needed to further differentiate between the parallel processing model and the Heuristic-First,Algorithmic-Second model.