优势语序还是优势解读?利用ERP考察汉语双论元歧义句的解歧过程

A Preference for Word Order or Animate Actor? An ERP Study on Processing Ambiguous Verb-final Constructions in Chinese

在听觉ERP实验中,被试听到NP1无生命NP2有生命V和NP1有生命NP2无生命V两种序列,句末动词将论元解歧为施事者或受事者。实验结果表明,该解歧过程受到优势语序(OSV)和优势解读(有生命施事者)的影响:在早期加工阶段,同时违反优势语序和优势解读的条件产生一个明显的前区负波;在晚期加工阶段,只是违反优势解读的条件产生了类似的前区负波。可见生命性在汉语句子加工中具有重要地位,其效应的时间进程为完善语言加工模型提供了启示。

The present auditory event-related brain potentials（ERPs） study investigated online argument interpretation in ambiguous verb-final constructions（NP1-NP2-Verb） in Mandarin Chinese, based on two aspects： a preference for OSV order and a preference for animate Actor arguments. The former is derived from the hypothesis that the processing system tends to prefer a simpler structure, and the latter is derived from the hypothesis that an animate Actor is preferred in transitive sentences. We thus adopted a 2 × 2 experimental design（word order × animacy of the arguments） to test our hypotheses and examine the language-specific role animacy plays in processing Chinese.Participants listened to sentences with NP（animate）- NP（inanimate）- Verb or NP（inanimate）- NP（animate）- Verb sequences embedded in minimal contexts. The sentence-final verb disambiguated the stimuli towards either an SOV or an OSV order. After listening to each context-stimulus pairs, participants performed an acceptability task and a subsequent comprehension task. While their accuracy was high for all conditions in sentence comprehension, there was an interaction of OSV order and animacy in sentence acceptability. Similar observations were also obtained from the ERP results（n = 25） at the disambiguating verb, revealing a strong interaction of word order and animacy. Between 450 and 700 ms post onset, we observed an anterior negativity for the condition violating both preferences vs. the condition fulfilling both preferences, i.e. S（inanimate）- O（animate）- Verb vs. O（inanimate）- S（animate）- Verb. The other two conditions, which satisfy only one of the two preferences, did not differ from one another but engendered a smaller anterior negativity in comparison to the most preferred condition, i.e. S（animate）- O（inanimate） – Verb / O（animate）- S（inanimate）- Verb vs. O（inanimate）- S（animate）- Verb. In the time window between 700 and 850 ms, we again observed an anterior negativity for the condition violating both preferences as opposed to that fulfilling both preferences. Interestingly, however, the other two conditions showed a reversed effect in comparison to each other, i.e. an anterior negativity for OSV vs. SOV. The reversed effect suggested that animacy finally overrode the OSV preference. In other words, animacy guided the processing system to fulfill the preference for animate Actors even though that would result in a non-preferred order（SOV）. Based on the topography of the effect, we argue that the negativity reflects reanalysis cost at the sentence-level rather than a pure lexical difference of the disambiguating verbs or a plausibility difference of the sentences. In sum, our data suggests that the preference for OSV order is only applied online when it is supported by animacy. When either of the two preferences is not satisfied, the processing system may not be committed to the OSV preference（450~700 ms）; instead it even choose a reversed order based on animacy（700~850 ms）. In contrast to languages previously examined, the present findings from Chinese are particularly interesting： animacy does not interact with the word order preference in other languages, whereas in Chinese, animacy could override the word order preference. Thus, the present data provides important evidence for the language-specific role animacy plays in processing Chinese.