中文句子中双音节兼类词句法分析历程初探

以既可用作名词，又可用作动词的汉语双音节兼类词为材料，采用移动窗口范式，考察了汉语句子中名词倾向性高和低两类兼类词句法分析的详细历程。结果发现，两类兼类词与各自的控制词之间，均未出现阅读时间上的显著差异，因此并不支持延迟模型所提出的兼类词句法角色延迟指派的观点。实验还发现，名词倾向性高的兼类词引起了更大的加工困难。此外，即便是名词倾向性低的兼类词，被试也仍为其指派名词的句法角色。花园路径模型和制约满意理论均不能单独和完整解释上述发现。文章指出，实验结果支持一种既不同于花园路径模型，又不同于制约满意理论的权变的理论。阅读中遇到兼类词时，读者能立即为其指派句法角色。这种指派过程既可以利用概率制约信息，也可以在句法分析原则指导下进行。句法分析原则是否起作用，取决于概率制约信息的制约强度

THE PARSING OF DISYLLABLE WORDS WITH SYNTACTIC CATEGORY AMBIGUITIES IN CHINESE SENTENCE READING

In Chinese language, there are some disyllable words, which are temporarily and syntactically ambiguous between noun and verb when being embedded in sentences. They are called words with syntactic ambiguities (WWSAs). Both high-noun-biased WWSAs (H-WWSAs) and low-noun-biased ones (L-WWSAs) were included in the present study for investigating their parsing processes and testing the theories of sentence processing. Thirty-six undergraduate students were asked to read sentences word-by-word as quickly as possible, without sacrificing their comprehension of each sentence. It was found that there was no significant difference of reading time between WWSAs and their control words, which was inconsistent with the prediction of the delay model that proposed that during the delay period at the onset of the ambiguity, very little analysis of the ambiguity takes place. It was also found that compared with their own control words respectively, L-WWSAs led to the longer reading time at two (i.e., the first and the third) regions of disambiguity, while H-WWSAs led to the longer reading time at three (i.e., from the first to the third) regions of disambiguity. These results showed that greater garden-path effects occurred for the H-WWSAs than for the L-WWSAs. Thus, the assigning processes of syntactic roles of WWSAs were influenced by their degree of noun biasing, which is predicted by the constraint satisfaction theory, but not by the garden-path model. In addition, subjects assigned the syntactic roles of noun even for L-WWSAs. It indicated that the parsing of WWSAs can be guided by the minimal attachment principle proposed by the garden-path model, when the degree of noun biasing of WWSAs can not provide any constraining cues for the resolution of syntactic ambiguities. All these results suggest that neither the constraint satisfaction theory nor the garden-path model can explain the parsing processes of WWSAs very well. Alternatively, readers may assign specific syntactic roles for WWSAs immediately, either by accessing probabilistic constrains information or by accessing specific parsing principles, depending on the strength of probabilistic constrains.