Abstract: Among thousands of words, native speakers can choose the right collocations to express the corresponding meanings in milliseconds. Existed theories cannot perfectly explain the mechanism behind it. It remains unclear how collocations are stored and extracted in human brain. Based on lexical decision task, our research revealed that high-frequency collocations significantly facilitated word recognition in Chinese native speakers. Surprisingly, high-frequency collocations in reversed order also moderately facilitated word recognition. However, non-collocations did not facilitate word recognition, even though they were semantically similar with the collocations. These results indicate that collocations are closely encoded in memory, regardless of its semantics. Furthermore, the connection between collocation words may be bi-directional and anisotropic, while the strength of such connection depends on the frequency of co-occurrence. Our findings provide solid evidence to the collocation research, and shed light to the principle of collocational encoding in human brain.

Key words: word recognition, Chinese, collocation, synonym, word order.

摘要： 采用词汇判断任务，探究汉语词汇识别中搭配信息的作用。研究表明：（1）汉语母语者对搭配信息敏感，高频搭配能显著促进词汇识别；（2）与高频搭配在语义关系上高度相近的词对词汇识别无促进效应；（3）以逆序形式呈现的高频搭配亦能促进词汇识别，但其效应弱于正序。本研究为词汇搭配知识的存储和加工方式的理论建构提供了依据：互为搭配关系的词语在词汇层面建立了具有方向性的联结，其联结强度取决于语言使用中的搭配共现频率；而在对搭配信息的前词汇加工过程中没有语义作用的参与。

关键词: 词汇识别, 汉语, 搭配, 近义, 词序。