Abstract: Idioms are fixed phrases and have a unified meaning. According to Multi-Constituent Unit (MCU) hypothesis, idioms maybe represented lexically as single representations during Chinese reading (Zang, 2019). There are two main structures of three-character idioms: modifier-noun (MN) and verb-object (VO) structures. Idioms with a MN structure are more like a long word due to the constraint of the modifier exerting over potential subsequent nouns than idioms with a VO structure, indicating that idiom structure might affect the processing of idioms. Furthermore, examining the role of idiom structure on its processing provides insight into the current theories of idiom representation. However, the previous research has not shown consistent results regarding this issue and the possible reason might be due to the familiarity of idioms not being carefully manipulated. The present study was to investigate whether idiom structure and familiarity exert an influence on idiom processing during Chinese reading. We manipulated structure (MN, VO) and familiarity (high, low) of idioms to clarify how the two variables affect idioms processing during Chinese reading. Idioms with a MN structure are comprised of a 1-character modifier and a 2-character noun (i.e., “贼+胆子”). Idioms with a VO structure are comprised of a 1-character verb and a 2-character object (i.e., “泼+脏水”). 128 idioms were selected as targets and embedded into 32 sentence frames. 104 students participated in the formal experiment and their eye movements were recorded when they read each sentence using Eyelink 1000 plus. Four files were constructed, with each file containing 32 sentences. Conditions were rotated across files according to a Latin Square design, and each participant read experimental sentences presented randomly from one of the four files. The 32 experimental sentences were mixed with 32 filler sentences, and 24 sentences were followed by yes/no questions. Besides, 12 practice sentences (including 4 question sentences) were presented at the beginning of the formal experiment. LMM models were carried out to analyze data. Results showed that the effects of familiarity were reliable, such that readers took significantly longer times processing low familiarity idioms than high familiarity idioms in all fixation time measures, indicating familiarity is an important factor during the processing of idioms. There was no reliable structure effect in first fixation duration, single fixation duration, gaze duration and skipping probability. However, the effect of structure was reliable in total fixation duration, a relative later eye movement measure, with MN structure associated with shorter time than VO structure, and this effect was modulated by idiom familiarity. Further analysis showed no significant difference between idioms with a different structure for high familiarity idioms, however, readers spent less total fixation duration on MN structure idioms than VO structure idioms for low familiarity idioms. In conclusion, our findings showed the structure of idiom affects the later processing of idioms during Chinese reading, and this effect was modulated by familiarity. Familiarity affects both early and later processing of idioms during Chinese reading. The current study has important implications for MCU hypothesis and theories of idiom processing. The results of current experiment support the hybrid model of idioms such that familiar idioms are represented as single units but unfamilar idioms are processed and modulated by their structure.

Key words: idioms, structure, familiarity , Multi-Constituent Unit hypothesis, Chinese reading

摘要： 为探讨结构和熟悉性在惯用语加工中的作用，本研究操纵惯用语的结构（偏正、动宾）和熟悉性（高、低），采用Eyelink 1000 plus记录104名大学生阅读包含惯用语的句子时的眼睛注视过程。结果发现：高熟悉惯用语的第一遍注视时间和总注视时间均短于低熟悉惯用语的；只有加工低熟悉惯用语时, 偏正结构惯用语的总注视时间短于动宾结构惯用语的。结果说明，结构只影响低熟悉惯用语的晚期加工，偏正结构比动宾结构惯用语语义整合更容易；熟悉性影响惯用语整个加工过程，高熟悉比低熟悉惯用语加工更快。实验结果支持惯用语的混合理论。

关键词: 惯用语, 结构, 熟悉性, 多成分单元假说, 中文阅读