关键词: 中文阅读, 边界范式, 词类预视效应, 眼动控制模型

Abstract:

A controversial issue in eye movement research investigating reading concerns whether high-level information (e.g., syntactic category information) of parafoveal words can be obtained prior to fixation. Researchers have demonstrated that readers could stably extract syntactic category information from the parafoveal words during English reading, and such findings are in favor of parallel graded processing model (e.g., SWIFT model). Unlike English, Chinese are not rich in inflectional or derivational indicators to specify words’ syntactic properties. For example, there is no inflection with the verb 包装 (pack) no matter whether this action is going to happen or has already happened, nor whether this word is used as a verb or a noun. Therefore, parafoveal syntactic cues in Chinese may be limited relative to English, meaning that the extraction and use of such information when making syntactic commitments may be delayed. The present study was to explore whether parafoveal word class information could be extracted during Chinese reading.
Using the boundary paradigm (Rayner, 1975), we manipulated the syntactic category (word class) of preview words to form three conditions: identical preview, preview with the same word class as the target word, preview with the different word class as the target word. 120 college students participated in the experiment. Eye movements were recorded using an Eyelink 1000 eye-tracker, and the recorded results (reading time and fixation probability) are the dependent variables. Participants were asked to read 45 sentences and answered a multiple-choice comprehension question if any (about 30%). Although participants were only exposed to each target word once, all sentences appeared in all preview conditions across three counterbalanced lists.
The results showed that readers spent similar duration when they fixed the target word regardless of whether the word class of the preview word in the parafovea was same as the target word or not, and there was no significant difference in fixation probability between them. Supplementary Bayesian analysis supported the null hypothesis. Moreover, all fixation durations were significantly shorter for the identical condition than for the other two previews, skipping rate was higher for identical condition, and regression probabilities were lower for identical condition.
The findings in the present study suggest that it is difficult for Chinese readers to extract high-level syntactic category information from parafoveal words. Our results provide support for the sequential attention shift model (e.g., E-Z reader model), which in favor of that high-level information processing occurs in the integration stage (I). The results provide evidence for improving computational models of eye movements about how to process lexical high-level information during reading.

Key words: Chinese reading, boundary paradigm, word class preview effect, computational models of eye movements