对汉字字形识别层次模型的实验验证

该研究采用Ｊｏｈｎｓｔｏｎ和ＭｃＣｌｅｌｌａｎｄ的实验范式以合体汉字为实验材料对层次模型进行了实验检验。结果表明，以假字掩蔽模式代替特征掩蔽模式使字目标对部件目标的字优效应明显下降。这一结果支持了层次模型，即合体汉字的字形加工以部件识别加工为中介。但本实验还发现了真字掩蔽模式下的字优效应量比假字掩蔽模式下明显减少，这一结果表明，在字加工水平中存在着觉察器间的相互竞争。因此，合体汉字字形识别的部件中介模型有待进一步完善。

EXPERIMENTAL TESTS OF THE HIERARCHICAL MODEL OF TWO -RADICAL CHINESE CHARACTER PERCEPTION

The hierarchical model of English word perception, presented by Johnson and McClelland ( 1980), suggested that identification of an English word is mediated by identification of its component letters. The model explains why people are more accurate in perceiving a briefly presented letter when it appears in a word than it appears alone (the word-letter phenomenon). The results of their experiment supported the model. Comparing a Chinese character composed with more than two radicals to an English word, we can see the structure of a Chinese character in as similar way to that of an English word. Is identification of a Chinese character mediated by identification of its component radicals? Suppose that the hierarchical model of English word perception could explain the procession of Chinese character perception, we could make two predictions which are compatible with the ones from Johnson and McClelland' s model: (1) The sizable character-radical phenomenon that can beobtained using a mask made up of radical features should be greatly reduced if the mask consists of complete radicals. (2) The size of the character-radical phenomenon should be the same whether a mask is a real Chinese character or a pseudo-character. The experiment reported in this paper confirmed prediction 1 , but prediction 2 was inconsistent with the result of our experiment. Nevertheless, the hierarchical model of Chinese character perception should be modified.