Representation and Access of Chinese Compound Words.

摘要

This dissertation reports on three studies of the cognitive representations and processes in visual recognition of Chinese two-character compound words. Because Chinese two-character compound words are composed of two spatially separated characters which themselves are words, there has been a debate about whether these words are represented and processed in the mental lexicon as unitary wholes or as combinations of the component characters (Zhang & Peng, 1992; Taft & Zhu, 1995).;The first two studies are concerned with whether Chinese two-character compound words have decomposed or holistic orthographic representations in the mental lexicon. Study 1 made use of the high frequency orthographic neighbor inhibition effect (Grainger & Jacobs, 1996; Davis & Lupker, 2006) in the masked priming paradigm. It was found that a high frequency orthographic neighbor (e.g., [special characters omitted] means virtually in English) inhibited recognition of a low frequency target word (e.g., [special characters omitted], means brief in English). The high frequency orthographic neighbor inhibition effect was supposed to be caused by lexical competition between prime and target words (Grainger & Jacobs, 1996). Therefore, the observation of inhibition in Study 1 suggests that the prime words have word-level representations that compete with those of target words for lexical access. It is argued that the activated word-level lexical representation is likely to be orthographic representation, rather than phonological and semantic representations, because the influence of phonology was found to be limited in a follow-up experiment, and the semantic relatedness between the prime and target words was low. Furthermore, because the inhibition effects were stronger for semantically opaque than transparent compound words, it seems that opaque words are more likely than transparent words to be represented as unitary units.;Study 2 adapted the transposed-letter similarity effect in English (Forster, Davis, Schoknecht, & Carter, 1987; Acah & Perea, 2008) into transposed-character similarity effect in Chinese to further examine the orthographic representation of Chinese compound words. Study 2 examined the processing of both transposable and untransposable compound words to distinguish the decomposed account (which assumes morpheme-to-word activation) and the holistic account (which assumes activation of word-level orthographic representation) of the mental representation of Chinese compound words. It was found that transposable word-word pairs (e.g., [special characters omitted], means tie in English-[special characters omitted], means lead in English) did not produce significant priming effects. This result is inconsistent with the decomposed account, but is explicable in terms of the holistic account. Two follow-up experiments show that the transposed-character similarity effects differ for transparent (e.g., [special characters omitted], means proud in English) and opaque words (e.g., [special characters omitted], means careless in English). Transposed nonwords show significant facilitation effect to the original words when the original words were opaque (e.g., [special characters omitted]) but not when they were transparent words (e.g., [special characters omitted]). It suggests that opaque words are more likely than transparent words to be represented as orthographic wholes. The findings of word-level orthographic representation and the influence of semantic transparency provide convergent evidence for the conclusions reached by Study 1.;The third study investigated whether or not the meaning of a component morpheme would be activated in the process of recognizing a Chinese two-character compound word. Current accounts of morphological processing disagree on whether morphological processing is form then meaning, or form with meaning (Feldman, O'Connor, & Del Prado Martin, 2009; Davis & Rastle, 2010). The focus of the debate is actually whether morpheme meaning activation occurs at early stage of compound word processing. This study attempted to contribute to that debate by trying to dissociate morphological processing of form and meaning. The method was to investigate masked priming effects produced by prime-target pairs that contained a pair of semantically related morphemes but without any overlap in orthography, phonology or whole-word meaning (e.g., [special characters omitted] /surprised in English/-[special characters omitted]/break one's promise in English/). Since there was no overlap between the prime and the target in form, any priming effects produced could be attributed unequivocally to activation of morpheme meaning (e.g., [special characters omitted], both means eat in English). The results of Study 3 show that morpheme meaning activation without form overlap occurred as early as the first 60ms of word processing. This finding is more consistent with the form-with-meaning than the form-then-meaning account.;In sum, the three studies suggest that, on one hand, Chinese compound words, particularly the opaque ones, seem to be represented as holistic orthographic units in the mental lexicon. On the other hand, the meanings of the component morphemes are activated in visual recognition of Chinese compound words, suggesting a decomposed access. The implications of these results for a model of Chinese compound word representation and access, as well as the dispute between morpho-orthographic and morpho-semantic decomposition accounts of morphological processing in other languages, were discussed.