中文字、詞在文章閱讀理解的背景下的認知處理歷程

Abstract

應用潛藏語意分析（Latent Semantic Analysis, LSA）到中文文章理解時，遭遇到的一個困難是在字的輸入部分，究竟是要用個別的單字或是由單字組成的詞。在本研究中由文章理解的建構-整合理論的觀點假設中文閱讀理解時以單字作為處理單位即可。在研究中以比較中文雙字詞在有足夠的文章背景下即在薄弱的文章背景下，其詞義及其構成單字的字義是否平行激發，以驗證此假設。實驗一以一次一個字的方式呈現中文文章，並檢驗雙字詞之單字字義、詞義激發與文章背景訊息對字彙判斷作業的促發效果，結果發現當SOA為500ms時，與文章背景訊息無關的單字字義以及文章的背景訊息都會被激發，SOA為1000ms時此激發產生的促發效果消失。另一方面，與文章背景無關的雙字詞詞義在SOA為200、500、1000ms時都產生負向的促發效果。此結果顯示中文的雙字詞中的單字字義是被獨立且平行激發的，與建構-整合模型的假設相符，但雙字詞則不符合。實驗二是雙字詞為背景來與比較雙字詞中構成單字的字義與雙字詞的詞義被激發的時間歷程，結果發現雙字詞中個別組合單字在SOA為200ms時，其與雙字詞詞義無關的字義會與雙字詞詞義皆被平行激發，且雙字詞中第二字的字義與雙字詞詞義的激發則皆會持續至SOA為1000ms時。綜合兩個實驗結果，我們推論在中文的理解歷程中，中文的認知處理單位可以是單字。

Researchers met one problem while applying Latent Semantic Analysis (LSA) to Chinese: what should be input, single Chinese character or multi-characters words? In the present study, Kintsch’s text comprehension paradigm (1985, 1988), Construction-Integration Model, was applied to test the hypothesis that Chinese readers process character by character under Chinese text reading comprehension. In Experiment 1, four different prime-target relationships (single character, two-character words, context, and control) and three different stimulus onset asynchrony (SOA) (200, 500, & 1000ms) were manipulated between the prime and the target character. The participants were asked to do lexical decision task under reading comprehension task. Results of Experiment 1 indicated irrelevant meanings of Chinese characters and context meaning were parallel activated at 500ms SOA under reading comprehension task and this effect decayed at 1000ms, but irrelevant meanings of two-character words were inhibited under reading across all three SOA’s. This result indicated Chinese character is more like English word under reading. In Experiment 2, four different prime-target relationships (the first character, the second character, whole word, and control) and three different stimulus onset asynchrony (SOA) (200, 500, & 1000ms) were manipulated between the prime and the target character. The participants make lexical decision task. The results of Experiment 2 indicated all meanings of single character in words were parallel activated while reading two-character words.