音形一致性對中文字辨識之影響

The Feedback Consistency Effect in Chinese Character Recognition: Evidence from a Psycholinguistic Norm

過去研究已指出語音表徵對閱讀習得與文字辨識的重要性。比如，形音一致性效果顯示，讀音一致的字母串（比如 -ean 在 lean, dean 與 bean 的讀音皆一樣）引發較快且正確的唸名反應。另外有研究指出，拼字形態較多樣的韻母（比如包含韻母 /ip/ 的詞可拼為 heap 或 deep）所構成的詞在詞彙辨識作業中需要較長的辨識時間，此效果稱為音形一致性。這些現象顯示文字訊息激發相關的語音、字形單位，不一致的對應降低交互激發之穩定性，並影響詞彙辨識的效率。然而後續研究未重複驗證音形一致性效果，其可能原因是音形一致性與其他變項共變，比如頻率、字形鄰項個數等等。詞彙變項之間的共變特性導致因素設計所需要之實驗材料受到限制，即研究者無法僅操弄音形一致性但控制所有的詞彙特性。本研究以大量詞彙庫與行為實驗來檢驗中文詞彙辨識中的音形一致性效應。180 位大學生參與詞彙辨識作業，實驗材料包含 3,423 個形聲字。反應時間以線性混合模型進行分析，除了音形一致性，另納入其他詞彙特徵為共變項，包含主／客觀頻率、筆畫數、部件結合度、語意歧義性、形音一致性……等。分析結果呈現典型的詞彙辨識效果，比如，高頻、以及高熟悉度的字引發較快的辨識時間。此外，線性混合模型分析也顯示，估計了各種詞彙變項之影響的同時，音形一致性具有顯著影響力，越一致的字引發越快的辨識速度。此現象顯示語音表徵會影響字形表徵的聯結方式，而且在文字辨識歷程中，兩種表徵會交互激發。

Researchers have demonstrated the importance of phonology in literacy acquisition and in visual word recognition. For example, the spelling-to-sound consistency effect has been observed in visual word recognition tasks, in which the naming responses are faster and more accurate for words with the same letters that also have the same pronunciation (e.g. -ean is always pronounced /in/, as in lean, dean, and bean). In addition, some studies have reported a much less intuitive feedback consistency effect when a rime can be spelled in different ways (e.g. /ip/ in heap and deep) in lexical decision tasks. Such findings suggest that, with activation flowing back and forth between orthographic and phonological units during word processing, any inconsistency in the mappings between orthography and phonology should weaken the stability of the feedback loop, and, thus, should delay recognition. However, several studies have failed to show reliable feedback consistency in printed word recognition. One possible reason for this is that the feedback consistency is naturally confounded with many other variables, such as orthographic neighborhood or bigram frequency, as these variables are difficult to tease apart. Furthermore, there are challenges in designing factorial experiments that perfectly balance lexical stimuli on all factors besides feedback consistency. This study aims to examine the feedback consistency effect in reading Chinese characters by using a normative data of 3,423 Chinese phonograms. We collected the lexical decision time from 180 college students. A linear mixed model analysis was used to examine the feedback consistency effect by taking into account additional properties that may be confounded with feedback consistency, including character frequency, number of strokes, phonetic combinability, semantic combinability, semantic ambiguity, phonetic consistency, noun-to-verb ratios, and morphological boundedness. Some typical effects were observed, such as the more frequent and familiar a character, the faster one can decide it is a real character. More importantly, the linear mixed model analysis revealed a significant feedback consistency effect while controlling for other factors, which indicated that the pronunciation of phonograms might accommodate the organization of Chinese orthographic representation. Our study disentangled the feedback consistency from the many other factors, and supports the view that phonological activation would reverberate to orthographic representation in visual word recognition.