語言理解的知覺特徵視覺化歷程：中英與英中雙語者的雙向比較

Visualization Process of Perceptual Features in Language Comprehension: A Bidirectional Comparison of Chinese-English and English-Chinese Bilinguals

詞彙理解的心智模擬理論認為，人類心智理解詞彙時會啟動視覺化歷程以模擬知覺經驗。心智模擬歷程應能整合多種知覺特徵，然而迄今只有支持單一特徵視覺化的實徵研究，多重特徵視覺化的課題尚待開發。Bocanegra等人（2022）發現雙特徵優勢僅在序列呈現條件出現，支持序列式特徵交集處理。本研究的前導實驗在即時呈現條件亦偵測到雙特徵優勢，顯示中文詞彙視覺化可能較快完成特徵整合；由於中文詞彙充斥多重特徵組合的複合詞，本研究預期中英與英中雙語者的視覺化歷程模式可能有所差異。本研究以呈現條件（即時／序列）為主要獨變項，招募98名英中雙語者與140名中英雙語者，分別執行中文與英文視覺化任務，比較兩組參與者在不同呈現條件的雙特徵或單特微優勢。四項預註冊假設皆未達預設顯著標準；敏感度分析顯示，增加英文視覺化作業的刺激數量，可達到偵測預期效果量的建議檢定力。納入語言熵的探索性分析發現，英中雙語者作業表現與語言熵呈現相關趨勢，但需要進一步驗證。輔助性語言流暢度測驗顯示英中雙語者的第一語言與第二語言流暢度差異不顯著，而且接近中文母語者的英文流暢度。三種呈現條件均未出現雙特徵優勢，揭示心智模擬理論預測及解釋多重特徵視覺化的侷限。語言差異的調節作用在本研究雖然尚未獲得確認，建議未來研究組合有效的測量指標，以語言熵指數估計調節作用。

The mental simulation theory of lexical comprehension holds that word understanding activates perceptual simulation. This process should integrate multiple perceptual features. However, empirical support has been limited to single-feature visualization. Multi-feature visualization remains largely unexplored. Bocanegra et al. (2022) found a dual-feature advantage only under sequential presentation. This supports sequential feature-concatenation processing. Our pilot study detected a dual-feature advantage under simultaneous presentation. Chinese lexical visualization may therefore achieve feature integration more rapidly. Compound words predominate in the Chinese lexicon and encode multiple perceptual features. We hypothesized that Chinese-English (CE) and English-Chinese (EC) bilinguals differ in their visualization processes. Presentation condition (simultaneous/sequential) was the primary independent variable. Ninety-eight EC and 140 CE bilinguals completed Chinese and English visualization tasks. Dual-feature and single-feature advantages were compared across three conditions. None of the four pre-registered hypotheses reached the predetermined significance criterion. Sensitivity analyses showed that increasing stimuli in the English visualization task would yield the recommended statistical power for the expected effect size. An exploratory analysis used language entropy as a continuous measure. EC bilinguals’task performance showed a trend toward correlation with language entropy scores. This finding requires further replication. A supplementary proficiency measure showed no significant L1-L2 difference for EC bilinguals. Their scores approached CE bilinguals’English proficiency. Dual-feature advantages were absent across all three presentation conditions. This reveals the limitations of mental simulation theory in predicting multi-feature visualization. The moderating role of language background remains unconfirmed. Future research should combine valid measurement indices. Language entropy is recommended as an effective index to estimate the moderation.