本研究操弄語意透明度，探討有臺灣手語經驗聽損者的中文構詞與神經電位運作機制。語意透明度詞分為透明詞（如平坦）與不透明詞（如花生），透明詞是指全詞語意可從組成字（詞素）語意得知，不透明詞則較難從詞素的語意獲得全詞語意。本研究假設：若詞彙以全詞表徵運作，透明詞與不透明詞的行為與腦電波結果應不會有差異；若會進行詞素分解後彙整，由於不透明詞詞素與全詞語意不一致，會需要較多認知資源彙整；若有些詞彙偏向詞素分解，有些以全詞表徵，兩種運作平行地激發，則為雙路徑模式。受試者包含28位聽力正常的聽常者，以及28位流利使用臺灣手語的聽損者。詞彙判斷作業中，受試者需判斷是否為合理的中文詞，之後在未預期下進行記憶作業，判斷詞彙是否出現過（新詞，舊詞），並評估當下記憶判斷的信心（記得/確定、知道/不確定）。結果發現在詞彙作業中，兩族群對不透明詞有較正確、較快反應時間，但腦電波結果相反，聽損者對透明詞有較負向電位，聽常者則是不透明詞。記憶作業中，兩族群對不透明詞有較高正確率與信心評估，但聽常者對不透明詞有較高敏感度，不透明詞的新詞比舊詞有更負向腦波，聽損者這兩項未有差異。這顯示出聽常者對詞彙進行詞素分解，不透明詞需要較多認知資源彙整，但也形成更凸顯的記憶表現；聽損者對透明詞採詞素分解，不透明詞則仰賴全詞表徵，支持雙路徑模式。此結果揭示出流利使用手語的聽損者，會發展出適合中文構詞知識與運作，以達詞彙理解的目的。

Purpose: In this study, we manipulated the semantic transparency of Chinese words to investigate how deaf individuals recognize morphologically complex Chinese words. We applied a lexical decision task and an incidental memory old/new recognition task by using behavioral and event-related-potential (ERP) recordings. The semantic transparency of Chinese words includes transparent and opaque words. Transparent words are words whose meanings can be derived from their constitutive characters (e.g., 平坦, flat), whereas opaque words are words whose meanings can rarely be predicted from any character (e.g., 花生, peanut). From this perspective, the similarity between transparent and opaque words is that they encompass two independent processes of constructing morpheme meanings and whole-word meanings, while the main difference is in the process of integrating conflicting meanings between morphemes and whole words. Opaque words suffer more conflict of meanings than transparent ones. The differences between transparent and opaque words reflect the direct impact of lexical semantics. We hypothesized that if participants processed any two-character Chinese word as a whole-word unit, transparent and opaque words would elicit similar behavioral results or exhibit similar electrophysiological waves. Conversely, if participants unavoidably decomposed a Chinese two-character word into two morphemes, the opaque words would require more effort to integrate the conflicting meanings between morphemes and whole words. For transparent and opaque words, differences would be evident in behavioral or electrophysiological results. Because one-to-one mappings between the Taiwanese Sign Language (TSL) lexicon and Chinese characters are lacking, some scholars have expressed concerns that earlier knowledge of TSL might disturb Chinese word learning. This study sought to provide clear scientific evidence concerning Chinese morphological processing among deaf signers. Methods: This study recruited 28 deaf signers with fluent TSL (16 females; age range: 25 to 55 years; mean age: 35.57 years) and 28 people with hearing and without any sign language experience (18 females; age range: 21 to 51 years; mean age: 31.46 years). Deaf participants were prelingually and profoundly deaf, with a hearing loss of 85 dB or more. These deaf participants acquired TSL as their first or predominant language (age of acquisition: 1 to 18 years; mean: 8.29 years) and completed at least senior high school. On a 0 to 5 scale (0 = low proficiency, 5 = high proficiency), deaf participants rated their proficiency in TSL communication and their reading abilities in written Mandarin on average at 4.04 and 3.71, respectively. The lexical decision task consisted of 120 two-character Chinese words. One hundred twenty two-character pseudo-words were generated by rearranging Chinese characters not used in this experiment. An additional 20 participants (6 males; mean age: 20.6 years) provided ratings of transparency (0 = most opaque, 10 = most transparent). According to the transparency results, 60 transparent words (mean = 6.98) and 60 opaque words (mean = 0.97) were significantly different. The frequencies of transparent words (32.33 per million) and opaque words (32.30 per million) were statistically nonsignificant. Participants decided whether the two-character words were real Chinese words or pseudo-words. After finishing the lexical decision task, participants undertook an incidental memory task in which they had to make an “old or new” judgment and then make a “remember (sure) or know (not sure)” judgment while presenting an instruction of memory certainty. Old words referred to the 52 Chinese words previously shown in the lexical decision task, whereas new words were the 52 Chinese words not shown in the lexical task. Half of the old words were transparent words, and the other half were opaque words. The electrophysiological data were recorded from 32 Ag/AgCl electrodes mounted on an electrode cap using SynAmps amplifiers and arranged according to the extended 10-20 location system. We focused on Fz, FCz, Cz, and Pz electrodes. In the lexical decision task, the electrophysiological signals between 100 and 600 ms after stimulus onset were separated into a series of 50-ms time windows to provide details regarding the time course of activation. In the memory task, the electrophysiological signals of a series of 50-ms time windows were analyzed between 300 ms and 600 ms after stimulus onset. Results: The behavioral results of lexical decision tasks indicated that the accuracies for the transparent and opaque words were 0.86 and 0.93, respectively, among the deaf group and 0.94 and 0.97, respectively, among the hearing group. The correct reaction times for the transparent and opaque words were 656 ms and 632 ms among the deaf group and 644 ms and 622 ms among the group with hearing. Both groups exhibited higher accuracy and faster reaction times for opaque words. However, the electrophysiological results revealed increased negative waves for transparent words between 150 ms and 600 ms on the Cz and Pz among deaf individuals and between 100 ms and 550 ms on the Fz for opaque words among participants with hearing. In the memory task, deaf participants recorded scores on the transparent and opaque words of 0.60 and 0.63 in accuracy, respectively, 1.13 and 1.15 in sensitivity, respectively, and 0.68 and 0.73 in memory confidence, respectively. For participants with hearing, their scores for transparent and opaque words were 0.60 and 0.69 in accuracy, respectively, 0.85 and 1.29 in sensitivity, respectively, and 0.70 and 0.83 in memory confidence, respectively. Both groups had higher accuracy and more confidence in recognizing opaque words. Deaf participants had different old and new word ERP patterns between 300 ms and 350 ms on the FCz and between 450 ms and 600 ms on the FCz and Cz. However, deaf participants exhibited no differences in sensitivity or ERP signals between transparent and opaque words. The group with hearing demonstrated higher sensitivity for opaque words and more negative ERP signals for opaque old words than for opaque new words, with signals ranging between 450 ms and 500 ms on the Pz and between 500 ms and 600 ms on the four electrodes. Conclusion: These findings indicated that individuals with hearing consistently decomposed whole Chinese words into smaller morphemes during lexical access processing. The incongruence between opaque words’ lexical and sublexical meanings became more distinct and easier to remember for people with hearing. The current study provides evidence of a dual-route view for deaf signers who decompose transparent words into smaller morphemes and process opaque words as whole-form units. The findings indicate that fluent deaf signers develop adaptive Chinese morphological knowledge and processes for lexical comprehension.