跨語言原型效應之電生理證據

Abstract

雙語者──尤其是外語能力較低者──有時會出現直接翻譯或第二語言類別化的錯誤，這可能源自於雙語者傾向選擇「原型翻譯對等（typical translation equivalent）」的緣故，然而，我們對此一現象的神經機制仍然所知甚少。為了瞭解此機制，本論文進行了兩個ERP實驗。第一個實驗探討了跨語言原型效應對中英雙語者進行中英動詞翻譯的影響。在實驗中受試者先閱讀一個中文動詞短語（如「開窗戶」或「開電腦」），接著閲讀一個英語動詞（例如， 「open」（原型翻譯對等）或 「turn on」（非原型翻譯對等）），並判斷該英語動詞是否為先前中文動詞（例如，「開」）適當的翻譯。研究結果發現，在適當翻譯對等的情況下，非原型翻譯比原型翻譯引發了更強的 N400反應，顯示雙語者在提取詞彙和語義上發生困難；此外，非原型翻譯也比原型翻譯引發了更強的f-PNP反應，顯示雙語者在整合非原型翻譯對等的過程中得抑制不適切的原型翻譯對等。這個結果可能與原型與非原型翻譯對等之間的靜息狀態激發（resting-level activation）差異有關。第二個實驗探討了跨語言原型效應對中英雙語者之第二語言類別化的影響。事物的中文標籤常蘊含指示類別成員的語義線索（例如，中文標籤「鴕鳥」包含了「鳥」這個類別成員線索）。先前的 ERP 研究發現，即使測試時啟動的類別為英文（例如，用「bird」 來代替 「鳥」），中文母語者仍能在對非原型成員進行類別化時從這些語義線索得到正回饋，進而導致其引發的N400 原型性效應減弱甚至消失。第二個實驗探討了雙語者的英文類別化是否會受到中文標籤所蘊含負回饋線索的影響。受試者先看到一個英文類別標籤（例如「bird」 或 「car」），接著看到一個圖像（例如，「鴕鳥」或「大卡車」），然後需要判斷該圖像是否屬於先前所呈現的類別。結果顯示，具中文負回饋線索的圖像（例如，受試者先看「car」，再看大卡車的圖案）比具正回饋線索的圖像（例如，受試者先看「bird」，再看鴕鳥的圖案）引發了更強的 N400反應，顯示雙語者需要付出額外的心力才能忽略中文類別表徵的干擾，並正確選取英文類別表徵。兩個實驗的結果凸顯出原型翻譯對等在跨語言處理中的雙重作用：即使在熟練中英雙語者的大腦裡，當原型翻譯對等與原文的語義範圍一致時，它會促進跨語言處理；反之，則產生干擾。Many erroneous literal translations or second-language miscategorization, often produced by low-proficiency bilinguals, can be attributed to a tendency to favor typical translation equivalents; however, the underlying neural mechanism remains poorly understood. This research investigated this typicality effect in real-time translation with the event-related potential (ERP) technique. The first ERP experiment examined the effect of crosslinguistic typicality on Mandarin Chinese-to-English verb translation in Chinese–English bilinguals. Participants were presented with a Chinese verb phrase (e.g., kāi chuānghù [“open the window”] or kāi diànnǎo [“turn on the computer”]) followed by an English verb (e.g., open [a typical translation] or turn on [an atypical translation]) and judged whether the English verb was an appropriate/congruent translation of the verb in the Chinese verb phrase previously presented (e.g., kāi). Compared to typical translation equivalents, atypical equivalents elicited an N400, indicating retrieval difficulty, and an f-PNP in congruent trials, reflecting the effort to inhibit typical translations and integrate atypical ones. This pattern may arise from differences in resting-level activation between typical and atypical equivalents. The second ERP experiment explored the crosslinguistic typicality effect on second-language categorization in Chinese–English bilinguals. Items’ Chinese labels often include morphological cues indicating category membership (e.g., the character niǎo [“bird”] is embedded in tuóniǎo [“ostrich”]). Prior ERP studies found that Chinese speakers use such cues to categorize atypical members, even when the priming category is in English (e.g., bird instead of niǎo), resulting in a reduced or absent N400 typicality effect compared to typical members. This experiment investigated whether Chinese–English bilinguals’ English categorization would be affected by misleading cues embedded in their Chinese labels. The participants were presented with an English category label (e.g., bird or car) followed by an image (ostrich or truck) and had to judge whether the image was a member of the previously presented category. Compared to items with facilitating Chinese cues (e.g., bird followed by an ostrich/tuóniǎo image), items with misleading Chinese cues (e.g., car followed by a truck/dàkǎchē image) elicited a posterior N400, indexing the effort to select the representations associated with English categorization over those associated with Chinese categorization. The results of both experiments underscore the double-edged nature of typical translation equivalents which either facilitate (when congruent) or interfere (when incongruent) with crosslinguistic processes, even in the brain of highly proficient Chinese–English bilinguals.