預測效果對第二語言情緒詞處理的影響─ERP研究

Abstract

本文旨在研究第二語言情緒預測語境下的情緒字詞處理，並探討二語人士是否能運用情緒預測語境預期後續字詞的情緒性。為了瞭解在微弱情緒預測語境及強烈情緒預測語境下的情緒字詞處理，我們操弄句子情緒預測程度(Valence Constraint)的強弱，其計算方式為──把每一個字詞作為某一特定句子結尾的克漏字機率(cloze probability)乘以每一字詞的正負向性，加總後得到此一句子在情緒預測上的數值。我們採用強烈或微弱情緒預測的語境，以情緒詞或中性詞作為結尾，產生通順或不通順的句子。我們讓22位母語非漢字圈的中文程度高級學習者以單字詞呈現在電腦螢幕上的方式閱讀這些句子，同時記錄大腦對句子結尾字詞的事件相關電位反應。 與先前的母語人士研究(Chou, 2016)相比，針對結尾詞的通順度效果，母語人士同時產生N400與LPC，而二語人士只引發N400的腦波成分，顯示二語人士在進行不通順結尾詞的語句處理時，較缺乏將其與前述語境進行重新分析、詮釋的歷程，但仍能展現其對語意錯誤的敏感度。針對通順的句子，在微弱情緒預測語境下的情緒字詞處理，無論在母語人士或二語人士，都引發P2及LPC成分。此結果與過去句子中的情緒詞研究相符，顯示情緒詞，與中性詞相比，能引發較多的注意力並使人對眼前情勢進行更深入的評估。 在強烈情緒預測語境之下，雖然母語人士及二語人士都分別對情緒詞和中性詞引發強度相似的P2及LPC成分，但配對比較(pairwise comparison)在兩組人之間的模式不盡相同。我們比較中性詞在微弱及強烈情緒預測語境中的P2峰值，發現只有在母語人士中產生顯著差異──在強烈情緒預測語境下的中性詞引發較大的P2成分；此外，我們比較中性詞在微弱及強烈情緒預測語境中的LPC峰值，兩組人士皆產生顯著差異──在強烈情緒預測語境下的中性詞引發較大的LPC成分。此結果顯示，在強烈情緒預測語境下，母語人士運用主動閱讀策略、引發對後續字詞的預測進行語言處理。二語人士可能因為語言處理的速度較慢、無法引發類似的效果，但仍能透過情緒語境所引發較強的動機及較大的腦內活動，進一步對之後的情緒事件做較好的準備。 而針對本篇研究的主要重點—第二語言情緒預測的效果，我們發現與情緒預測相符的狀況相比，與情緒預測不符的情況在頭皮前區引發一個達邊緣顯著(P=0.05)較大的晚期正波。在先前母語人士的研究中亦引發此一情緒預測的前區正波(frontal Positivity)效果，顯示大腦對非預期但通順的情緒資訊所進行的額外加工處理歷程。總體而言，本篇研究擴展了我們對第二語言情緒處理的認識：高級程度的第二語言者，如同母語人士，能夠運用不同的神經心理處理策略、藉由語言達成有效的情緒預測結果。

This study investigates emotional responses to L2 input with contextual support and asks whether L2 speakers use context to anticipate the emotional content of the upcoming word. Event-Related Potentials were recorded from 22 advanced learners of Taiwan Mandarin (mean age: 28.04 years old, range: 20-45 years) while they read sentences ended with plausible or implausible emotional or neutral words. Additionally, we manipulated the amount of Valence Constraint (VC) for the sentence (strong vs. weak), by summing over valence values of possible sentence endings weighted by cloze probability. Participants read each sentence word by word from a computer screen and judged its plausibility. Comparing with prior results of native (L1) speakers reading the same sentences, we found that while L1 speakers showed both N400 and LPC plausibility effects, L2 learners only elicited N400 plausibility effect. Focusing on plausible sentences only, effect of Word type showed qualitatively similar P2 and LPC responses in L1 and in L2 under weak valence constraint sentences. Consistent with previous findings, emotion words elicited enhance attentional resources and a continuation of deeper evaluation than neutral words. As for strong valence constraint sentences, equal responses to neutral and emotion endings were discovered in both group respectively; however, pairwise comparisons demonstrated distinct pattern within each Valence constraint and Word type combination between L1 and L2. Critically, pairwise comparison to P2 amplitude between Neutral word under weak valence constraint and under strong valence constraint sentence showed significant difference in L1 only, with larger P2 to neutral word under strong valence constraint sentence. Besides, pairwise comparison to LPC amplitude between Neutral word under weak valence constraint and under strong valence constraint sentence showed significant difference in both group. The result suggested that with rich emotional context native speakers tend to apply active reading strategy to process linguistic information in a predictive way. Although such effect was not observed in learners of Mandarin, emotional contexts provide enhanced motivation and, in turn, increase individual’s interior tension as a preparation for the subsequent emotional events. To the main interest of this study, we found a marginal significance (P=0.05) to emotionally-unexpected condition than emotionally-expected condition over anterior scalp region within later temporal window. The same effect, Frontal Positivity, has been observed in previous research with native speakers, indicating an additional cost for processing unexpected yet plausible item on emotional level. Overall, our result added to second language studies by suggesting that L2 speakers, as well as L1 speakers, could generate emotional expectation during incremental language presentation in a less predictive yet functional way.