情緒在年齡相關語句閱讀預測歷程中的角色：來自腦電位的證據

Abstract

本研究探討年輕與年長成人在語言理解過程中，預測性處理動態如何受到情緒的調節。實驗招募27位健康年輕成人（年齡介於20–30歲，平均為24歲）與30位年長臺灣華語母語者（年齡介於60–77歲，平均為69歲），在閱讀具情緒性或中性語句時記錄其事件相關電位（ERP）反應。語句在語意限制強度（constraint strength）上有所操弄，並以預期或不預期的情緒性或中性詞彙為結尾。 結果顯示，無論年齡組別，受試者對於出現在情緒性語境中的預期情緒詞，皆誘發較中性語境中更大的前額P200振幅，而該效應在年長組表現為較不前額的分佈，反映出情緒語境所驅動的早期注意力投入。此外，兩組皆在N400時間窗展現典型的預期效應：預期結尾相較於不預期結尾誘發較小的N400負波，且此效應不受語境情緒性的影響。特別是在中性語境中，來自語意限制性高的語句（Neu_SCE）的預期結尾，誘發較語意限制性低語句（Neu_WCE）為小的N400振幅，該效應在兩組中皆有顯現；然而，在情緒語境下，年輕組則未顯現類似的N400減弱。在前額ERP效應方面，兩組皆展現語境限制性調節的影響，但情緒調節的模式在年齡組別間有所不同。在中性語境中，兩組對於來自高限制性語境的預期結尾（Neu_SCE）皆呈現顯著較大的前額負波，相較於其他中性語句；此外，年長組對於低限制性語境中的不預期結尾（Neu_WCU）亦額外誘發一個前額正波。相對地，在情緒語境中，兩組的反應模式出現差異：年輕組對來自低限制性語境的預期結尾（Emo_WCE）表現出較大的前額負波，且未對不預期結尾產生正波；年長組則在高限制性語境的不預期結尾條件（Emo_SCU）中，顯現顯著的前額正波。這些結果說明，情緒對預測性語言處理的調節存在年齡差異，且表現出不同的電生理反應型態。在未進行情緒操弄的情況下，兩組皆能依據語句中的語意限制性建立語意整合，進而影響隨後詞彙的語境回溯處理。當給予足夠時間使情緒逐漸累積，情緒可能會在整個語境中增強語意訊息的整合與建構，進一步提升年輕人在高、低限制性語境中的再處理歷程。相對地，年長組僅在高限制性情緒語境中（而非中性語境）對不預期中性結尾詞產生前額正波（AP），此結果與先前研究一致，並指出情緒參與可能有助於提升年長者的預測性語言處理能力。整體而言，本研究強調了預測性語言處理在整個生命歷程中的適應性，並突顯了情緒如何在強語境限制下，透過不同的機制調節年輕與年長成人的語言理解歷程。

This study examined how predictive dynamics during language comprehension is modulated by emotion in young and older adults. Event-Related Potentials were recorded from 27 healthy young (aged 20–30 years, M=24) and 30 older Taiwan Mandarin natives (aged 60–77 years, M=69) as they read emotionally or neutrally framed sentences, varying in constraint strength and ending with expected or unexpected emotional or neutral words. Enhanced anterior P200 amplitudes to emotionally expected words in emotional relative to neutral contexts were observed in both young and older adults, with a less frontal distribution in older adults—suggesting early attentional engagement driven by emotional context. The results showed that expected endings elicited reduced N400 amplitudes compared to unexpected endings across context emotionality in both age groups. In particular, expected endings following strongly constraining neutral contexts (Neu_SCE) elicited less negative N400 amplitudes than those following weakly constraining neutral contexts (Neu_WCE) in both age groups, whereas this N400 reduction was not observed under emotional contexts in young adults. In addition, both age groups exhibited anterior constraint effects, but the pattern of emotional modulation under strong constraint differed between young and older adults. In neutral contexts, both age groups exhibited greater anterior negativity for expected endings following strongly constraining contexts (Neu_SCE) compared to the other neutral conditions. Older adults also showed an additional anterior positivity to unexpected endings in weakly constraining contexts (Neu_WCU). In emotional contexts, the pattern diverged: young adults showed enhanced anterior negativity for expected endings in weakly constraining contexts (Emo_WCE) but no anterior positivity for unexpected endings, whereas older adults exhibited a distinct anterior positivity to unexpected endings following strongly constraining contexts (Emo_SCU). These results indicate distinct patterns of emotional modulation in predictive dynamics during sentence comprehension. Without systematic emotional manipulation, both young and older adults were able to integrate semantic information based on strong contextual constraints, which in turn influenced context revisiting when reading upcoming words. With sufficient time for incrementing emotionality, emotion may globally amplify message-level consolidation throughout the context, enhancing revisiting processes not only in strongly but also in weakly constraining sentences for young adults. In contrast, for older adults, unexpected neutral words elicited an Anterior Positivity (AP) when ending an emotionally, strongly constraining contexts—a pattern absent in neutral contexts—replicating previous findings. In light of this, the observed AP effect may suggest that predictive processing during reading becomes more available for older adults when emotionally engaged. These findings underscore the malleability of predictive language processing across the lifespan and highlight age-related differences in how emotion modulates such mechanisms under strong contextual constraint.