記憶材料與呈現模式調節工作記憶負荷相依的神經活動

Abstract

工作記憶讓個體得以短暫地維持多個項目資訊，以引導後續的目標導向行為。工作記憶的其中一項主要特性為容量限制，雖然過去研究經常探討工作記憶容量個體差異的神經基礎，對於個體而言，工作記憶容量可能會受到記憶項目的特性所影響。本論文使用腦電圖和腦磁圖，探討在延遲反應工作記憶作業中，不同的記憶材料（英文字母、規則圖形和抽象圖形）和呈現模式（同時呈現或順序性呈現）如何影響高或低工作記憶負荷的工作記憶表現和神經振盪活動。研究一的行為結果顯示，相較於簡單物體（字母和規則圖形），複雜物體（抽象圖形）的工作記憶容量在較低的工作記憶負荷就達到容量限制，導致簡單物體較複雜物體呈現隨負荷差異較大的工作記憶容量差異。腦電圖的結果和行為測量的結果一致，顯示在工作記憶維持階段，簡單物體較複雜物體呈現較大的負荷相依（高負荷相對於低負荷）的後側腦區alpha power下降。這些結果顯示 alpha 振盪以內容特定的方式追蹤工作記憶容量，不只反映了儲存項目的數目，也反映了記憶材料的特性。研究二中，腦電圖的結果顯示工作記憶維持階段負荷相依的 alpha振盪活動會受到刺激呈現模式的影響，同時呈現模式相較於順序呈現模式較大的後側腦區alpha power下降。此外，順序性呈現在記憶提取階段的N3rs振幅反映了行為的新近效應。多變量模式分析顯示，在中央順序呈現模式中，目標物的序列順序資訊可從記憶搜尋的神經活動中被解碼，且在解譯分析中的嘗試次分類保真度可以預測嘗試次的反應時間。這些結果顯示蘊含在記憶登錄時刺激呈現模式中的空間和順序資訊得以繼續影響後續記憶維持和提取的歷程。在研究三中，行為和腦磁圖的資料複製研究一的結果。感測器層次的結果追蹤了行為測量，在工作記憶維持階段，簡單刺激（英文字母）相較於複雜刺激（抽象圖形）有較大的負荷相依的 alpha power下降和gamma power增加。此外，神經來源定位支持先前的研究結果，顯示枕葉區和後側頂葉區域可能是內容特定工作記憶容量的相關腦區。最後，左側海馬迴theta-gamma的相位-振幅聯結強度可能反映了內容特定的工作記憶容量。這些結果提供了記憶材料會調節不同頻帶的神經振盪活動間協調的神經證據。總結而言，此論文闡明了刺激材料和呈現模式如何調節工作記憶處理時的神經活動，以及不同頻帶的振盪活動在工作記憶維持階段如何互動和協調。

Working memory (WM) enables individuals to maintain information of multiple items for a brief period, guiding goal-directed behaviours. One key feature of WM is its capacity limit. Although the neural basis of WM capacity is often studied by examining individual differences, capacity may also vary depending on the properties of memoranda for a given individual. This dissertation investigates how different memory materials (English letters, regular shapes or abstract shapes) and presentation modes (simultaneous or sequential) influence WM performance and oscillatory activities using electroencephalography (EEG) and magnetoencephalography (MEG) in the delayed response tasks with high and low WM load. In Study I, the behavioural results showed that WM capacity for complex objects (i.e., abstract shapes) reached a plateau at a lower set size compared to simple objects (i.e., English letters and regular shapes), resulting in a larger difference for capacity measures for simple objects relative to complex objects. Our EEG data mirrored the behavioural measures in capacity, showing a larger load-dependent (high versus low WM load) decrease in posterior alpha (8-13 Hz) power for simple objects than complex objects during WM maintenance. These findings suggest that alpha oscillations track WM capacity in a content-specific manner, reflecting not only the number of stored items but also the property of memory material. In Study II, the EEG results demonstrated that the load-dependent alpha oscillatory activity was sensitive to the stimulus presentation mode, as manifested by a larger attenuation of posterior alpha power during WM maintenance when stimuli were presented simultaneously during encoding compared to when presented sequentially. Additionally, the behavioural recency effect was reflected in N3rs amplitudes during WM retrieval for the sequential presentations. Multivariate pattern analysis revealed that the information of serial position of the target during WM encoding can be decoded from neural activity during WM search, and the trial-wise classification fidelity of target position in the decoding analysis predicts the trial-wise RTs in the center-sequential presentation. These findings suggest that the spatial and the ordinal information embedded in the stimulus presentation during encoding continues to affect processing during subsequent WM maintenance and retrieval. In Study III, our behavioural and MEG results at the sensor level replicated the findings from Study I. Consistent with capacity measures, we found a more pronounced load-dependent (high versus low WM load) decrease in alpha (8-13 Hz) power and an increase in gamma (30-100 Hz) power for simple objects (i.e., letters) than complex objects (i.e., abstract shapes) during WM maintenance. Furthermore, the source localisation supported the prior findings that the occipital and the posterior parietal regions may be the locus related to the content-specific WM capacity. Finally, our source-level data demonstrated that the coupling strength of theta-gamma PAC in the left hippocampus may reflect content-specific WM capacity. The findings provided neural evidence that memory material modulates the coordination between oscillatory activities of various frequency bands. Together, this dissertation sheds light on how memory materials and presentation modes modulate neural activities during WM processing, and how oscillatory activities at different frequency bands interact and orchestrate during WM maintenance.