|Abstract: ||鑒於關係子句的複雜性及類型豐富性，其相關研究已經成為一個非常重要的議題。以往研究發現，雖然主語關係子句和賓語關係子句僅在詞序上不同，但存在明顯 的難易差別。因此，兩類關係子句的比較研究為理解大腦如何處理不同難度的句子提 供了重要材料。儘管大量來自印歐語的研究一致表明賓語關係子句比主語關係子句更 難，但在漢語裡，關於二者的加工偏好仍有待釐清。雖然已有研究嘗試借助行為或腦 電波實驗來解決這一重要的語言學議題，但結論不一;加上目前尚未有腦成像方面的 證據支持漢語兩類關係子句存在難易差別。因此，本論文旨在通過一系列的自定閱讀 速度(第二章)以及功能性核磁共振造影實驗(第三和四章)揭示漢語主語關係子句 和賓語關係子句之間的加工偏好模式以及對應的大腦處理機制;進而從單句子的層面 探討大腦是如何通過動態性調整來實現對不同難度句子的加工(第五章)。
同時，本論文首次為漢語關係子句存在加工偏好的處理模式提供了清晰的腦成像 證據。在理解更難的漢語關係子句時，激活了更多與句子處理相關的腦區域面積，特 別表現在左額下迴 (left inferior frontal gyrus)和左顳上迴(left superior temporal gyrus)兩大腦區域。左額下迴和左顳上迴在句子處理歷程中分別發揮了句法結構重組 和詞彙意義整合的作用。同時，從左額下迴到左顳上迴這一特定腦連接對理解漢語關係子句起著重要作用。兩大關鍵腦區域的相互作用確保了從記憶中檢索到的詞彙信息被統一為連貫有意義的序列，使其具有完整的句法結構和語意解釋，從而達到理解漢語關係子句的目的。
綜上所述，本研究不僅證實了漢語兩類關係子句存在明顯的難易差別及存在特定 的加工偏好模式，也揭示了左額下迴和左顳上迴在理解漢語關係子句中的重要作用。 同時，本研究從單句子層面的觀察發現關係子句的不同處理難度可以調節左額下迴和 左顳上迴這兩大腦區域之間的連接強度。從本論文的系列行為和腦成像實驗中可看 出，句法結構且語義內容存在差異的句子，即使差異很小，其難易程度也可反映在腦 活化的模式中。因此，除了傳統對句子難度的測量指標(例如閱讀時間的長短和眼睛 注視時間的長短)，動態腦連接模式的改變可能也是一個區分句子難度的有效指標。;The processing of relative clauses (RCs) has been an important topic owing to its grammatical complexity as well as its rich typology. Extensive research has found that there is a processing asymmetry between the two most common types of RCs, subject-extracted relative clauses (SRCs) and object-extracted relative clauses (ORCs). Comparison of SRCs and ORCs, which differ minimally in word order, provides us valuable materials to investigate how the human brain interprets sentences with different levels of processing difficulty. A great deal of research has indicated that SRCs are consistently easier to comprehend than ORCs in Indo- European languages, which seems to raise a very attractive proposition of a universal preference for SRCs; however, it has been a long-standing debate about whether Chinese SRCs are also easier to process than ORCs. Moreover, the neural correlates underlying the processing of Chinese RCs have been poorly understood before. Therefore, one aim of this dissertation is to investigate the processing preference pattern between Chinese SRCs and ORCs by conducting a series of self-paced reading experiments (Chapter 2) and also explore the neural mechanisms underlying the comprehension of Chinese RCs by running fMRI experiments (Chapter 3 and 4). Another aim is to elaborate on how the human brain dynamically adjusts itself for comprehending Chinese RCs with different levels of processing difficulty by applying a newly-developed single-trial analysis (Chapter 5).
The present findings from a series of self-paced reading experiments (Chapter 2) demonstrated that there was a processing asymmetry between Chinese SRCs and ORCs and Chinese ORCs were easier to comprehend than Chinese SRCs. The ORC preference in Chinese thus poses a challenge to the theoretical approaches that predict a universal SRC advantage and also highlight the importance of investigating language-specific processing mechanisms in sentence comprehension. Besides, these behavioral findings set a reference for interpretations of the neural mechanisms underlying the processing of Chinese SRCs and ORCs.
The neuroimaging studies (Chapter 3 and 4) then presented clear neuroimaging evidence, for the first time in the literature, to support the ORC preference in Chinese, as reflected in enhanced activation of the LIFG and the LSTG during the comprehension of Chinese SRCs than ORCs. In addition to crucial roles of the LIFG and the LSTG, these neuroimaging results also indicate that enhanced effective connectivity from the LIFG to the LSTG may prominently contribute to the comprehension of Chinese SRCs that are more difficult than ORCs.
Further, through a single-trial analysis by combining fMRI and eye-tracking data (Chapter 5), the strength of the connectivity from the LIFG to the LSTG was found to be significantly correlated with the processing difficulty of Chinese SRCs. The findings further confirm our assumption that the effective connectivity from the LIFG to the LSTG plays an important role during the processing of RC structures. Enhanced strength of this connectivity is claimed to reflect increased integration demands and restructuring attempts for sentence comprehension.
In sum, the findings of this dissertation contribute to the understanding of the subject-object asymmetry in Chinese and the neural mechanisms underlying the comprehension of Chinese RCs. Moreover, the additional sources of evidence from a single-trial analysis reveal that brain activation in between the LIFG and the LSTG can dynamically attune to different levels of processing difficulty. Besides, the collected behavioral and neuroimaging evidence from the same set of stimuli in a series of studies demonstrate that grammatically different sentences but with subtle differences in semantic content can be detected in brain activation patterns. Therefore, in addition to traditional measures (i.e., reading time, eye fixation time), the dynamic networking pattern change might be another useful index to differentiate the difficulty level of sentence comprehension, as what we found in the current neuroimaging studies.