|題名: ||The cognitive effect of english garden-path sentences on the comprehension of efl learners in Taiwan|
|作者: ||楊惠筠;Yang, Susan Hui-yun|
范瑞玲;Fahn, Rueih-lirng;林永煬;Lin, Yung-yang
|關鍵詞: ||歧義句;限定關係子句結構;句法分析策略;腦磁圖儀;M300成分;garden-path sentence;restrictive relative clause constructions;parsing strategy;Magnetoencephalography (MEG);M300 Component|
|上傳時間: ||2010-01-10 23:12:11 (UTC+8)|
|摘要: ||本論文探討英文能力中級生與進階生大腦對於主詞與受詞省略限定關係歧義句的延遲處理現象與認知歷程的差異性，採用句法分析策略論，使用非侵入性腦磁圖儀（Magnetoencephalography, MEG）來了解歧義現象所產生的認知歷程。操弄主詞與受詞省略限定關係結構句最後一個動詞出現的時間，研究歧義句所產生的誤導效應。使用訊號源分析軟體的Dipole Modelling來研究三個時區（300-500毫秒、500-700毫秒、700-900毫秒）的偶極密度時，發現左右大腦顳葉都有認知反應。當兩組受試者閱讀第一部分主詞省略的關係句時，T值檢定發現右顳葉在潛時300-500毫秒處理歧義句有顯著（p=0.0246）。由配對T值檢定也發現進階生在於時區700-900毫秒左右顳葉處理歧義句也有顯著差（p=0.026）。兩組受試者閱讀第二部分受詞省略的關係歧義句時，發現在左顳葉700-900毫秒的T值檢定有顯著性（p=0.0189）。進階生在閱讀第一與第二部分關係歧義句發現T值檢定於時區300-500毫秒時，右顳葉有顯著差異（p=0.008），顯示受詞省略的關係歧義句比主詞省略的關係歧義句難解。另外使用Sigma Plot分析軟體來解析兩組受試者於時區150- 300毫秒的振幅平均值，以期瞭解反應的差異性。配對T值檢定發現中級生在閱讀第一部分主詞省略關係結構句時，其右顳葉於潛時區間150-300毫秒偵測歧義與非歧義句的差異有顯著性（p<0.05）。第二部分的實驗結果也發現進階生閱讀受詞省略的關係歧義句時有顯著的M300成分，顯示出句法結構的再分析技巧，這證明了延遲現象。由振幅平均值發現進階生對於非歧義句的瞭解大於中級生，而且兩組受試者的右顳葉於時區150-300毫秒與300-500毫秒發現差異有顯著性（p<0.05）。結果證明了進階生閱讀關係歧義句的誤導效應與句法結構的再分析歷程。綜合第一部分與第二部分實驗的結果，左右顳葉對於複雜的關係子句處理都有認知反應，但是對於歧義句的偵測則是在右顳葉。結果顯示，相較於中級生，進階生較持續性地運用句法分析策略來分析句子，閱讀歧義句時的容忍度較強，閱讀非歧義句的理解肯定度也較高。|
This thesis on the cognitive effect of garden-path (GP) sentences investigates the Principle-based Approach used by intermediate and advanced EFL learners reading sentences embedded with subject-reduced and object-reduced relative constructions. Magnetoencephalography (MEG) was used to examine the delayed sentence comprehension processing caused by garden-path attachment ambiguity. In both types of relative constructions, each final critical word was presented on the screen at the same time to subjects. Based on dipole density analysis, both the left and the right temporal areas were involved in the responses to the experimental sentences. For the purposes of analysis, the dipole numbers in the latency shift from 300 to 900 ms were divided into three time windows. For the experiment of Session I, the t-test revealed a significant difference (p= 0.0246) between the responses of the two proficiency groups to garden-path sentences at 300-500 ms in the right hemisphere. However, a paired t-test revealed a significant difference (p=0.026) in the activation of the left and the right temporal areas in the advanced group at 700-900 ms. For the experiment of Session II, the t-test showed a significant difference (p=0.0189) at 700-900 ms in the left temporal area between the two groups’ responses to the GP experimental sentences. Finally, in comparing the results of Session I and Session II, the t-test showed that there was a significant difference (p=0.008) at 300-500 ms in the right temporal area in the advanced group’s responses to the two different types of GP sentences in the sessions. This indicates that the degree of difficulty of GP sentences in Session II was greater than that of Session I. Sigma Plot analysis was used to investigate the mean amplitudes in the time window from 150 to 300 ms. In the detection of the differences between GP and Non-GP sentences, two significant differences were found in the two sessions. The paired t-test results of Session I demonstrated that there was a significant difference (p< 0.05) in the intermediate group’s responses to GP and to Non-GP sentences in the time window from 150 to 300 ms. As for the inter-group variability, it was found that in Session II there was a significant difference (p< 0.05) in the right temporal area in both time windows between the two groups’ responses to sentences with the GP effect. A significant difference in the M300 component was found in the responses of different EFL proficiency groups towards GP sentences. The MEG results prove that the ambiguity in GP sentences causes a reanalysis over time. In the results of these two sessions, it was seen that, although both the left and right temporal areas were activated to different degrees in response to restrictive relative constructions, the subjects’ detection of the difference between GP and Non-GP sentences was mainly in the right temporal area. It is thus concluded that higher proficiency indicates a higher degree of consistency in using a serial parsing strategy, greater tolerance to work with problems like the GP effect, and a greater degree of certainty as shown by the mean amplitudes.