| 摘要: | 本研究利用中文雙字詞的語意透明度(semantic transparency)，探討再認記憶中的鏡像效應(mirror effect)是否會因為記憶登錄時對語意透明詞(semantic transparent words)以及語意不透明詞(semantic opaque words)的不同處理方式不同而產生。 實驗一及實驗二分別以受試者內及受試者間的實驗設計操弄記憶材料的語意透明度，並要求受試者進行『記得/知道判斷』(Remember/Know judgment)，藉以量測受試者對這兩類詞記憶的主觀經驗。實驗結果觀察到鏡像效應的產生，相較於語意透明的中文雙字詞，語意不透明的中文雙字詞有較高的正確再認率(hit)以及較低的錯誤再認(false alarm)率。此外，不透明詞相較於透明詞，在正確再認中有較多記得(Remember)反應，錯誤再認中有較少的知道(Know)反應，顯示在提取不透明詞的過程中，雙歷程理論(dual-process theory)中的清晰記憶歷程(recollection)佔了較多比例；提取透明詞時，則以熟悉歷程(familiarity)為主。 實驗三利用事件相關腦電位(ERPs)，觀察透明詞與不透明詞所激發的事件相關電位新舊效果是否有所不同。實驗發現不透明詞相較於透明詞有較大的左側頂葉區新舊效果(left parietal effect)，顯示提取不透明詞時清晰記憶歷程所佔的比重較多；在中額葉區新舊效果(mid-frontal effect)上，兩類詞間則未產生差異。當比較兩類詞中正確接受與正確拒絕(correct rejection)波形差異時，正確拒絕波形的差異在兩類詞之間達到顯著，而正確接受部分則未達顯著，因此，實驗三中的事件相關腦電位新舊效果可能肇因於透明詞與不透明詞兩者之間對做出正確拒絕反應的標準不同。 為將兩類詞之間的決斷標準設置在同一水準上，實驗四要求受試者在新舊判斷之外額外做出信心程度的評量，並取高信心程度的反應作比較。結果發現整體鏡像效果依然存在，高信心程度組中的正確拒絕比例依然是以不透明詞較高，確認此效果並非來自於受試者對正確拒絕部分判斷標準的不同。 實驗五則使用事件相關腦電位觀察受試者的高信心程度反應，結果發現在左側額葉區新舊效果及中額葉區新舊效果均產生差異，不透明詞所產生的效果均較透明詞大。不透明詞相較於透明詞，無論在清晰記憶或是熟悉歷程上均有較高的比例，驗證此處新舊效果並非來自於決斷標準的不同。 以上實驗顯示，操弄語意透明度作為變項時，可觀察到再認記憶中的鏡像效應，根據先前在語言學上對透明詞以及不透明詞的研究結果，不透明詞因全詞意義與部件意義不相關，全詞表徵不易受到部件表徵干擾；而透明詞因部件意義與全詞意義高度相關，全詞表徵可經由相關的部件意義組合而成，因此容易受到非目標的激發干擾。不透明詞不易受部件表徵干擾的特性，可能使其較透明詞具有較大的區別性(distinctiveness)，因而產生較佳的記憶效果。Semantic transparency is an important property of compound word. By definition, the meaning of transparent words can be inferred by their constituents; in contrast, the meaning of opaque words cannot be derived from their constituents. The experiments of this thesis aimed to examine whether and how the mirror effect, which has been reported in previous recognition memory studies that manipulated the frequency and neighborhood size of the materials, would be induced by the semantic transparency of Chinese two-character words. In Experiments One and Two, participants engaged in a study-test recognition memory test, where the semantic transparency was manipuoated as a with-subejcts and a between-subjects factor, respectively. The Remember/Know procedure was employed in both experiments to examine the contributions of recollection and familiarity to the memory perforamcnes for these two types of words. A reliable mirror effect was observed in both experiments. In comparison to transparent words, opaque words elicited a higher hit rate and a lower false alarm rate, In addition, the higher hit rate for opaque words mainly comes from a larger proportion of Remember response, while the higher false alarm rate for transparent words mainly comes from a larger prorion of Know response., These findings are consistent with the hypothesis that the different processing of Transparent and Opaque words will lead to mirror effect. The findings are also consistent with the dual-process account for the mirror effect. In Experiment Three, subjects engaged in the same procedure as Experiment 1, except that the Remember/Know procedure was not employed. Instead, the ERPs were recorded at test, aiming to examine the electrophysiological correlates of the mirror effect einduced by semantic transparency. The behavioral data gave rise to a mirror effect. In the ERP data, the left parietal old/new effect, thought to reflect recollection-based recognition, was larger for opaque words than for transparent words. The data therefore was consistent with Experiment 1 in showing that opaque words elicited more recollective experience than transparent onew. However, the mid frontal old/new effect, thought to reflect familiarity, was however equivalent for both types of words. In addition, it was found that the difference between the parietal effect for these two types of words mainly comes from the waves asoociated with correct rejections. These results suggested that participants might have adopted different criteria while making memory judgments to these types of words. In Experiment Four, a procedure that incorporated Remember/Know response and cofident judgment was employed. Subjects made five response categories, i.e., Rememebr, Sure Old, Not-Sure Old, Not-Sure New, Sure New, to transparent and opaque words. It was found that opaque words received more Remember Hits than transparent words, as in Experiments One and Two. However, for correct rejections to these two types of words, there was a higher proportion of confident rejection (Sure New) for transparent words than for opaque ones. This finding therefore support that the ERP findings in Experiment Three might have come from different criteria aopted by subjects for these two types of words. To examine this possibility, ERPs were recorded in Experiment 5 while subjects made confidence old/new judgments to transparent and opaque words. The ERP analysis only included the response categories of confident hits and confident correct rejections. The left parietal old/new effect was still larger for opaque words than transparent words, suggesting that this pattern of results was not due to the different cirteirial for these two types of words. Neverhteless, the mid frontal effect was found to be larger for opaque words than for transparent words. In addition, the difference in mid frontal effect mainly comes from correct rejections. Across the five experiments, opaque words gave rise to better memory performance and more recollection than transparent words. A plausible interpretation for these results is offered in the thesis as follows. The better memory performance for opaque words may result from the distincivess associated with opaque words during the time of encoding. On the other hand, the transparent words might have elicited a more flent processing than opaque words, such that the resulting familiarity led to the higher false alarm rate for trasnpsrent words than for opaque words. |
