中文字形結構的偵測：以視覺搜尋斜率值作為結構間相似性的指標

Structure Detection of Chinese Characters: Visual Search Slope as an Index of Similarity between different-Structured Characters

本研究以三個視覺搜尋實驗來檢驗中文字形結構對搜尋效率的影響。在實驗一中受試者的作業是在一群相同的中文字（干擾字）中儘快判斷是否有一個不同的字（目標字）。每次嘗試所呈現的刺激總數不定，隨機呈現在5×5的矩陣方格內，而由反應時間與刺激總數的函數圖形可推知內在的處理速率。干擾字與目標字可能共享相同的結構（同構組），也可能為不同的結構（異構組）。所檢驗的字形結構為葉素玲等人（1997；1999）以字形分類作業經多向度量尺法（MDS）所得的兩種結構向度（水平／垂直、開放／包圍）。實驗二改將刺激以固定密度的方式呈現，受試者的作業是搜尋預先指定的目標字，同構組與異構組的嘗試次數相等，並將目標字與干擾字的字頻控制在更接近的範圍。即使這些改變，兩個實驗的結果皆顯示異構組比同構組有較快的搜尋速率。實驗三將葉素玲等人（1999）經群聚分析所得的五種結構類別兩兩配對呈現，結果顯示當目標字與干擾字二者的結構愈相似，所得的搜尋斜率值愈大，因此可藉由搜尋斜率值作為字形結構之間相似性的指標。文中並討論本研究結果的意涵及其應用。

Three visual search experiments were conducted to examine the effect of the configurational structures of Chinese characters on search efficiency. In Experiment I, the subjects' task was to detect a different character among a homogeneous set of distracting characters presented in an imagery 5 × 5 grid. The set size varied from trial to trial, and the reaction times were measured. The target and the dis tractors either shared the same structure or had different structures. The horizontal-vertical and open-enclosed structural dimensions that were revealed in the character space of Yeh, Li, and Chen (1997; 1999) were examined. In Experiment 2, the subject searched for a predesignated target character among a distracting field of constant density. Better controls of the familiarity of same-structured and different-structured trials and the range of character frequency were made to exclude the possible confounding variables in Experiment I. The results from both experiments showed that it was more efficient to detect a target character when its structure was different from that of the distractors than when their structures were the same. In Experiment 3 the five structure types explicitly specified in Yeh et al. (1999) were paired as the target and distractors. As the structures between these pairs were more similar, the search slopes were steeper. The search slopes thus were served as a diagnostic tool for the similarity of characters with various structures. Implications of these results were discussed in the text.