中文字形分類系統的再確立：類別與字數的控制

Classification of The Shapes of Chinese Characters: Verification by Different Predesignated Categories and Varied Sample Sizes

為確認葉素玲、李金鈴、陳一平（1997）就中文單字形狀的相似性判斷分組實驗所建立之中文字形分類系統，本研究針對該文實驗的兩個可能缺失予以改進。一是字形取樣的類別，本研究參考Liu（1984）的11類再加一類共選定12類字形樣本；三是在各類別內選取字數不等的字形，以去除受試者藉由各類別相同且固定的字數而猜出研究者事先選定的類別的可能性。受試者的作業是將中文字依其形狀的相似性作分組，所得再進行群聚分析以及多向度量尺法（MDS）的分析。所得結果與先前的研究結果頗為一致，左右、厂型、L型、上下，包型仍是五個主要的群聚，而由MDS所得的三軸可界定為上下／左右、開放／包圍與方／不方。

To verify the classification system of the shapes of Chinese characters identified by Yeh, Li. and Chen (1997) this study was designed to avoid the two possible shortcomings in the previous study. Twelve predesignated categories of characters adopted from Liu (1984) were selected to cover a larger spectrum of Chinese characters. In addition. the set size in each category was varied to prevent subjects' guessing as might have been caused by the regularity of the constant set size. As with the previous study. the subjects were asked to group Chinese characters according to the degree of similarities in their shapes. Hierachical cluster analysis and nonmetric multidimensional scaling were conducted and the results were consistent with the original report by Yeh et al.. Horizontal, L-shaped. 厂-shaped. vertical and enclosed characters formed 5 major clusters. Horizontal-vertical, open-enclosed, and square-nonsquare were identified as the 3 axes in the character space.