過去我們的研究（葉素玲、李金鈴、陳一平，1997；Yeh, Li, & Chen, 1999, Yeh & Li, 2002）顯示，熟練的中文讀者在判斷中文字形的相似性時，會根據字形的結構特性，而非根據相同的筆畫或部件。本研究進一步探討國小學生對中文字形相似性的分類。做法是令國小一、三、五年級的學童針對不同的中文單字做字形分類的作業，將這些字根據其形狀分類，把相似的字形分在同一類，不相似的分在不同類，並報告其所採用的分類原則。以分類結果計算相似性分數的高低，並依此相似性分數作群聚分析（cluster analysis）。所得結果顯示，隨著學齡的增加，識字量增加；而字形相似性判斷的依據也由局部細節漸次擴增至部件與結構。由一年級到五年級，以書法特徵與筆畫分類字形的傾向漸次遞減，而以結構、部首、部件（構形、非構形、相似、相關等部件）分類字形的傾向漸次遞增。與過去針對幼稚園孩童與大學生的研究相比，最大的不同在於國小學童會以相似部件及部件之間的關係作為分類的原則，且似乎無法避免字義的涉入。隨著學齡的增加，愈能察覺字形的結構特性，並將之運用於字形分類作業中。

Our previous studies (Yeh, Li, & Chen, 1997, 1999; Yeh & Li, 2002) showed that skilled readers of Chinese used the overall structure of the character in their judgments of visual similarity in Chinese characters. The purpose of this study was to further examine the perception of Chinese characters for children in elementary school, especially, the role of structure in their judgments of visual similarity of Chinese characters. School children of grade 1, 3, and 5 conducted a sorting task based on the similarity in the visual form of Chinese characters. The results of cluster analysis showed a perceptual trend from local features/strokes to more globally defined patterns (such as components and structures) with the increase of age and the number of characters these children possessed. Different from the previous results obtained from kindergartners and college students, children in elementary school had a higher tendency to sort characters based on similar components and the relationship between components, and they seemed unable to avoid the involvement of semantics in the shape-sorting task. As age and vocabulary knowledge increase, the tendency of extracting the character structure and using it in the shape-sorting task increases.