兒童性格結構：五大模型的本土化檢驗

THE STRUCTURE OF PERSONALITY IN TAIWANESE CHILDREN: AN INDIGENOUS LEXICAL APPROACH TO THE BIG FIVE MODEL

本研究使用中文性格形容詞，探討特質論的五大模型（Goldberg, 1993）是否可以描述台灣兒童的主要性格差異。國小五、六年級學童1,779位以150項中文性格形容詞來評定自己個性，其中309位學童並由其級任老師使用同一套形容詞加以評定其性格。性格形容詞主要採自兒童適用的《國語日報辭典》與兒童知覺同儕之個性的用語。就教師評定的資料進行因素分析，結果發現兒童之性格因素主要包括和善性、聰穎開放性、嚴謹自律性、神經質、及外傾支配性等五項因素。此一因素結構大抵吻合五大模型，且可詮釋圓周環現象。就學生自評之資料進行因素分析，則發現兒童性格因素主要包括和善性、神經質、內外向、聰穎嚴謹性、愛現自恃性、攻擊性等六項。自評的性格結構不吻合五大模型，而且隨性別而有所差異。本研究依據因素分析結果編製量尺，發現性格特質的評定具有自評與他評的聚合與區分效度，顯示兒童性格因素結構並非語意類似性的產物。各項性格特質間隔一年的穩定係數至少高於.43，顯示兒童個性具持續性。文中並就檢驗五大模型是否適用於華人所涉及的概念與測量問題加以討論。

Searching for personality attribute lexicons listed in a popular Chinese children dictionary found that nearly all of them could be classified into the Big Five factors. This was also true for the children' spontaneous perception of their classmates' personalities. Like English, there were more Chinese adjectives associated with agreeableness and conscientiousness. These findings provided preliminary support for the lexical hypothesis underlying the Big Five model (Goldberg, 1993). To test whether children's personality structure fits the Big Five model, one thousand and eight hundreds elementary school students with between I 1 to 13 years of age were sampled from 54 classrooms of 27 schools across Taiwan. They rated their own personalities for each of the 148 trait adjectives using a five-point scale. Of these adjectives, Ill were drawn from a children's dictionary and 16 were purely from children's spontaneous descriptions of their classmates. Besides self-rating, five or six students of each class (total N = 309) were also rated by their teachers using the same set of adjectives. Exploratory factor analysis was conducted to delineate the structures of these two sets of rating. The Scree test and meaningfulness of factors were used as criteria to decide on the appropriate number of factors. The oblique rotation that is less likely to distort the structure of data was adopted to generate the factor structure of personality. The results obtained from the two sources of data had different implications for the Big Five model. The result from the teachers' ratings of children revealed the Big Five factors: agreeableness, conscientiousness, intellect/openness, neuroticism, and extraversion/surgency. The extraversion/surgency factor referred to an active or dominant tendency that may be involved in kind or malicious behavior. More than twenty trait adjectives that had loading on the extraversion/surgency factor also had positive or negative loading on the agreeableness factor, thereby leading to a quasi-circumplex structure in which personality attributes appeared in a circular order around the orthogonal coordinates of extraversion/surgency and agreeableness. Because many personality-attribute terms relate in a complex manner to more than one Big Five factor, it seemed likely to find other circumplex structures. To some extent, circumplex structures can be interpreted in terms of the framework of the Big Five model. All of these findings from teachers' ratings of students seemed to be rather consistent with the lexical perspectives on the Big Five factor structure (Saucier & Goldberg, 1996). On the other hand, the factor-analytic results for students' self-rating data did not fit well with the Big Five model. Six factors were obtained for the sample of all students: agreeableness, neuroticism, extraversion. conscientiousness/intellect-openness, narcissistic tendency, and aggressiveness. The contents of several factors, including narcissistic tendency, aggressiveness. and extraversion, somewhat varied with sex. Agreeableness factor for students' self-rating tended to be a unipolar factor that was somewhat independent of aggressiveness, whereas agreeableness obtained from teachers' ratings tended to be a bipolar factor. Because the results from self-rating and teachers' rating were different, the conceptual and measurement issues involved in testing the cross-cultural generality of the Big Five model were discussed. Based on the results of factor analysis, several scales of measuring major traits were also constructed to test the convergent and discriminant validity of trait ratings. All ratings of major traits were found to be valid. More important, all of them showed moderate to high stabilities over one year. These findings provided the supportive evidences of constructive realities for the high-level traits in the children.