過去研究指出友誼能提供規範、資訊與資源，為青少年的學業成 就帶來正向影響，且此影響在低學業成就的青少年身上更加彰顯(Frank et a1. 2008) ，由此觀點，友誼應有助於縮減青少年因家庭背景不同帶來的學業成就差距。然而，相似性原則卻指出，若家庭背景或學業成就是青少年據以選擇朋友的關鍵因素，那麼友誼反而可能因此擴大不同背景青少年既有的學業成就差距。為驗證此二對立競爭之研究假設，本研究使用「台灣青少年成長歷程研究計畫」資料中台北縣、台北市及宜蘭縣共72 個班級、2 ， 527 位青少年的資料，利用二階段多層次動態社會網絡分析探究青少年的家庭社經背景、友誼網絡與學業成就在國中三年期間的動態改變歷程。研究結果顯示，家庭社經背景、教育期望與學業成就是青少年據以選擇友誼的重要因素，這樣的相似性選擇使得正向同儕影響在高成就的青少年身上更為彰顯。亦即，背景好、成績好的青少年在某種程度上透過相似性選擇鞏固其既有的學業成就優勢，階層式的友誼配對使得背景好的青少年從友誼網絡中獲益更多，兩者加乘，青少年的友誼網絡最終很可能是擴大而非縮小既有的教育成就差距。

Friendship is documented to have positive peer effects on adolescent academic performance by providing norms, information, and resources, and these kinds of peer effects are even more significant on low-achieving students if they can nest in a high-alChieving student group (Frank et al.,2008). The homophily principle that dominates the adolescent-friendship mixing pattem, however, indicates that adolescent friendship might be sorted by student family background and achievements, which implies that friendship might not be able to reduce the achievement gap and could even make the gap wider. To test these two competing hypotheses, we conduct a dynamic network analysis to investigate how family background influences the co-evolution process of students' friendship networks and academic performance by using data from the Taiwan Youth Project, which includes a research sample of 2,527 students nested in 72 classes. Findings from a series of two-step multilevel network analyses demonstrate that students might self-segregate by selecting homophilious friends. This segregation makes low-SES and low-performing students less likely to be nominated by high-SES or high-performing students; differences in probabilities hence exist for different achievement students to improve their academic performance. As a resu1t, adolescent friendship networks in the junior-highschool years may eventually help high-performing students maintain and exacerbate their achievement advantages.