過去的實證研究已經一再地指出，偏差友伴與個人偏差行為間具
有強的關連性。但，當今社會學和犯罪學領域的研究中，卻有兩個相
互競爭的觀點嘗試解釋兩者間的關連。一為「影響性效果」，此觀點
認為偏差友伴確實地影響著個人的偏差行為發展；另一個觀點為「選
擇性效果」，主張個人會選擇跟自己相似的人結交朋友，或個人會「投
射」認為朋友的行為應跟自己是相似的，基於此觀點，偏差友伴與個
人偏差行為的高相關性只不過是「自我選擇」下的產物。本研究使用
2,280 位來自於「台灣青少年計畫」的台灣青少年樣本，並配上以班
級為單位的友誼網絡資料，用以檢測偏差友伴和個人偏差行為間的實
際連結。為了排除「選擇性」及「投射性」的偏誤，本文除了使用貫
時性資料外，並且使用兩種不同偏差友伴的測量方法，一為傳統的自陳式測量，另一種為社會網絡測量的技術。分析結果顯示，縱使「選
擇性」、「投射性」和「樣本流失性」的偏誤排除後，偏差友伴和個人
偏差行為間仍具有實際的關連性。另外，本文進一步發現，自陳式測
量相較於社會網絡測量法，更有可能高估偏差友伴對偏差行為的解釋
力。根據分析之結果，本研究建議，當使用自陳式測量的偏差友伴預
測個人偏差行為時，應同時在分析模型中考慮「自我控制」和「父母
監控行為」對個人偏差行為的效果，用以減低自陳式測量所造成的「投
射性」偏誤。

Consistently, empirical findings have indicated that the presence
of deviant peers is one of the strongest predictors of adolescent delin-
quency. While this association is widely accepted, the mechanisms at
work have been disputed. Two competing approaches, the influence
and selection approaches, have received the most scholarly attention.
The influence model argues that an association with deviant peers has
a real and causal impact on adolescent delinquency. Conversely, the
selection model posits that perceived associations are the result of
homophily, as people tend to form or select friendships with others
who share similar characteristics. Using a sample of 2,280 Taiwanese adolescents from the Taiwan Youth Project along with friendship net-
work data, this study reexamines the association between deviant
peers and delinquency. In order to rule out potential biases, this study
uses both perceptual and direct measures of deviant peers. The per-
ceptual measure of deviant peers refers to an individual’s report of
his or her friends’ delinquent behavior, whereas the direct measure is
formed through an analysis of the friendship network data. Control-
ling for selection and projection effects, the results show that the
effect of deviant peers on delinquency remains significant. The
results also indicate that using a self-report measure of deviant peers,
in contrast to a social network analysis, potentially overestimates the
deviant peers-delinquency relationship. Finally, this study’s results
support the combined use of both self-report measures of deviant
peers in conjunction with self-control and parental monitoring mea-
sures to best account for the influence of deviant friendships on
delinquency.