The study used survey data from several large samples of at-risk youth and logistic regression analysis to test whether factors derived from strain, social learning, and social control theories contribute to the probability of gang membership. The study also tested whether peer delinquency, respondent's delinquency, and gang membership are observed measures of the same latent construct. Findings include: (1) The revised strain theory measure in which financial goals and educational means are disjointed contributes significantly to the probability of gang membership; (2) In social control theory, the likelihood of gang membership increases among youths whose mothers lack knowledge about where they are or whom they are with; (3) Significant correlates in learning/differential association theory include peer delinquency and a belief that it is "OK to get around the law"; (4) Girls are as likely as boys to identify with gangs; and (5) Many parental factors are irrelevant to gang membership, including how well the youth communicates with, shows respect for, and identifies with his or her parents. Tables, references