The present study empirically examined three theoretical approaches designed to predict risk for delinquency during adolescence: an individual difference perspective, a social interactional model, and a social contextual approach. Hypotheses derived from each perspective were tested using two independent samples of early adolescents followed over a 4-year period. Six-hundred sixty-seven children (in sixth grade at Time 1), and their parents comprised the first sample (Project Family); and 451 children (in seventh grade at Time 1), their parents, and a close-aged sibling made up the second sample (Iowa Youth and Families Project). Results from a series of structural equation models suggested that a social contextual approach provided the best fit with the data across both samples and genders. Consistent with the social contextual approach, results indicated that a lack of nurturant and involved parenting indirectly predicted delinquency by increasing children’s earlier antisocial behavior and deviant peer relationships; child antisocial behavior also predicted similar decreases in nurturant parenting over time. Both child antisocial behavior and deviant peer affiliations at Time 2 predicted delinquency 1 year later. Implications for theoretical development and future research priorities are discussed.