This article explores the plausibility of the conflicting theoretical assumptions underlying the main criminological perspectives on juvenile delinquents, their peer relations and social skills: the social ability model, represented by Sutherland's theory of differential associations, and the social disability model, represented by Hirschi's control theory. It does so by applying the principles associated with Granovetter's theory of strong and weak ties, drawing out its theoretical implications for the criminological theories and deriving a number of hypotheses. The article uses "strategic research materials" in the form of a data set strongly biased towards the social disability model and containing information on co-offending relations among a population (N = 580) of juveniles in a Swedish town during a three-year period. Results from cross-sectional empirical analyses using, among other things, the Quadratic Assignment Procedure, clearly support the social ability model. The robustness of the initial analyses is checked through longitudinal analyses applying actor-oriented statistical models for network evolution. The article's implications for criminological, as well as multidisciplinary, research are discussed. (Contains 6 notes, 4 figures and 6 tables.)