Achieving consensus on the definition and measurement of social competence (SC) for preschool children has proven difficult in the developmental sciences. We tested a hierarchical model in which SC is assumed to be a second-order latent variable by using longitudinal data (N = 345). We also tested the degree to which peer SC at Time 1 predicted changes in positive adjustment from Time 1 to Time 2, based on teacher and peer ratings. Using a multiple-method data-collection strategy, information for three subdomains of SC (social engagement/ motivation, profiles of social interaction and personality assets assessed with Q-sorts, peer acceptance) were collected across consecutive years in preschool programs. Longitudinal confirmatory factor analyses (CFAs) demonstrated invariance of both the measurement and the structural models across age levels and yielded a cross-time path weight of 0.74 for the second-order factor. Analyses of latent means suggested significant increases in SC scores from the first year to second year of participation, and longitudinal cases in their second year of participation had higher scores than did age peers who entered the program as older children. Finally, Time 1 SC predicted increases from Time 1 to Time 2 for SC-relevant indicators rated by teachers and peers (standardized path coefficient of 0.29, p less than 0.001). (Contains 2 figures and 2 tables.)