Children's social skills are an important class of learned behaviors that facilitate success in the classroom; the primary method used in the assessment of social skills involves having parents or teachers complete standardized checklists using judgments of frequency or intensity. Children's (N = 1,102) social skills were modeled as time-varying predictors of student achievements within a latent growth curve model that allowed for estimation of student level variation and the possibility of non-linear achievement growth across 4-5 years of age and grades one, three, and five. Separate models were examined to determine whether ratings provided by mothers' accounted for more achievement score variance than ratings provided by teachers', and multi-group time-varying conditional latent growth curve models were investigated for boys and girls separately by informant type. Results indicated that children's social skills accounted for levels of achievement score variance that were most pronounced at pre-school age, that teachers' ratings of children's social skills generally accounted for more achievement score variance than those obtained by mothers' regardless of the child's sex, and that the explanatory power of social skills for boys and girls was dependent upon the type of achievement considered. (Contains 1 figure and 4 tables.)