One hundred-one fourth, fifth, and sixth graders at a Southern California elementary school were randomly assigned by classroom to either a social skills training program or a no treatment control condition. On the basis of teacher ratings, subpopulations of subjects were identified: 17 Underachievers, 40 Disruptive children, 29 children with Multiple Problems, and 15 Exceptional children. In the intervention classes, subjects were randomly assigned to groups of 6 subjects each and received training for one hour twice a week for 12 weeks. Measures of social skills, peer popularity, teacher ratings, and grade-point averages were collected to evaluate the intervention. The results indicated that subjects improved most in the area not targeted as a problem for the subject; i.e., Underachievers improved their social relationships, Disruptive children increased in academic skills, and Exceptional children increased significantly in both areas. A discussion followed of the optimum intervention strategy, building on children's existing strengths vs. targeting deficits.