This study examined the effects of a computer-assisted social skills intervention on the social competence and behaviour adjustment of students with special needs. The sample consisted of 114 male students, divided into two groups: 52 students with learning disabilities (LD) (mean age = 12.23; SD =1.12) and 62 students with behaviour disorders (BD) (mean age = 12.53; SD = 1.34). The students were trained with the “I Found a Solution” computer-assisted social skill package. The intervention consisted of the computer-assisted program and teacher-guided work. Each student was assessed, before and after the training, on: teacher-rated self-control and maladjusted behaviours, peer-rated social acceptance and self-rated loneliness. Through conceptualisation based on the students' loneliness and externalising maladjustment, four subgroups were identified: (a) relatively adjusted students; (b) externalising and not-lonely students; (c) lonely and non-externalising students; and (d) lonely and externalising students. The analyses included subgroups b, c and d, excluding the relatively adjusted subgroup, adding up to a sample of 83 students (35 students with learning disabilities and 48 students with behaviour disorders). A series of two-way ANOVAs with repeated measures (grouping by subgrouping, with pre-/post-training scores as the repeated measures) revealed that following training, students in both groups felt less lonely and were more accepted by peers. Teachers rated them as demonstrating higher levels of self-control and decreased externalising and internalising maladjustment. The differential impact of the training between groups (LD/BD) and among subgroups (b/c/d) highlighted the significance of the subtyping approach for understanding the children's social learning profiles and for directing effective social skill programming.