A repeated-measures, waiting list control design was used to assess efficacy of a social skills intervention for autistic spectrum children focused on individual and group LEGO play. The intervention combined aspects of behavior therapy, peer modeling and naturalistic communication strategies. Close interaction and joint attention to task play an important role in both group and individual therapy activities. The goal of treatment was to improve social competence (SC) which was construed as reflecting three components: (1) motivation to initiate social contact with peers; (2) ability to sustain interaction with peers for a period of time: and (3) overcoming autistic symptoms of aloofness and rigidity. Measures for the first two variables were based on observation of subjects in unstructured situations with peers; and the third variable was assessed using a structured rating scale, the SI subscale of the GARS. Results revealed significant improvement on all three measures at both 12 and 24 weeks with no evidence of gains during the waiting list period. No gender differences were found on outcome, and age of clients was not correlated with outcome. LEGO play appears to be a particularly effective medium for social skills intervention, and other researchers and clinicians are encouraged to attempt replication of this work, as well as to explore use of LEGO in other methodologies, or with different clinical populations.