Much of the comparative research on stimulus overselectivity has been flawed by either failure to control for chronological age and language ability of the subjects or reliance on the controversial technique of matching on mental age. The present study investigated the prevalence of overselectivity in autistic, trainable mentally retarded, and nonhandicapped children demonstrating some expressive speech. The ages of the children were between 6 years-6 months and 9 years-3 months. Thus, chronological age and language ability were controlled, rather than allowed to vary unsystematically. Results indicated no significant differences between the autistic and TMR samples, but significant differences between the handicapped samples and the non-handicapped group. Some, but not all, of the handicapped children displayed overselectivity.