Background: To examine longitudinal associations between diagnosis, joint attention, play and imitation abilities and language outcome in infants with autism and pervasive developmental disorder. Methods and Procedures: Experimental measures of joint attention, play and imitation were conducted with a sample of infants with autism spectrum disorder at age 20 months. Language outcome was assessed at age 42 months. A within-group longitudinal correlational design was adopted. Outcomes and Results: Language at 42 months was higher for children with a diagnosis of pervasive developmental disorder than for children with a diagnosis of autism. Language at follow-up was also positively associated with performance on experimental measures of joint attention and imitation, but not with performance on experimental measures of play and "goal detection" at 20 months, nor with a non-verbal intelligence quotient, although these associations were not examined independent of diagnosis. However, floor effects on the measure of play at 20 months and the small sample size limit the conclusions that can be drawn. Conclusions: Individual differences in infant social-communication abilities as well as diagnosis may predict language outcome in preschoolers with autism spectrum disorders. Attention should be directed at assessing these skills in 2- and 3-year-old children referred for a diagnosis of autism spectrum disorder. Imitation and joint attention abilities may be important targets for early intervention. (Contains 2 tables and 2 figures.)