Background: Deficits in joint attention (JA) and joint engagement (JE) represent a core problem in young children with autism as these affect language and social development. Studies of parent-mediated and specialist-mediated JA-intervention suggest that such intervention may be effective. However, there is little knowledge about the success of the intervention when done in preschools. Aim: Assess the effects of a preschool-based JA-intervention. Methods: 61 children (48 males) with autistic disorder (29-60 months) were randomized to either 8 weeks of JA-intervention, in addition to their preschool programs (n = 34), or to preschool programs only (n = 27). The intervention was done by preschool teachers with weekly supervision by trained counselors from Child and Adolescent Mental Health Clinics (CAMHC). Changes in JA and JE were measured by blinded independent testers using Early Social Communication Scale (ESCS) and video taped preschool teacher-child and mother-child play at baseline and post-intervention. Clinical trials registration: Clinicaltrials.gov: NCT00378157. Results: Intention-to-treat analysis showed significant difference between the intervention and the control group, with the intervention group yielding more JA initiation during interaction with the preschool teachers. The effect generalized to significantly longer duration of JE with the mothers. Conclusions: This is the first randomized study to show positive and generalized effects of preschool-based JA-intervention.