Twenty-two children with autism spectrum disorders who had not responded to supported behaviour management strategies for severe dysomnias entered a double blind, randomised, controlled crossover trial involving 3 months of placebo versus 3 months of melatonin to a maximum dose of 10 mg. 17 children completed the study. There were no significant differences between sleep variables at baseline. Melatonin significantly improved sleep latency (by an average of 47 min) and total sleep (by an average of 52 min) compared to placebo, but not number of night wakenings. The side effect profile was low and not significantly different between the two arms.