The article reports the results of a pilot study comparing traditional behavioural approaches and natural play interventions for young children with autism over a 10 week period. Two matched groups of eight young children with autism participated. Using a crossover design, children in both groups showed positive gains in compliance, attending, play and communication with their therapists and parents. Improvements in attending and compliance were higher following the behavioural condition compared with the natural play condition. Seven participants had reduced autism scores after the intervention. The findings suggest that behavioural and play approaches affect behaviour in different ways and that autistic symptomatology of young children may be amenable to treatment. The discussion focuses on the active ingredients of treatments and the need to base efficacy research on well-planned treatment comparisons.