Few attempts have been made to conduct randomised control trials (RCTs) of interventions for pre-school children with autism. We report findings of a pilot RCT for a parent training intervention with a focus on the development of joint attention skills and joint action routines. Twenty-four children meeting ICD-10 criteria for childhood autism (mean age = 23 months) were identified using the CHAT screen and randomised to the parent training group or to local services only. A follow-up was conducted 12 months later (mean age = 35 months). There was some evidence that the parent training group made more progress in language development than the local services group. However, the present pilot study was compromised by several factors: a reliance on parental report to measure language, non-matching of the groups on initial IQ, and a lack of systematic checking regarding the implementation of the parent training intervention. Furthermore, three parents in the local services group commenced intensive, home-based behavioural intervention during the course of the study. The difficulties encountered in the conduct of RCTs for pre-school children with autism are discussed. Methodological challenges and strategies for future well-designed RCTs for autism interventions are highlighted.