Tests of mediation in treatment trials can illuminate processes of change and suggest causal influences in development. We conducted a mediation analysis of a previously published randomised controlled trial of parentmediated communication-focused treatment for autism against ordinary care, with 28 children aged 2–5 years (Aldred et al. in J Child Psychol Psychiatr 45:1–11, 2004). The hypothesised mediating process, targeted by the intervention, was an increase in parental synchronous response within parent-child interaction. The results showed partial mediation, with change in synchrony accounting for 34% of the positive intervention effect on autism symptomatology (Autism Diagnostic Observation Schedule communication and social domain algorithm); the result was confirmed by bootstrap estimation. Improved parental synchronous response to child communication can alter short-term autism symptom outcome with targeted therapy.