Forty two preschool children referred to a speech therapy department were randomly allocated to receive intensive individual speech therapy or the more traditional once weekly approach. Boys and minority ethnic groups were referred most frequently. Speech therapy improved expression more than comprehension, as measured on Reynell scales. The mean improvements were 0.5 SDs (95% confidence intervals (CI) 0.3 to 0.7) and 0.3 (95% CI 0.1 to 0.5) respectively. There was a greater improvement in children receiving intensive compared with weekly therapy in the expression scores (0.8 SDs (95% CI 0.5 to 1.1) v 0.2 SDs (95% CI -0.1 to 0.5]. White and non-white children had similar improvements in comprehension scores but white children had greater improvement in expression scores (1.1 SDs v 0.3 SDs). This difference was seen in both therapy groups. Randomised trials are useful in the evaluation of speech therapy in children.