Objective: To compare routine speech and language therapy in preschool children with delayed speech and language against 12 months of “watchful waiting.” Design: Pragmatic randomised controlled trial. Setting: 16 community clinics in Bristol. Participants: 159 preschool children with appreciable speech or language difficulties who fulfilled criteria for admission to speech and language therapy. Main outcome measures: Four quantitative measures of speech and language, assessed at 6 and 12 months; a binary variable indicating improvement, by 12 months, on the trial entry criterion. Results: Improvement in auditory comprehension was significant in favour of therapy (adjusted difference in means 4.1, 95% confidence interval 0.5 to 7.6; P=0.025). No significant differences were observed for expressive language (1.4, −2.1 to 4.8; P=0.44); phonology error rate (−4.4, −12.0 to 3.3; P=0.26); language development (0.1, −0.4 to 0.6; P=0.73); or improvement on entry criterion (odds ratio 1.3, 0.67 to 2.4; P=0.46). At the end of the trial, 70% of all children still had substantial speech and language deficits. Conclusions: This study provides little evidence for the effectiveness of speech and language therapy compared with watchful waiting over 12 months. Providers of speech and language therapy should reconsider the appropriateness, timing, nature, and intensity of such therapy in preschool children. Continued research into more specific provision to subgroups of children is also needed to identify better treatment methods. The lack of resolution of difficulties for most of the children suggests that further research is needed to identify effective ways of helping this population of children.