Objective: To determine the clinical effectiveness, cost-effectiveness and service users' views of enhanced early communication therapy by speech and language (SL) therapists compared with attention control (AC). Design: Successful feasibility study followed by a randomised trial with economic evaluation, and nested qualitative study using 32 individual interviews. Setting: Twelve English NHS hospital and community stroke services. Participants: One hundred and seventy adults with aphasia or dysarthria admitted to hospital with stroke, December 2006 to January 2010. Eligibility determined by NHS SL therapists. Seventeen people declined follow-up. Interventions: A best-practice, flexible intervention by NHS SL therapists, up to three contacts per week for up to 16 weeks compared with a similar number of AC contacts by employed visitors. Main outcome measures: Primary outcome was blinded, functional communicative ability 6 months post randomisation on the Therapy Outcome Measure activity subscale (TOM). Secondary outcomes were participants' perceptions on the Communication Outcomes After Stroke scale (COAST); carers' perceptions of participants from part of the Carer COAST; carer well-being on Carers of Older People in Europe Index and quality-of-life items from Carer COAST. Serious adverse events (SAEs) were recorded. Economic evaluation: participants' utility (European Quality of Life-5 Dimensions), service use and cost data from medical records and carers, and a discrete choice experiment. Results: Intervention typically started after 2 weeks, providing 22 contacts. Both groups improved on the TOM. The estimated 6 months' group difference [95% confidence interval (CI)] was 0.25 (-0.19 to 0.69) points in favour of SL therapy. Sensitivity analyses adjusting for baseline chance imbalance or not imputing values for decedents further reduced this difference. Per-protocol analyses rejected a possible dilution of therapy from controls refusing allocation and receiving NHS SL therapy. There was no evidence of added benefit of therapy on any secondary outcome measure or SAEs, although the latter were less frequent in the therapy group [odds ratio 0.42 (95% CI 0.16 to 1.1)]. Regardless of group allocation, interviewed participants reported positive impacts on their confidence and mood, identified drivers for change and valued early and sustained contact. Health economic analysis indicated a high level of uncertainty. Early enhanced SL therapy for communication is likely to be cost-effective only if decision-makers are prepared to pay ≥ £25,000 to gain one unit of utility. Conclusions: These findings exclude the possibility of a clinically significant difference of 0.5 points on the TOM. There was no evidence, on any measure, of added benefit of early communication therapy beyond that from AC. It is unclear whether therapy is more or less cost-effective than AC. Early, frequent contact was highly valued by users and had good uptake. Functional communication improved for both groups, plausibly due to natural recovery and early and regular opportunity to practise everyday communication with a professional (therapist/visitor). There is no evidence to recommend enhancing the provision of early communication therapy by a qualified SL therapist over and above usual care. SL therapy service reorganisation should consider skill mix and timing within a stepped care model and should take place within the context of a trial. Trial registration: Current Controlled Trials ISRCTN78617680. Funding: This project was funded by the NIHR Health Technology Assessment programme and will be published in full in Health Technology Assessment; Vol. 16, No. 26. See the HTA programme website for further project information. The Stroke Association funded part of the excess treatment costs. © Queen's Printer and Controller of HMSO 2012.