OBJECTIVES:  Communication impairment after stroke affects everyday activities and social participation. Speech and Language therapy (SLT) has waiting lists of 6 months plus in the National Health Service (NHS), implying a high opportunity cost. The aim was to evaluate the cost effectiveness of SLT versus attention control (AC), at six months, for people with communication difficulties due to aphasia/dysarthria following stroke. METHODS:  The economic analysis was an integral component of a randomised, controlled, pragmatic trial comparing SLT (n= 85) and AC (n=85). The perspective was the NHS and social care, patients and families.  The time horizon was 6 months (baseline to end of scheduled follow up). Resource use and health status (EQ-5D) were collected on all participants recruited into the trial. Utility values were estimated from the EQ-5D and associated population tariffs. The setting was inpatient and community/primary care in North West England; data were collected between 2006 and 2010. Unit cost data are for 2008-2009.  Regression models estimated incremental or costs and outcomes for the ICER, adjusted for predefined covariates.  Incremental costs and outcomes were bootstrapped to derive cost effectiveness acceptability curves, net benefit statistics and probability that SLT was cost effective. RESULTS:  The net cost of SLT was £110 (95% percentiles: -£640 to +£861). The net utility was 0.01 (95% percentiles: -0.03 to +0.04).  SLT is only likely to be cost effective if decision makers are willing to pay £25,000 or more to gain a 1 point increase in utility (p=0.50). The cost effectiveness of SLT depends on the outcome measure used and the baseline severity of stroke. CONCLUSIONS: The primary and sensitivity analyses indicated a high level of uncertainty suggesting it is not possible to conclude whether therapy is more or less cost effective than attention control.