From 1979 to 1985, 2435 patients thought to have had a transient ischaemic attack or minor ischaemic stroke were allocated at random to receive long term blind treatment with either aspirin 600 mg twice daily (n = 815), aspirin 300 mg once daily (806), or placebo (814). Treatment continued with about 85% compliance until September 1986 (mean four years). The odds of suffering one or more of four categories of event--namely, non-fatal myocardial infarction, non-fatal major stroke, vascular death, or non-vascular death--were 18% less in the two groups allocated to receive aspirin than in the group allocated to receive placebo (2p = 0.01). The more relevant but less frequent composite event of disabling stroke or vascular death was reduced by only 7%; this reduction was not significantly different from zero, but nor was it significantly different from a 25% reduction. There was no definite difference between responses to the 300 mg and 1200 mg daily doses, except that the lower dose was significantly less gastrotoxic.