BACKGROUND Platelets are involved in myocardial infarction but evidence of prediction of infarction by measures of platelet function are sparce. METHODS Platelet aggregation to thrombin and to ADP in platelet rich plasma was recorded for 2176 men aged 49-65 years in the Caerphilly cohort study. RESULTS Results from 364 men were excluded, 80 of whom had not fasted before venepuncture; most of the others were excluded because antiplatelet medication had been taken shortly before the platelet tests. During the five years following the platelet tests 113 ischaemic heart disease (IHD) events which fulfilled the World Health Organisation criteria were identified--42 fatal and 71 non-fatal. No measure of platelet aggregation was found to be significantly predictive of incident IHD. The possibility that platelet function is predictive for only a limited time after it is characterised, and that prediction falls off with time, was tested. When IHD events are grouped by their time of occurrence after aggregation had been measured, the test results show a gradient suggestive of prediction of early IHD events. Thus, 24% of the men who had an event within 500 days of the test had had a high secondary response to ADP while only 12% of those whose IHD event had been 1000 or more days after the test had shown a high platelet response at baseline. The trend in these proportions is not significant. CONCLUSIONS Platelet aggregation to thrombin and ADP in platelet rich plasma was recorded in the Caerphilly cohort study. No measure of aggregation was found to be predictive of IHD.