To determine whether the site of acute myocardial infarction (AMI) can be predicted on the basis of a previous coronary angiogram, 184 consecutive angiograms obtained between March 1972 and August 1990 in 92 patients who had undergone coronary angiography both before and after AMI without intervening bypass surgery or angioplasty were evaluated. Median time between the first coronary angiography and AMI was 26 months (range 1 to 144). On the first angiogram, most patients (89%) had 1- or 2-vessel disease, and 56 (61%) had an abnormal ventriculography. Seventy-two segments (78%) responsible for a future AMI were not significantly stenosed. On the second angiogram, AMI was related to the previously most stenotic segments in only 29 patients (32%). For these patients, median time between first coronary angiography and AMI was slightly shorter (22 vs 28 months; p = 0.04). The severity of the narrowing on the first angiogram was a poor predictor of subsequent AMI. It is concluded that in a selected, medically treated cohort, AMI is frequently related to a segment that was not the most stenotic one or was not even significantly stenosed at previous angiography, particularly with a long interval between the first angiogram and AMI.