To determine the diagnostic accuracy of physicians for identifying patients with acute myocardial infarction (AMI) without an electrocardiogram (ECG). All patients in Göteborg with suspected AMI below 75 years of age who called for an ambulance or came directly to one of the two city hospitals with a delay time of less than 2 h 45 min from the start of symptoms. As part of the TEAHAT study (comparing rt-PA and placebo in AMI), we asked physicians to judge on a 1-5 scale (1 = no suspicion; 5 = convinced) how strong their suspicion of AMI was prior to interpreting the ECG. Among patients evaluated outside hospital with 4 or 5 on the scale, i.e. either a strong suspicion of AMI or the physician felt convinced about the diagnosis, 45% had ST elevation and 48% developed AMI during the first 3 days in hospital. The corresponding values for patients evaluated in hospital were 67 and 70%, respectively. We found that physicians could not accurately distinguish patients with AMI from those without based on clinical criteria without the help of an ECG.