Of 2608 consecutive patients with acute myocardial infarction, 24 developed subacute free wall rupture (= 0.92%; 95% C.I. = 0.6-1.4). Clinical manifestations varied widely (shock on admission; 25% of cases; severe arrhythmias followed by shock: 17%; shock during hospital stay: 42%; symptoms suggestive of infarct extension without shock: 17%). The electrocardiograms were confusing rather than revealing: 56% of patients showed new ST segment elevations of 0.2 to 1 mV in the infarct-related leads, while autopsy or creatinine phosphokinase evidence of infarct extension was missing. In the first 21 cases, therefore, no definitive diagnosis was made before autopsy. Using 197 infarct patients in cardiogenic shock or with infarct extension during the acute stage, i.e. a patient group with comparable clinical manifestations, as control group, a logistic regression model was generated in which the variables age, lateral wall involvement and history of hypertension were used for estimating the probability of subacute rupture. In fact, probability may rise to more than 40% in major subgroups. As death occurred after a median interval of 8 h (45 min-6.5 weeks) following the onset of rupture symptoms, echocardiography must be performed urgently in all cases presenting symptoms of shock or infarct extension. Pretest probability which can be roughly estimated from our model as well as sensitivity and specificity of individual echocardiographic or clinical parameters are indispensable for correct therapeutic decisions. The routine application of this algorithm in our department contributed to a timely diagnosis in the last three consecutive cases of whom one patient survived.