Numerous variables have been identified as having prognostic value after infarction. The significance of each of these depends both on the time after the event when the observation is made and the length of follow-up. Although much prognostic information has been published, its validity is difficult to establish because the criteria for infarction have seldom been stated, the effect of treatment has been ignored and the case material has been either ill-defined or atypical. Most of the most powerful predictors of medium and long-term prognosis are not susceptible to correction, e.g. age, previous myocardial infarction, cardiomegaly, enzyme levels, intraventricular conduction defects and indices of left ventricular dysfunction. Others, e.g. "warning arrhythmias" may be treatable but there is, as yet, no evidence that suppressing them improves prognosis. The most clinically relevant prognostic factors are those which are of relatively high predictive value but are potentially correctable. These include smoking, hypertension, diabetes, life-threatening arrhythmias, and exercise-induced ischaemia, as manifested by angina and ST changes. Secondary prevention is most likely to be successful if appropriate approaches are aimed at specific subsets. Thus, beta-blockade, anti-arrhythmic therapy, platelet active agents, and surgery may each be beneficial in different although possibly overlapping sub-sets. Even if it is shown that any one of these forms of therapy produces a statistical benefit when given to the whole post-infarct population, it does not follow that it should be given to all members of such a population.