Sets of the most commonly monitored cardiovascular data have been collected retrospectively in four categories of patients following cardiac surgery: survivors and non-survivors with and without inotropic support. A data set was recorded in the surviving categories when the patient had achieved an optimum cardiovascular state, and in the nonsurviving categories 2 h before death. Three statistical methods of discriminating between the categories are compared. The best discriminants of cardiovascular status are peripheral skin temperature, mean arterial blood pressure, and urine output; there is little to be gained by using more than three variables. A method is suggested of reducing the relevant information in cardiovascular data to a single variable which can be used to predict future status.