Cardiac output in normal subjects and in patients with various cardiovascular abnormalities was determined from indicator dilution curves recorded from the ear and/or an artery using five different methods for curve-area analysis. These methods were the conventional Hamilton, the gamma-function model, the tanks-in-series model, the simplified gamma method and the Dow formula. In normal subjects and in the various groups of patients, excluding cases with shock, a close agreement was found between estimates by the Hamilton method and all others. In all comparisons with the Hamilton technique, the gamma function, tanks-in-series and simplified gamma models systematically slightly overestimated the cardiac output. Estimates made by the non-Hamilton methods on patients with shock were markedly higher than those based on the Hamilton technique. It is concluded that, in such cases, values for output derived by the simplified gamma approach are more likely to approximate to the cardiac output than those based on the Hamilton technique. A simple approach to the identification of abnormal curves is described.