The ideal diagnostic test should discriminate unerringly between diseased and healthy individuals. When such a test is available, there is no statistical problem. The search for an unerring test yields a series of tests that achieve partial success. There is, therefore, a need for an index to rate diagnostic tests in an objective manner and to provide a means of deciding whether two diagnostic tests really differ in their capacity to discriminate between healthy and diseased individuals. Diagnostic tests undergo modifications in attempts to improve their performance. The medical researcher needs a statistical tool to assist him in detecting as early as possible whether a particular modification has led to an improvement in the results obtained with the diagnostic test. There are two questions that arise in appraising a diagnostic test to which it is not possible to give completely satisfactory answers. First, whenever a diagnostic test is subjected to study, the question arises as to how sure the experimenter can be that his controls are really healthy and that all his