Operations research is a novel approach to the problem of making sound decisions in complex matters. Conceived during the last war by scientists helping the war effort, it has proved its usefulness in many fields, as different as hunting enemy submarines and determining optimal rates of industrial production. Very briefly, the operations research approach consists in determining one's objectives, selecting meaningful measures of achievement, formulating a mathematical model of the problem to show both the variables under control and those not subject to control, and indicating the results of alternate courses of action (2). Problems of roentgen diagnosis are many, but they do not greatly resemble those of finding hidden submersibles, except for the trying necessity of making important decisions when information available is incomplete or uncertain. The art of diagnosis is the art of evaluating equivocal evidence. Problems of patients with partial proof provide the pitfalls and pleasures of roentgen iconography, the pride of predictions come true, and the pains, and lessons, of insight by hindsight. If it were otherwise, it would not be the same; if roentgen findings would always lead to conclusions that are obvious and certain, examining patients would be no more interesting than examining metal castings. Since the normal merges imperceptibly with the abnormal in size, shape, structure, and function, examinations in clinical medicine cannot make sharp distinctions. The body weight, the blood hemoglobin level, or a chest roentgenogram reveal the obviously healthy and the obviously sick but leave us with the puzzle of a third group, those with equivocal findings. Some aspects of the problem of “borderline” observations are shown as simple numerical concepts, such as sensitivity and specificity of a given method, and others. Since diagnoses are predictions, only future events can make them true or false; true positive, false positive, true negative and false negative refer to the way a diagnostic opinion fares in the light of later findings; a “positive” reading in a patient who turns out to have been “positive” constitutes a true positive. Diagnostic statements are confirmed by events, not authority. A clinical test, or sign, is said to be sensitive if it helps find most of those affected among those examined—it yields few false negatives; unless such a test is also specific, it may include among the sick many who are not (many false positives). Roentgen signs of pulmonary emphysema in aging people could be said to be sensitive, but not very specific; we miss few of those who have the disease, but may suspect many who do not have it. A test is called wholly specific (3) if it is positive only in those having a certain disease—it gives no false positives; unless it is sensitive at the same time, such a test may give many false negatives.