In investigating the effects of a treatment or experience on a group of patients, it is usually necessary to observe a control group. The comparison between the two groups may be influenced by the characteristics of the subjects: for example, their sex, age, body weight, and original clinical state. In order to secure comparability for such relevant "extraneous" factors, the controls are frequently matched with the experimental subjects. An alternative is to select the controls at random from an appropriate population, and to rely on statistical analysis to make allowance for the effects of concomitant variables. Medical research workers often prefer to select matched instead of randomly selected controls. The usual reasons seem to be that :