In Reply.
—Leviton and colleagues question our choice of a control group.1 As noted by Kelsey et al,2 recognized authorities in the area of comparative studies, "Choice of the most appropriate control group is one of the most difficult and controversial aspects of study design... [and] in many situations no single control group is obviously best. In such situations it is helpful if... more than one case-control study of an area of interest is undertaken by different investigators using different types of control groups...."2 The approach suggested by Leviton et al is one way of investigating the problem, but not the only way as they aver. If controls are selected on the basis of epilepsy as Leviton et al suggest, there is no way to determine whether psychopathologic factors differ in epileptic and nonepileptic suicide attempters, nor can the effect of chronic psychosocial stressors related to having