The purpose of the comparison group is to provide greater con-fidence that outcomes are a function of the intervention. For example, psychiatric inpatients are involved in a 12-week series of group therapy sessions in addition to their normal treatment regimen. The patients have not been randomly assigned to the group, though pre- and post-measures of symptom severity are taken from the intervention group and a nonintervention group. This nonequivalent design allows for both between-group and within-group comparison of symptom severity. The design, how-ever, is susceptible to selection bias (where differences between groups are not randomly distributed between them), which can be mitigated or controlled for by the use of some type of matching.