Immediately before their second group therapy session, 10 newly formed inpatient therapy groups were randomly assigned to complete either collective or individual art projects. The members of a group in the collective-project condition completed a single art project as a group. Each member of a group assigned to the individual project condition completed an individual art project. In the therapy session which followed, groups in the collective-project condition showed a decrease in physical distance between group members relative to a baseline therapy session and relative to groups in the individual-project condition. These groups also scored higher on a questionnaire measure of cohesion. The questionnaire and intermember-distance measures of group cohesion correlated highly. Intermember distance returned to baseline during a follow-up therapy session. No differences were found during the therapy sessions in the frequency of three types of verbal behavior. It was concluded that collective art projects may be used to increase cohesion in ongoing therapy groups. Further exploration of intermember distance as a measure of group cohesion was suggested. (Author)