Drop-outs in group psychotherapy are an everyday clinical problem. A premature quitting by one patient may have great consequences, especially in closed groups, not only for the drop-out but also for the remaining group members. Compared to the clinical relevance, empirical research on the conditions leading to a premature termination is quite meagre. Empirical basis of the study presented here are 58 groups with a total of 445 patients with a variety of symptoms and personality structures. All groups started with a 3 month inpatient period and were continued as a closed outpatient group for about 2 years. 110 patients (24.7%) left the group at least 3 months before the agreed end of the group. Empirical research on conditions leading to the premature drop-out is based on three perspectives: 1. patient-oriented: The values of certain patient variables e.g. symptoms, personality structure, socio-economic status etc. are considered as potential predictors. 2. institution-oriented: Here group characteristics are considered as the pool of potential predictors: e.g. group size, decisions made under time pressure, cotherapy etc. 3. group-oriented: The values of patient variables for one patient in relation to other group members (e.g. "isolated position") is considered here. From each perspective we were able to find potential relevant predictors.