Between 30 and 60% of patients drop-out of institutional psychiatric treatment. There are few studies on this issue and these have not provided a clear understanding of this fact. Although it is a different therapeutic setting, there are many studies on patients' dropout in psychotherapy: the influence of many patient sociodemographic variables such as gender, age, diagnostic, were studied without providing strong and regular links with early dropout. Other, more relational variables (such as object relation and interpersonal functioning), gave stronger results although insufficiently confirmed by different studies. A third kind of variable involves the concrete relationship between patient and therapist (therapeutic alliance, patient's expectations) and provided interesting results (but not easy to use in institutional treatment). The aim of this study is to provide data to understand patients' institutional dropout in a French psychiatric centre. The latter's aim is psychosocial rehabilitation for schizophrenic and borderline patients. Thirty percent of these dropout during psychosocial treatment. According to the specificities and aims of this psychiatric centre, we hypothesize that there are strong links between relational dimensions (objet relation, interpersonal functioning), subjective evaluation of pathology intensity, and early dropout. Thirty-one subjects; 65% schizophrenic, 23% borderline, 13% other (according to the ICD10 criteria); 71% females; mean age 34 years (min=23; max=55); mean education level=3.4 (2 years of high school university). to have dropped out before 6 months' of the treatment, or continuing the treatment after 6 months (mean of treatment for all patients=15 months). patients present in the service for less than 6 months). At the beginning of the treatment, each patient (informed consent provided) underwent a psychological assessment with: Échelle d'aptitude psychosociale (EAPS) for assessing interactional functioning; SCL90-R for assessing the intensity of psychopathology; TAT (with Social Cognition and Object Relation Scale [SCORS] scales) and Rorschach (with Mutuality Of Autonomy [MOA] scale) for assessing object relations. After 6 months of treatment, each patient was evaluated with a five-point scale (dropout and continuity scale), which assessed the investment in the treatment (criteria: dropout at 3 or 6 months or continuity; according or not to the centre's professionals; level of assiduity). We have correlated this variable with EAPS, SCORS and MOA. In addition, we have calculated statistical relationships between age, gender, diagnostic, education level and early dropout. Correlation was found neither between interactional functioning (EAPS) and dropout nor between object relations (SCORS and MOA) and dropout. Correlations were found between the dropout and intensity of the psychopathology (SCL90-R): the more the patient sees himself suffering, the more he invests the centre and the less he drops out (Spearman R=0.37, P<0.05). No differences were found between the dropout (N=10) and continuity group (N=21) regarding age, gender and diagnostic. However, a correlation was found regarding the education level: the more patients are educated, the more they continue the treatment (R=0.45; P<0.05). The dropout (and the continuity of treatment) seems more likely related to concrete variables such as psychological and relational suffering, educational level in this study, than structural psychological variables such as object relation, relational skills, and diagnostic. Other studies are necessary for a better understanding of these drops out. An interesting way should be the study in institution of the therapeutic alliance.