Objectives Among interventions that have been shown to be efficacious in the treatment of depression, behavioural activation (AC) is receiving increasing attention as the evidence supporting its effectiveness continues to accumulate. Although the efficacy of AC for the treatment of depression has been established through numerous randomized controlled trials, studies evaluating the effectiveness of AC when implemented in mental health settings are rare and there is insufficient supportive data. This step is, however, essential to the validity and the generalization of the treatment to the reality of clinical settings. This study focuses on AC applied to take into account the reality of clinical settings and patients seeking treatment. It evaluates the effectiveness of group-based AC for the treatment of severe depression in a clinical setting in a heterogeneous population in terms of diagnosis (unipolar and bipolar depression) and comorbidity (Axis I and II). Methods A sample of 45 participants with severe depression was recruited in a psychiatric hospital. Participants received a 10 sessions group intervention of AC. Questionnaires were administered to obtain pretreatment, post-treatment and four-week post-treatment data. The impact of the intervention was observed on measures of depression, behavioural activation, reinforcement, anxiety, social adjustment and quality of life. Various moderation effects associated with the heterogeneity of the sample were tested on the evolution of depressive symptoms. The integrity of the treatment administered by the therapists and the acceptability of the intervention by participants were also documented. Results Mixed model analyses of variance were performed to assess whether (a) AC caused a significant change at the end of treatment on depressive symptoms, behavioural activation, reinforcement, anxiety, social adjustment and quality of life and whether (b) gains were maintained after four weeks. A significant change was obtained between the pre-post measures on the average score of all these variables, with the exception of a subscale of the quality of life measure. Analyses were also performed to verify various moderating effects on the evolution of depressive symptoms, level of activation and reinforcement. No interaction effects are observed on depression, activation and reinforcement measures. There is no significant difference according to pretreatment severity category, diagnosis (unipolar vs bipolar), presence of comorbidity (other Axis I and/or Axis II disorder) or co-morbidity of Axis II disorder. As for the activation measure in people with bipolar depression versus unipolar depression, it should be noted that the result is at the threshold of statistical significance. Conclusion The results support the effectiveness of group-based AC for the treatment of severe depression in clinical settings in a heterogeneous population, as well as for the maintenance of gains after four weeks. The effectiveness of AC was also observed across all associated psychosocial measures.