Introduction. Cognitive-behavioural therapy (CBT) seems to offer effective treatment for pathological gambling (PG). However, it has not yet been established which techniques yield the best results, or whether exposure and response prevention (ERP) techniques are of additional use. Objectives. To evaluate clinical and socio-demographic characteristics of a PG sample at baseline, comparing cognitive-behavioural group intervention, with and without exposure, with response prevention (CBT + ERP vs. CBT), to compare the results of therapy and to assess pre-post changes in psychopathology between both groups. Design. We applied a quasi-experimental design comprising intervention on the independent variable, but without random assignment. Methods. The sample comprised 502 males with PG, consecutively admitted to a specialist unit, who received standardized outpatient CBT group therapy in 16 weekly sessions. Scores on the Symptom Checklist-Revised (SCL-90-R), the Temperament and Character Inventory-Revised (TCI-R), the South Oaks Gambling Screen (SOGS), and other clinical and psychopathological scales were recorded. Results. Pre-post changes did not differ between groups, except for SCL paranoid ideation, being greater in the CBT therapy group. The risk of relapse during treatment was similar in the CBT + ERP and CBT patients. However, compliance with treatment was poorer in the CBT + ERP group, who presented higher drop-out rates during treatment. Drop-out during therapy was associated with shorter disorder duration and higher scores on the TCI-R novelty seeking scale. Conclusions. Although the two CBT programs elicited similar therapy responses, patients receiving CBT alone showed higher adherence to therapy and lower drop-out rates. © 2011 The British Psychological Society.