The aim of the study was to compare the fidelity of manualised group memory rehabilitation programmes for participants with neurological disabilities. A sample of 11 neurological patients with memory problems, enrolled in a randomised controlled trial comparing compensation, restitution and self-help treatments, were observed during group sessions. Time-sampling was used to record the activity of the participants and the content of the discussion at one minute intervals. There was a significant difference between groups in the amount of time the group leader and participants spent talking (p < .001). Participants in the compensation and restitution groups spent significantly more time in memory rehabilitation discussion than participants in the self-help group (p < .001). There was also a significant difference between the amount of time spent discussing internal and external memory aids in the compensation and restitution groups (p < .001). These results support the fidelity of the interventions provided. This study also highlights the usefulness of time-sampling as a method to record the content and activity in rehabilitation groups.