This research explored domestic abuse perpetrator group dynamics with especial attention to the vital role ofthe therapeutic alliance and the constructions men have of their own narratives; it provides important lessons. for future efficacy research with its attention to group process rather than a focus on measurable outcomes. Three Programmes participated and differed from each other on characteristics such as duration and approach; one was mainly therapeutic with additional cognitive-behavioural aspects (Programme A), one was mainly cognitive-behavioural with additional therapeutic aspects (Programme B), and one was the Home Office accredited Integrated Domestic Abuse Programme (IDAP) (cognitive-behavioural only) (Programme C). The data set was comprised of interviews with men from each programme (completers and non-completers, n=20), 50 hours of reviewed sessions from each Programme, and a Staff Day. The method of analysis for the interview data was based on the psychoanalytic defended subject, Free Association Narrative Analysis. RESULTS Key findings from this research are: that there were therapeutic alliances in the group between the facilitators and the participants and in one instance between participants; that even a strong alliance needed sufficient time to be of greater benefit to the participant; and that negative and neutral alliances generally appeared unsupportive of treatment goals. Facilitation and structural factors facilitating a robust alliance were: attention to feelings, leavings and loss; consistently high quality facilitation; boundaried compassion; gender equality; group alliance; appropriate facilitator sharing; organisational clarity; motivational interviewing; individual session work; longer length and content flexibility. Factors found to undermine the alliance were: inconsistent facilitation; disrespectful communication; inattention to feelings, loss and leavings; pedagogy; shaming; content rigidity and lack of organisational clarity. The term 'boundaried compassion' was developed to bridge the necessity to label some behaviours as unacceptable whilst being compassionate to the person. Additional findings: the high quality facilitation in the context ofa brief intervention (15 weeks for Programme B) was experienced as genuinely traumatic for most of their respondents. Combining CBT and other therapeutical approaches appears complimentary. When men were committed to the process of change, they fell into two categories: putting a 'leash' on (controlling) their anger, and 'breaking down the mask' of their defences (changing fundamental attitudes and beliefs, including those regarding masculinity).