Background: Conversation therapy for aphasia is a complex intervention comprising multiple components and targeting multiple outcomes. UK Medical Research Council (MRC) guidelines published in 2008 recommend that in addition to measuring the outcomes of complex interventions, evaluation should seek to clarify how such outcomes are produced, including identifying the hypothesized mechanisms of change. Aims: To identify mechanisms of change within a conversation therapy for people with aphasia and their partners. Using qualitative methods, the study draws on behaviour change theory to understand how and why participants make changes in conversation during and after therapy. Methods and Procedures: Data were derived from 16 participants (eight people with aphasia; eight conversation partners) who were recruited to the Better Conversations with Aphasia research project and took part in an eight session conversation therapy programme. The dataset consists of in-therapy discussions and post-therapy interviews, which are analysed using Framework Analysis. Outcomes and Results: Seven mechanisms of conversational behaviour change are identified and linked to theory. These show how therapy can activate changes to speakers' skills and motivation for using specific behaviours, and to the conversational opportunities available for strategy use. Conclusions and Implications: These clinically relevant findings offer guidance about the processes involved in producing behavioural change via conversation therapy. A distinction is made between the process involved in motivating change and that involved in embedding change. Differences are also noted between the process engaged in reducing unhelpful behaviour and that supporting new uses of compensatory strategies. Findings are expected to have benefits for those seeking to replicate therapy's core processes both in clinical practice and in future research.