Rationale, aims and objectives This paper addresses the challenge of investigating fidelity in the implementation of a complex rehabilitation intervention designed to increase the level of independence in personal activities of daily living of stroke patients living in UK care homes. A programme theory of intervention fidelity was constructed to underpin a process evaluation running alongside a cluster randomized trial of the rehabilitation intervention. Methods The programme theory has been constructed drawing on principles of realist evaluation. Using data from in-depth semi-structured interviews (n = 17) with all occupational therapists (OTs) and critical incident reports from the trial (n = 20), and drawing from frameworks for implementation, the programme theory was developed. Results The programme theory incorporates four potential mechanisms through which fidelity within the trial can be investigated. These four programme theory areas are (1) the balancing of research and professional requirements that therapists performed in a number of areas while delivering the study interventions; (2) the OTs rapport building with care home staff; (3) the work focused on re-engineering the personal environments of care home patients; and (4) the learning about the intervention within the context of the trial and its impacts over time. Conclusions These findings characterize the real-world nature of fidelity within intervention research, and specifically the negotiated nature of implementation within clinical settings, including individual patients' needs. This research adds to the evidence base because current frameworks for fidelity neglect the importance of learning over time of individuals and across the time span of a trial.