BACKGROUND: Changing the physical environment is one way to promote physical activity and improve health, but evidence on intervention effectiveness is mixed. The theoretical perspectives and conceptual issues discussed or used in evaluative studies and related literature may contribute to these inconsistencies. We aimed to advance the intervention research agenda by systematically searching for and synthesising the literature pertaining to these wider conceptual issues. METHODS: We searched for editorials, commentaries, reviews, or primary qualitative or quantitative studies in multiple disciplines by electronic searches of key databases (MEDLINE and MEDLINE In-Process, Web of Science, Cochrane Reviews, ProQuest for dissertations, Health Evidence, EPPI-Centre, TRID and NICE) and snowballing. We extracted theoretical and conceptual material and used thematic analysis in an in-depth, configurative narrative approach to synthesis. RESULTS: Our initial searches identified 2760 potential sources from fields including public health, sociology, behavioural science and transport, of which 104 were included. By first separating out and then drawing together this material, we produced a synthesis that identified five high-level conceptual themes: one concerning outcomes (physical activity as a behaviour and a socially embedded practice), one concerning exposures (environmental interventions as structural changes) and three concerning how interventions bring about their effects (the importance of social and physical context; (un) observable mechanisms linking interventions and changes in physical activity; and interventions as events in complex systems). These themes are inter-related but have rarely been considered together in the disparate literatures. Drawing on these insights, we present a more generalisable way of thinking about how environmental interventions work which could be used in future evaluation studies. CONCLUSIONS: Environmental and policy interventions are socially embedded and operate within a system. Evaluators should acknowledge this, and the philosophical perspective taken in their evaluation. Across disciplinary fields, future studies should seek to understand how interventions work through considering these systems, the context in which interventions take place, and the (un) observable mechanisms that may operate. This will help ensure that findings can be more easily interpreted and widely applied by policymakers. We hope that highlighting these conceptual issues will help others to interpret and improve upon a somewhat contested evidence base.