In this paper we reflect upon and problematize the ways in which 'patient involvement' is interpreted in a substantial proportion of the research literature on involvement and shared decision making. Drawing upon an analysis of this literature we raise concerns about the 'medicalization of involvement' embedded in, and reproduced by, some dominant research lenses, suggesting that this medicalization has powerful discursive and material effects. For example, we suggest that it tends to normalize and arguably trivialize intrinsically problematic and contentious concepts such as 'patient preferences' and, at the same time, to obscure the full range of possibilities for reciprocity in the exchanges between the medical world of the professional and the experiential and narrative world of the patient. We argue that richer conceptualizations of collaboration in clinical work are both possible and very much needed, and we indicate some examples of scholarly resources and perspectives that point towards richer and more defensible accounts of involvement. Overall we call for more attention to the idea of 'epistemic involvement' and much greater cross-fertilization between different epistemological paradigms in this area of research.