Evidence receives elevated attention in teacher education today, and this attention fuels the urge to identify the impact of evidence for teaching. However, this identification turns out to be difficult: Evidence is only infrequently used by teachers to explain or justify action, and this empirically documented situation is argued to be due to a missing connection between evidence and contingent decision-making in teaching, and a lack of recognition of characteristics of teaching in research that produces evidence. The argument points to a schism between the knowledge researchers construct and the knowledge that teachers use for teaching. To address this schism, educational research requires a robust conception of knowledge, and a recognition of differences between knowledge-that and knowledge-how is argued to contribute to such robustness. The paper introduces a Foucault perspective that recognises this difference, and illustrates the relationship of knowledge-that and knowledge-how based on Foucault's concept of care of the self. Care of the self is described from an educational perspective to highlight that teachers develop professional knowledge on teaching through reflective knowledge-how, and to describe evidence as a resource in this development when it enables the problematisation of personal knowledge.