This paper raises questions about the epistemological foundations of evidence-based medicine (EBM). We argue that EBM is based upon reliabilist epistemological assumptions, and that this is appropriate - we should focus on identifying the most reliable processes for generating and collecting medical knowledge. However, we note that this should not be reduced to narrow questions about which research methodologies are the best for gathering evidence. Reliable processes for generating medical evidence might lie outside of formal research methods. We also question the notion of the knower that is assumed by EBM. We argue that EBM assumes an enlightenment conception of knowers as autonomous, substitutable individuals. This conception is troubled by the way that clinicians learn the role of anecdote in health care and the role of patient choice, all of which bring into play features of clinicians and patients as situated individuals with particular backgrounds and experiences. EBM's enlightenment conception of the knower is also troubled by aspects of the way evidence is produced. Given these limitations, we argue that EBM should retain its reliabilist bent, but should look beyond formal research methodologies in identifying processes that yield reliable evidence for clinical practice. We suggest looking to feminist epistemology, with its focus on the standpoints of individual situated knowers, and the role of social context in determining what counts as knowledge.