Ethnographic research was conducted in the thoracic ward of a Norwegian university hospital in order to study collaborative medical problem solving. As a general principle, evidence-based medicine is supposed to lead the process of medical problem solving. However, medical problem solving also requires evidence of a different kind. This is the more concrete form of evidence, such as X rays and other representations, that guides medical practice and makes sure that decisions are grounded in sound empirical facts and knowledge. In medicine, 'evidence' is on the one hand an abstract category; on the other hand, it is a tool that is practically enacted during the problem-solving work. Medical evidence does not 'show itself'. As such it has an emergent quality. Medical evidence has to be established and made practically useful in the collaborative settings by the participants in order to make conclusions about diagnoses and treatment. Hence, evidence is an interactional product; it is discursively generated and its applicability requires discourse. In addition, the production of medical evidence requires more than medical discourse and professional considerations. This paper looks at the production processes and use of medical evidence and the ambiguous meaning of this term in practical medicine.