This article reviews consistent research findings concerning the assessment of clinical competence during the clerkship phase of the undergraduate medical training programme on issues of reliability, validity, effect on training programme and learning behaviour, acceptability and costs. Subsequently, research findings on the clinical clerkship as a learning environment are discussed demonstrating that the clinical attachment provides a rather unstructured educational framework. Five fundamental questions (why, what, when, how, who) are addressed to generate general suggestions for improving assessment on the basis of the evidence on assessment and clinical training. Good assessment requires a thoughtful compromise between what is achievable and what is ideal. It is argued that educational effects are eminently important in this compromise, particularly in the unstructured clinical setting. Maximizing educational effects can be achieved in combination with improvements of other measurement qualities of the assessment. Two concrete examples are provided to illustrate the recommended assessment strategies.