"Academic drift" is a term sometimes used to describe the process whereby knowledge which is intended to be useful gradually loses close ties to practice while becoming more tightly integrated with one or other body of scientific knowledge. Drift in this sense has been a common phenomenon in agriculture, engineering, medicine and management sciences in several countries in the 19th and 20th centuries. Understanding drift is obviously important, both to practitioners concerned that higher education should be relevant to practice, but also to historians who seek to make sense of long-term trends in knowledge-production. It is surprising, therefore, that although the existence of drift has been widely documented, remarkably little attention has been given so far to explaining it. In this paper I argue that drift is not an invariant universal tendency but a historically specific one which arises under particular circumstances. I outline a model of institutional dynamics which seeks to explain why drift has occurred at some institutions but not others. In the second section I explore the implications of the model for educationists and policy-makers concerned with the reform of higher education in these areas.