For the traditional Chinese medicine (TCM), a prescription for a patient often contains several herbs. Some herbs are often used together in prescriptions, and these herbs can be considered as a functional group. In this paper, we propose an approach for discovering herbal functional groups from a large set of prescriptions recorded in TCM books. These functional groups are allowed to overlap with each other. Our approach is validated with a simulation study and applied to a data set containing thousands of TCM prescriptions. Copyright © 2011 John Wiley & Sons, Ltd.