There are many facets to cancer prevention: a good diet, weight control and physical activity, a healthy environment, avoidance of carcinogens such as those in tobacco smoke, and screening of populations at risk to allow early detection. But there is also the possibility of using drugs or naturally occurring compounds to prevent initiation of, or to suppress, tumour growth. Only a few such agents have been used to date in the clinic with any success, and these include non-steroidal anti-inflammatory drugs for colon, finasteride for prostate and tamoxifen or raloxifene for breast tumours. An ideal chemopreventive agent would restore normal growth control to a preneoplastic or cancerous cell population by modifying aberrant signalling pathways or inducing apoptosis (or both) in cells beyond repair. Characteristics for such an agent include selectivity for damaged or transformed cells, good bioavailability and more than one mechanism of action to foil redundancy or crosstalk in signalling pathways. As more research effort is being targeted towards this area, the distinction between chemotherapeutic and chemopreventive agents is blurring. Chemotherapeutic drugs are now being designed to target over- or under-active signalling molecules within cancer cells, a philosophy which is just as relevant in chemoprevention. Development of dietary agents is particularly attractive because of our long-standing exposure to them, their relative lack of toxicity, and encouraging indications from epidemiology. The carcinogenic process relies on the cell's ability to proliferate abnormally, evade apoptosis, induce angiogenesis and metastasise to distant sites. In vitro studies with a number of different diet-derived compounds suggest that there are molecules capable of modulating each of these aspects of tumour growth. However, on the negative side many of them have rather poor bioavailability. The challenge is to uncover their multiple mechanisms of action in order to predict their efficacy, to learn how to use them effectively in combination, and in some cases to redesign them to improve potency or bioavailability. These ideas are illustrated by dietary agents such as indole-3-carbinol (I3C), epigallocatechin gallate (EGCG), curcumin and resveratrol, all of which appear to have a number of different molecular targets, impinging on several signalling pathways. Ultimately it may be possible not only to suppress tumours and to extend quality of life by administering appropriate diet-derived molecules, but also to refine the definition of a cancer chemopreventive diet.