Physical activity is associated with a reduced risk of all-cause and colonic cancers, and it seems to exert a weaker effect on the risk of breast, lung and reproductive tract tumours. This review examines possible mechanisms behind the observed associations. Restriction of physical activity by pre-existing disease may contribute to the association with lung cancers, but seems a less likely explanation for other types of tumour. Indirect associations through activity-related differences in body build or susceptibility to trauma seem of minor importance. Potential dietary influences include overall energy balance and energy expenditure, the intake and/or bioavailability of minerals, antioxidant vitamins and fibre, and the relative proportions of protein and fat ingested. Links between regular exercise and other facets of lifestyle that influence cancer risks are not very strong, although endurance athletes are not usually smokers, and regular leisure activity is associated with a high socioeconomic status which tends to reduce exposure to airborne carcinogens, both at work and at home. Overall susceptibility to cancer shows a 'U'-shaped relationship to body mass index (mass/height2) reflecting, in part, the adverse influences of cigarette smoking and a tall body build for those with low body mass indices and, in part, the adverse effect of obesity at the opposite end of the body mass index distribution. Obesity seems a major component in the exercise-cancer relationship, with a particular influence on reproductive tract tumours; it alters the pathways of estradiol metabolism, decreases estradiol binding and facilitates the synthesis of estrogens. Among the hormonal influences on cancer risk, insulin-like growth factors promote tumour development and exercise-mediated increases in cortisol and prostaglandin levels may depress cellular components of immune function. However, the most important change is probably the suppression of the gonadotropic axis. Apparent gender differences in the benefits associated with regular exercise reflect gender differences in the hormonal milieu and also a failure to adapt activity questionnaires to traditional patterns of physical activity in females. The immune system is active at various stages of tumour initiation, growth and metastasis. However, acute and chronic changes in immune response induced by moderate exercise are rather small, and their practical importance remains debatable. At present, the oncologist is confronted by a plethora of interesting hypotheses, and further research is needed to decide which are of practical importance.