Although participation in many sporting activities has increased dramatically in recent years, the study of injuries sustained during training or participation is still in its infancy. The most commonly used strategy is to describe the characteristics of a suitable case-series. This approach is relatively easy to implement, can be used to estimate the total morbidity load in a population, and can identify the relative frequency of various types of injury. However, the case series method cannot validly identify risk factors for injury or athletes at high risk; similarly, it cannot be used to estimate the absolute level of risk associated with sports participation. Finally, the population from which the injuries arose is often difficult to identify, and the series may not be representative of all injuries occurring in that population, and this may produce quite misleading results. In contrast, a variety of epidemiological designs may be employed to address questions of aetiology and to identify high risk groups of athletes. With careful attention to the underlying population denominators, one may estimate the relative or absolute risk of injury for athletes with given risk characteristics, defined by type and intensity of their participation in sports or by their individual physiology. This is achieved by inclusion of suitable control subjects in the epidemiological sample; these controls may be uninjured athletes or random samples of the general population. The comparison of injured and uninjured groups permits valid inferences to be drawn concerning risk factors, avoiding the many potential biases which affect inferences drawn from injured athletes only.