In this paper the example of falls in the elderly is used to demonstrate methodologic difficulties that arise in the epidemiologic study of frequent and recurrent health problems. Issues discussed include whether the relevant outcome is the state of being a faller or the rate at which falls occur, misclassification of self-reported outcome data, the inadequacies of current terminology for describing certain study types and measures of frequency and effect from studies of recurrent events, the potential for outcome to influence exposure status in cohort studies of recurrent health problems, and the question of controlling for falls occurring prior to the study period. It is concluded that epidemiology needs to develop a framework for studying frequent and recurrent health problems.