The findings of the elder abuse surveys conducted in the United States are not strictly comparable, although they may be combined to suggest a more organised approach to elder abuse research. This paper presents a model of elder abuse derived from those studies, and suggests that abuse is the predictable outcome of the interactions between and among elders' and abusers' characteristics. A list of the research hypotheses suggested by the model is presented, two forms of analysis that could be used to test them are described, and the implications of this model for scholars and practitioners are outlined.