Medical intervention commonly addresses issues which can rarely be controlled and measured experimentally. The zeal for standardization of the intervention may lead to premature closure of a development process, in which the intervention method and its implementation are omitted from systematic evaluation. However, evaluation strategies other than the experimental ones do exist and can overcome these limitations. Action research is presented as an alternative approach for evaluation of medical interventions. This strategy denotes the study of a social situation, intended to improve the quality of action. The two central concerns--improvements in practice and increased knowledge and understanding--are linked together in an integrated and dynamic cycle of activities, in which each phase learns from the previous one and in turn shapes the next. The stages of research involve problem identification, planning, action and evaluation. A model for medical action research, intended to facilitate evaluation of every step, is presented and discussed, illustrated by an example from primary health care research. Evaluation must account for the following elements of the process: problem identification, summarizing previous experience, determining the aims of intervention, planning and development of the intervention method, design and articulation of the intervention strategy, implementation of action, and redefining the problem. Potentials and limitations of the action research strategy are discussed.