The author argues that exercise may be a unique health behavior, leaving researchers with problems in understanding exercise compliance. New definitions, new methods, and broader perspectives are needed. Four theories of compliance are described, and problems of research methodology are discussed. (Author/MT)