This paper presents a critical overview of current concepts and analytic practices in stress research and considers how they can be changed to make the research more consistent with core sociological interests. An overarching concern of the paper is the analytic use of basic information about people's social and institutional affiliations and statuses. It is important that such information be treated not simply as data that need to be controlled statistically; we must examine the bearing of these data on each domain of the stress process: the exposure to and meaning of stressors, access to stress mediators, and the psychological, physical, and behavioral manifestations of stress. The conceptualization and measurement of stressors should move away from their focus on particular events or chronic strains and should seek instead to observe and assess over time constellations of stressors made up of both events and strains. Moreover, the effects of the mediators--coping and social support--are evaluated most fruitfully in terms of their effects in limiting the number, severity, and diffusion of stressors in these constellations. Finally, sociological stress researchers should not be bound to outcomes that better serve the intellectual interests of those who work with biomedical and epidemiological models of stress, nor should the research be committed exclusively to a single outcome.