Evidence relating stress with illness has emerged from a variety of experimental, clinical and epidemiological research strategies. Psychobiological stress responses emerge through an imbalance between demands and psychosocial resources, and many factors relevant to this transaction have been identified, including the chronicity and predictability of stimulation, opportunities for control, psychological coping responses and the availability of social supports. By contrast, the mechanisms through which stress responses may increase risk of illness are poorly understood, and investigators are frequently obliged to fall back upon a poorly defined 'biological predisposition' to account for individual differences in susceptibility to disease or variations in clinical course. It is argued in this review that several distinct cognitive-behavioural and psychophysiological mediating processes may be postulated, and that these are relevant to different types of influence on health (acute and chronic effects, causal and facilitatory processes, impact on aetiology, maintenance, relapse or recurrence). It is through the recognition and identification of separate processes that questions concerning the specificity of illness outcome, personal resilience, the limits of the impact of psychosocial factors, and the appropriate forms of therapeutic intervention, may be resolved.