It is important that the effects of work on mental health are investigated when work practices are changing rapidly and there is decreasing job security. This has been examined in the Whitehall II Study, a cohort study of 6895 male and 3413 female, London-based civil servants, aged 35-55 years at baseline in 1985. Work characteristics were measured by modified Karasek indices in a self-report questionnaire. Psychiatric disorder was measured by the 30-item General Health Questionnaire (GHQ). In longitudinal analyses in men and women, high work social support predicted lower GHQ scores, and high job demands predicted higher GHQ scores at follow-up. High work social support and high skill discretion were protective against taking short spells of psychiatric sickness absence. The protective effects of social support at work and the potential risk of job demands have implications for management, job design, training, and further research.