Background. Few studies have examined prospectively both the direct and buffering effects of types of social support and social networks on mental health. This paper reports longitudinal associations between types of social support and psychiatric morbidity from the Whitehall II study.

Methods. Social support was measured by the Close Persons Questionnaire and psychiatric morbidity by the General Health Questionnaire at baseline (1985–1988) and at first follow-up (1989) in 7697 male and female London-based civil servants aged 35–55 years at baseline. The cohort was followed up and baseline measures were used to predict psychiatric disorder measured by the General Health Questionnaire at second follow-up (1991–1993).

Results. Longitudinal analyses showed that low confiding/emotional support in men and high negative aspects of close relationships in men and women were associated with greater risk of psychiatric morbidity even after adjustment for baseline General Health Questionnaire score. There was no evidence of a buffering effect among men or women who experienced life events or chronic stressors. Controlling for a personality measure of hostility did not affect the observed relations.

Conclusions. The present findings illustrate that different types of support are risk factors for psychological distress and that they operate in different ways for men and women. Direct effects of emotional support are predictive of good mental health in men and negative aspects of close relations predict poor mental health in both men and women. Emotional support is predictive of good mental health in women whereas, confiding alone is not.