Soc Psychiatry Psychiatr Epidemiol (2001) 36: 582 – 589 © Steinkopff Verlag 2001 ■ Abstract Background According to the buffer hy- pothesis, social support or social network may affect mental health by buffering the negative effect of stress on mental health. Previous studies have mostly been cross-sectional or have been done on selected popula- tions, and the results have been conflicting. Methods The buffer hypothesis was tested in three population-based health surveys which took place in 1987, 1990 and 1993 in five coastal municipalities in Finnmark, Norway. All persons aged 40–62 years and a random sample of those aged 20–39 years were invited, and 77 %, 74 % and 70 % attended the three health surveys, respectively. Those who had attended the health surveys more than once and had answered the key questions about mental health and social network, social support (measured both as instrumental support as well as emotional support) and stress were included in the analyses. Stress was divided into acute stress, (somatic stress, civil stress and work stress) and chronic stress (having a chronic disease, dis- ability pension, being a single parent or providing long term nursing care for someone in the family). Growth curve analyses using mental distress as the dependent variable were used, analysing the interactional effects of various types of stressors and social network or social support. Results When all possible stressors and the sum of social network/social support were taken into consid- eration, total social support/network buffered the dete- riorating effect of total stressor score upon mental health. The effect was weak but significant, and stronger for women than men. When each stressor was analysed separately, only a significant buffer effect of social net- work could be detected for work stress. For one of the chronic stressors, receiving a disability pension, a buffer effect could be demonstrated for both social network and instrumental support. Conclusion The results pro- vide some support for the buffer hypothesis, and indica- tion of specificity in the interactions between stressor and social network/social support was found.Women,in general, had a larger buffering effect from their social network than men. ■ Key words Mental distress – Social support – Social network – Stress Introduction The interactions between mental health, social network, social support and stress are complex. Two hypotheses exist – the main or direct effect hypothesis and the buffer hypothesis (Cohen and Wills 1985). The main or direct effect hypothesis assumes that social support or social network influences mental health regardless of the level of stress. The buffer hypothesis assumes that the effect of social network and social support is to buffer or moderate the effect of stress on mental health. Empirical research that has attempted to determine which of the two hypotheses is the most correct has been hampered by the lack of agreement about how to define and measure the various concepts involved. In a previ- ous article, in which the direct effect hypothesis was tested in a prospective population sample (Olstad et al. 1999), the concepts of social support and social net- work were discussed.We found that social network and social support had very little direct predictive effect upon mental distress 3 years later. When focusing upon the buffer hypothesis, the con- cept of stress also needs to be clarified. Early research on the effect of the buffer hypothesis mainly focused upon ORIGINAL PAPER R. Olstad · H. Sexton · A. J. Søgaard The Finnmark Study. A prospective population study of the social support buffer hypothesis, specific stressors and mental distress Accepted: 5 September 2001 SPPE 503 R. Olstad Institute of Clinical Medicine University of Tromsø 9037 Tromsø, Norway R. Olstad () · H. Sexton Psychiatric Research Center for Finnmark and Troms Regional Hospital in Tromsø – Åsgård 9291-Tromsø, Norway E-Mail: reidun.olstad@rito.no A. J. Søgaard National Health Screening Service PO. Box 8155, Dep 0033 Oslo, Norway