Background. While high levels of social support (SS) are associated with a decreased risk for major depression (MD) or less depressive symptomatology, and stressful life events (SLEs) have a substantial causal relationship with MD, uncertainty remains as to whether a main-effect or a buffering model best explains the nature of the relationship among SS, MD and SLEs.

Method. Using two waves of interview data on 2163 female twin pairs from a population-based twin registry, and discrete time survival analysis with both logistic and linear regression models, we examine the ability of interactions between eight dimensions of SS and 16 categories of stressful life events to predict MD onset and levels of depressive symptomatology.

Results. In the presence of a significant effect of a SLE on MD (β [ges ] 1·00), we found evidence for seven interactions out of a possible 93, of which none involved buffering effects. Similarly, examination of depressive symptomatology detected a total of two interactions (both buffering) out of possible 28. We found no evidence, beyond what would be expected by chance, for the existence of buffering effects where either MD or depressive symptomatology was used as the dependent variable.

Conclusions. There is little evidence to suggest the presence of the buffering effect of social support in the face of adverse life events for women. We suggest that it is important to use alternative models (multiplicative and additive) to examine data, to investigate the match between stressors and social resources, and to investigate fully whether detected interactions actually represent a buffering effect.