Cassel (1976), Cobb (1976), and Caplan (1974) all emphasize that much or
most of the beneficial health effects of social relationships are due to their
buffering properties in the presence of stress, hence the label social support,
and some (Kaplan et al 1977:49) even assert that "social supports are likely to
be protective [of health] only in the presence of stressful circumstances."
Thus, much of the subsequent literature focused on the issue of buffering vs
main effects. Literature reviews by the first author and others (e.g. House
1981, House & Kahn 1985, Cohen & Wills 1985) reveal that, while neither
type of support effect is found uniformly, few if any studies fail to find
evidence for either main or buffering effects of social relationships on health. I
Thus, it now seems that the more appropriate research question is not whether
both effects exist, but when, how, and why each occurs.
At this point, we would hypothesize that buffering effects are most often
found when there is a strong stressor to which people with varying degrees of
involvement in social relationships are exposed, or when the measure of
social relationships is phrased in terms of buffering (e.g. "How much are
people helpful in times of stress?"). Reviews of nonexperimental studies by
Cohen & Wills (1985) and Kessler & McLeod (1985) find evidence of
buffering effects primarily in studies using measures of perceived availability
of support in times of need or stress. Measures of the existence of social
relationships more often produce only main effects, although being married
considered by itself produces buffering effects in three of four studies reviewed
by Cohen & Wills ( l985:Table 1). Retrospective measures of the
actual provision of support in the face of stress show mixed relationships with
health and well-being, probably because they are confounded with the severity
of stress. Longitudinal studies, with appropriate controls for stress levels,
should find clearer evidence of beneficial effects of supportive behavior in the