This research addresses measurement issues, derives a model of the effects of life events on mental health, and tests hypotheses regarding the role of social supports in modifying these effects over time. Analyses were based on a longitudinal study of persons (N = 2,234) sampled from a general population in Seattle, Washington, as part of Rand's Health Insurance Experiment. In addition to sociodemographic variables, measures of physical limitations, mental health, social supports, and stressful life events were fielded twice (one year apart), using previously validated measures. Data for a random half of the sample were analyzed to evaluate different approaches to defining and scoring life events and social supports and to fit a model of the main and interactive effects of these variables on mental health. The best-fitting model was then cross-validated, using the remaining half of the sample. Results support the following conclusions: (1) Social supports predict improvements in mental health over time, (2) life events and physical limitations predict a deterioration in mental health over time, (3) the negative effects of life events and physical limitations on mental health do not vary according to amount of social support, and (4) differences in measurement strategies for life events and social supports produce some variance in results, but not in conclusions about whether effects on mental health are additive or interactive.