Based on the premise that acute and chronic stresses stimulate and suppress cortisol secretion, respectively, and the hypothesis that marriage provides a buffer to stress, we tested whether extreme values of serum cortisol concentrations would be less likely in married women than in unmarried women. Three hundred women were recruited from two central Connecticut communities. Cortisol was measured in overnight urine samples using liquid chromatography-tandem mass spectrometry. Information on each subject's demographic characteristics, such as income and education level was collected. Mean log urinary cortisol was virtually identical in married and unmarried women, however, as predicted, the variance was significantly larger in the unmarried group (p = 0.01). After adjustment for potential confounders, multivariate logistic regression still revealed that absolute deviation of log(10) cortisol from the mean was smaller for married versus unmarried women (p < 0.01); deviation from the mean cortisol was also higher for non-working than working women. These results support the idea that marriage and employment reduce the extreme levels of cortisol secretion, and by extension, this may reflect differences in levels of stress in married and in working women compared to unmarried and non-working women.