The relation between psychological stress at work and menstrual function was examined for 276 healthy, working, premenopausal women who participated in the California Women's Reproductive Health Study in 1990-1991. Subjects collected daily urine samples and completed a daily diary for an average of five menstrual cycles. Metabolites of estrogen and progesterone were measured in the urine, and computer algorithms were developed to characterize each cycle as ovulatory or anovulatory and to select a probable day of ovulation. A telephone interview collected information about psychological stress at work as well as other occupational, demographic, lifestyle, and environmental factors. Logistic regression was used to model stressful work and risk of anovulation (> or = 36 days without ovulating) and measures of within-woman cycle variability. Repeated measures analyses were performed on other menstrual cycle parameters. Stressful work (high demand in combination with low control) was not strongly related to an increased risk for anovulation or cycle variability or to any of the following cycle endpoints: short luteal phase (< or = 10 days), long follicular phase (> or = 24 days), long menses (> or = 8 days), or long cycle (> or = 36 days). However, women in stressful jobs had a more than doubled risk for short cycle length (< or = 24 days) compared with women not working in stressful jobs (adjusted odds ratio = 2.24, 95% confidence interval 1.09-4.59).