The present study was designed to test the relation between stressful life events experienced during pregnancy and the risk of preterm delivery and shortened duration of pregnancy. We collected data prospectively in a general population sample, including repeated questionnaire measures of exposure to stressful life events during pregnancy. Between August 1989 and September 1991, 8,719 Danish-speaking women with singleton pregnancies attended antenatal care. Of these women, 5,873 (67%) completed all questionnaires. When indicating an event, the woman was asked to rate the amount of stress induced by this event. Measurement of gestational duration was primarily based on early ultrasound scan. When we evaluated life events independently of the individual's appraisal, we found no association with duration of gestation or risk of preterm delivery. In contrast, life events assessed by the subject as highly stressful were associated with shorter mean duration of gestation and increased risk of preterm delivery. This association was observed primarily with events experienced between the 16th and 30th week of gestation. Women who had one or more highly stressful life events had a risk of preterm delivery 1.76 times greater than those without stressful events (95% confidence interval = 1.15-2.71). We found no evidence for a buffering effect of social support.