We recruited 221 women with no known fertility problems who were discontinuing contraception to conceive a pregnancy. The primary objective of this prospective study was to estimate the incidence of very early pregnancy loss by using a highly sensitive and specific assay to detect the pregnancy hormone chorionic gonadotropin in first morning urine specimens. We found the risk of early pregnancy loss, defined as loss within 6 weeks of the last menstrual period, to vary by season of conception, with a large amplitude and some consistency across the 3 years of the study. Such a seasonal pattern may reflect the contribution of an environmental factor that varies with season. The peaks in risk ranged from early September to early December. A seasonal pattern of early pregnancy loss should contribute to a corresponding lagged seasonal pattern in livebirths. Accordingly, we looked for corroborative evidence in regional birth data from the same years. There was some correspondence, but this was largely limited to the first year of the study.