OBJECTIVE To evaluate the test properties of the postcoital test (PCT). METHODS Retrospective analysis of prospectively collected data on a cohort of infertile couples with complete follow up. METHODS Fertility clinic of a Dutch university hospital. METHODS A continuous series of 224 couples of whom 24 were excluded for reasons of anovulation, coital problems, proven sterility or incomplete follow-up. METHODS Cumulative pregnancy rates in relation to PCT results with and without treatment for 'cervical factor' infertility. RESULTS The predictive values of normal and abnormal PCTs were 0.54 and 0.58 overall and 0.74 and 0.47 if only untreated women were considered. Sensitivity and specificity were, respectively, 0.47 and 0.65 for all women and 0.54 and 0.68 for untreated women only. Likelihood ratios for normal and abnormal PCTs were 0.83 and 1.32 overall and 0.67 and 1.72 in untreated women. CONCLUSIONS The PCT has poor predictive power. This and the psychological impact on subfertile couples attest to the need for more rigorous study designs in evaluating this test.