Objective: To develop a reliable sperm test that would predict pregnancy rate in assisted reproductive technologies. Design: Blind prospective cohort study. Setting: Tertiary-care, university hospital-affiliated IVF program. Patients: One hundred nineteen sperm samples were obtained from 110 males from couples undergoing IVF or GIFT (ART). Sperm samples were washed by Percoll, incubated at 24 degrees C for 4 hours, and an aliquot of the same sperm suspension was used for ART incubated at 40 degrees C for 4 hours (stress test). Stress test scores are expressed as the ratio of final to initial motility. Results: Of 119 ART cycles, 24 resulted in pregnancy. Of 24 pregnancies, 23 occurred in cycles that used sperm samples with str ess test scores greater than or equal to 0.75 and only one with a stress test score <0.75. The negative predictive value of the test, defined as the absence of pregnancy with scores <0.75, was 98% and the positive predictive value, defined as the occurrence of pregnancy with scores greater than or equal to 0.75, was 36%. Logistic regression analysis indicated that the stress test score alone was correlated significantly with pregnancy after ART. Conclusion: These results indicate that str ess test scores <0.75 are predictive of poor pregnancy outcome in ART.