A formalin-inactivated Rift Valley fever vaccine prepared in primary monkey kidney cells has been used to protect laboratory workers from disease since 1967. A similar but improved vaccine was prepared in 1978-1979 using well characterized diploid fetal rhesus lung cells. In initial clinical trials reported here, the new vaccine elicited high levels of plaque neutralizing antibodies and caused only minimal local reactions at the injection site. Significant variability was observed in the geometric mean titre evoked by various vaccine lots. This variability had not been predicted by conventional pre-filtration or pre-inactivation virus infectivity assays, or the results of animal potency tests. These findings emphasize the need for statistically valid human potency testing and the development of accurate predictive preclinical measurements for this and other vaccines.