A new approach for testing hypotheses about modern human origins using molecular divergence dates is presented. Coalescence times from many unlinked loci are needed to test the alternative models. Hypotheses are evaluated on the basis of their differing predicted distribution patterns of coalescence times from multiple genes. No single coalescence time from one genetic system is sufficient to reject any of the three alternative models. Several nuclear datasets give recent dates for human genetic ancestors, at approximately the mitochondrial coalescence time, while some nuclear datasets support older dates. Given the overall distribution of available mitochondrial and nuclear coalescence times, the rapid replacement hypothesis is the likeliest model for modern human origins. The unusual nature of the human mitochondrial pattern is highlighted by comparative data from nonhuman hominoids. To understand the pattern of modern human genetic variability better, more nuclear data from all hominoid species are needed.