Legitimate genotype frequency estimation for multiallelic loci relies on component allele frequencies, as population surveys represent only a fraction of possible DNA profiles. Multilocus genotypes from two ethnic human populations, African American (n=195) and U.S. Caucasian (n=200), were compiled at 13 STR loci that are used worldwide in forensic investigation (D3S1358, vWA, FGA, D16S539, TH01, TPOX, CSF1PO, D8S1179, D21S11, D18S51, D5S818, D13S317, and D7S820). Sex-specific AmpFlSTR multiplexes provided stringent PCR-based STR typing specifically optimized for multicolor fluorescence detection. Heterozygosity at each STR locus ranged from 0.57 to 0.89 and encompassed from seven (TH01) to twenty-one (D21S11) alleles. Homozygosity tests, tests based on the distinct numbers of observed homozygous and heterozygous classes, log likelihood ratio tests, and exact tests assessed that the degree of divergence from theoretical Hardy-Weinberg proportions for all 13 STRs does not have practical consequence in genotype frequency estimation. Departures from linkage equilibrium, between loci, that imposed significance to forensic calculations were not indicated by observed variance of the number of heterozygous loci or Karlin interclass correlation tests. For forensic casework, reliable multilocus profile estimates may be obtained from the product of component genotype frequencies, each calculated through application of the Hardy-Weinberg equation to population database allele frequency estimates reported here. The average probability that two randomly selected, unrelated individuals possess an identical thirteen-locus DNA profile was one in 1.8x10(15) African Americans and one in 3.8x10(14) U.S. Caucasians.