Forensic DNA Tests and Hardy-Weinberg Equilibrium

all ofwhom have in common only that they identify themselves culturally as black. The term "black" has more cultural than biological meaning. The same is true for the term "Hispanic." Since the methods of Devlin et al. do not detect deviations from H-W for groups as internally diverse as blacks and Hispanics, one wonders whether the same methods would detect deviations even if all U.S. citizens were combined into a single group that is known to be heterogeneous at these loci (8). 6) Devlin et al. give no analysis of their metds' statistical power to detect deviations from H-W. They should have performed a simulation study of artificially generated data, with sample sizes corresponding to the actual sample sizes, with varying amounts of deviation from H-W. Applying their methods to these data sets would show how small a deviation from H-W can be detected. 7) For the Caucasian data set, Devlin et al. observe an increase in the ratio of the observed to the expected number of heterozygotes as a function of the difference X between fragment pair lengths (in kilobases) for three different probes (D17S79, D14S13, and D2S44) and approximate the pattern of increase by a logistic curve (their figure 3). They interpret the observed patterns as a result only of coalescence. While coalescence contributes to the patterns observed, it would not seem to explain why the patterns are so different for the three loci. Their estimated thresholds for coalescence range from 0.099 kb for the D17S79 locus to 0.434 kb for the D2S44 locus. If the patterns were a simple function of the physical properties of DNA separation on a gel and of human visual discrimination, the thresholds for coalescence should be independent of the probed locus. Factors other than coalescence appear to be important; population-level processes may be among them. For example, pairs of fragments of similar length (small I) could have more recent common ancestry and be more geographically concentrated than pairs of fragments of greatly differing length (large a). 8) The complete set of data available to Devlin et al. could have been used in a simple way to detect excessive or deficient homozygosity relative to H-W. According to the points plotted in their figure 3, as Or increases, the observed heterozygosity comes to match and often to exceed the expected heterozygosity under H-W. If only coalescence were responsible, then for large T, the observed heterozygosity should randomly fluctuate above and below the expected heterozygosity; if other factors besides coalescence, such as population substructure, were at work, the observed heterozy-