Update to Konigsberg and Ousley's "Multivariate Quantitative Genetics of Anthropometric Traits from the Boas Data" (1995)

Our 1995 paper was entirely based on anthropometric data from American Indians that had been collected under Franz Boas's direction for the 1893 World Columbian Exposition in Chicago, Illinois. The original data sheets had been rediscovered by Richard Jantz and had been carefully entered into a computer database, which served as the basis for a number of projects. Our 1995 publication was but one such project. Given the 102-year lapse between the Columbian Exposition and our 1995 paper, we might not expect that the passage of 15 more years would have brought much change to the interpretations of our basic results. We would be correct in this particular expectation but wildly incorrect in the supposition that the particular question we were examining has lain dormant. In the 1995 paper we asked, What is the relationship between phenotypic and additive genetic variance-covariance matrices? Our results supported an argument that the matrices were proportional, such that P^-1G was a diagonal matrix with a single common trait heritability on the diagonal. This finding has implications for biological distance analysis, because the proportionality then also extends to Mahalanobis distances. Specifically, the genetic Mahalanobis distance matrix is then equal to the phenotypic distance matrix divided by the common trait heritability. The proportionality also applies to F_ST such that the phenotypically measured F_ST divided by the heritability is equal to the true F_ST. And finally, when P and G are proportional, multivariate phenotypic and genetic allometry coefficients are equal.

Our examination of proportionality between the P and G matrices was largely based on Cheverud's (1988) paper in which he suggested that "the genetic-variance/ covariance matrix may be roughly estimated by 0.35P" (Cheverud 1988: 965). Cheverud based this argument on an analysis of phenotypic and genetic correlations among traits within 41 animal data sets. It is interesting to note that in the same year that our publication appeared, Roff (1995) published an article demonstrating close correspondence between genetic and phenotypic correlations in the sand cricket. The following year Roff (1996) published a broader study that showed comparability of genetic and phenotypic correlations across 51 plant and animal taxa. Waitt and Levin (1998) showed across a sample of 27 plant species that genetic and phenotypic correlations were significantly associated. Although more recent studies have focused primarily on the evolution of the G matrix (Begin et al. 2003; Steppan et al. 2002), they have continued to provide support for proportionality of the P and G matrices. One notable exception to these confirmations is the study by Ousley (1997), who found that G and P are not proportional for dermatoglyphic finger ridge counts, quasicontinuous traits that are established by the sixth fetal month. Although there does seem to be continuing support for many of the results from our 1995 paper, we would be remiss were we not to point out that the last sentence of our paper made a completely erroneous statement. We wrote that "the Boas data present an unequaled chance to examine population dynamics among individuals for whom standard marker locus or DNA data are unavailable." Although it is true that the Boas data continue to be unequaled, it is not true that DNA data are unavailable from this sample. Soon after the publication of our article, Richard Jantz found that hair clippings from many of the individuals in the Boas sample had been preserved at the American Museum of Natural History. Baker (2001) later performed extractions and analyzed mtDNA variation in two groups (Choctaw and Meominee) not included in our 1995 analysis.