- Commentary on Derish and Sokal's "A Classification of European Populations Based on Gene Frequencies and Cranial Measurements: A Map-Quadrat Approach" (1988):Evidence for the Lack of Taxonomic Structure but the Congruence between Genetic and Morphometric Information
Due to an error at the printer, the original article page range (530–532) was incorrect. The correct pagination is 531–533 and is reflected in this updated article. Click here for the corrected PDF.
Published in 1988, the paper by Derish and Sokal on the "classification on European populations" represented a methodological turning point in the assessment of the possible relationships (co-variation extent) among genotype (gene frequencies), phenotype (morphometrics), culture (language), and geography within a relatively limited and coherent eco-spatial unit. Notably, this exploratory study examined the variation patterns expressed by 59 gene frequencies (sampling 22 genetic loci) and 10 selected cranial measurements in population samples from 97 European localities encompassing (at that time) 29 countries, the investigated area having been arbitrarily structured into 85 unitary quadrats.
By simply reading the first introductory sentences, the main critical question was: How can it work? This is the question mostly because the tools commonly used at the time for combined data analysis— the NT-SYS library of numerical taxonomy programs set by Rohlf (1985)—were in their "infancy" stage (the electronic version for PC was made available only later; see Rohlf 1990). In addition, previous work from the same Sokal, as well as from other researchers (e.g., Guglielmino-Matessi et al. 1979), had pointed out the complex nature and not yet fully understood background of morphometric (notably, craniometric) variation, which represented an additional complicating factor in such an investigation. Finally, the quadratting procedure appeared to give little hope for a consistent picture of spatial correlation between gene frequencies and morphometric record—how could the use of a grid with cells of 5 × 5 degrees (each roughly covering 240,000 km2), where both gene frequencies and craniometric data were only available for a subset of cells and sample sizes varied considerably, have been a reliable background for a successful analytical work? The nature itself of the investigated topic, dealing with the "classificatory schemes for organizing human variation"—that is, indirectly with the concept of [End Page 531] "race" as perceived at the time and discussed during the two preceding decades—was too sensitive to risk a flop.
Rereading this seminal paper twenty-four years later proves how our initial concerns were unjustified. Derish and Sokal elegantly showed that the quadratting procedure, even if introducing some possible artefacts (p. 822), was much more effective than the usual ethnic classification at a continental scale of the European populations traditionally used in similar analyses. Besides other interesting results, by itself, this is rather important evidence which, while based on a quite simple data set, unambiguously illustrates the feeble consistency and the risks of an oversimplistic and often typologically-based thinking deeply rooted in our cultural background when dealing with human population biohistories.
As expected, of course Derish and Sokal failed to reveal a distinct "taxonomic structure" (hierarchical sense) within and among the quadrats, but rather showed a continuum in variation resulting from the complex dynamics having affected the bio-cultural history of the European populations since the early Holocene, notably throughout the last three millennia. As a whole, the noise introduced by the pseudo-random sampling was indeed less strong than the noise inherent to the classical group classification in which culture overcame biology and masked/exalted biological differences. Finally, the polarity congruence is noteworthy between the genetic and morphometric information first noted by the authors among most of the quadrats.
A few years after their work, the geometric morphometrics "colored" books made their appearance, thus providing new and more robust statistical tools for size and shape analysis (Bookstein 1991; Marcus et al. 1993, 1996). Nonetheless, the firm methodological background of multivariate statistics supporting Derish and Sokal's work remained (and still partially remains) a highly valuable reference approach. Mostly, the authors showed how to proceed in data mining and how to build a solid interpretative framework by exploring bio-cultural scenarios integrating at a continental scale...