Prisoners of Abstraction? The Theory and Measure of Genetic Variation, and the Very Concept of “Race”

It is illegitimate to read any ontology about “race” off of biological theory or data. Indeed, the technical meaning of “genetic variation” is fluid, and there is no single theoretical agreed-upon criterion for defining and distinguishing populations given a particular set of genetic variation data. By analyzing three formal senses of “genetic variation,” viz., diversity, differentiation, and heterozygosity, we argue that the use of biological theory for making claims about race inevitably amounts to a pernicious reification. Biological theory does not force the concept of “race” upon us; our social discourse, social ontology, and social expectations do. We become prisoners of our abstractions at our own hands, and at our own expense.

1. This position may seem initially implausible (consider the facts about the relationship between socially identified race and, e.g., health outcomes, noted below), but Mills notes that a certain kind of Marxist, committed to the view that class is explanatory of all important differences in life prospects, might deny the existence of race as a category with any real impact, collapsing the life-prospect issues that follow race into those that follow class (Mills 1988, p. 49).

2. Lewontin (1972) does not explicitly deny the existence of population structure in humans—indeed, we are unaware of any biologist who has done so. But his statistical results and criticism of the “race” concept as socially destructive and as (nearly) biologically useless are often interpreted this way, and it is this interpretation that Edwards seems to have in mind. See Kaplan (2011) for discussion.

3. Contemporary humans vary, on average, by about 1:1,000 nucleotides; this is around an order of magnitude less variation than occurs in many other species (see, e.g., Li and Sadler 1991; Cognato 2007). All the measures of genetic variation discussed below rely on this small amount of variation when applied to humans. This variation is not distributed equally in humans; people whose ancestors are of recent African origin, for example, differ on average by about 1:900 nucleotides; people whose ancestors were of recent European origin differ by only about 1:1,600 nucleotides. Edwards is correct when he notes that “It is not true, as Nature claimed, that ‘two random individuals from any one group are almost as different as any two random individuals from the entire world’” (2003, p. 801) but not perhaps in the way he intended; interestingly, the “average” person of recent African descent is more likely to share an arbitrary allele with an “average” person of recent European descent (will differ less) than he or she is with another arbitrarily chosen person of African descent (African v. European difference is about 1:1050, compared with 1:900 difference within Africa) (see Yu et al. 2002).

4. Even in the professional literature, there is substantial confusion surrounding these concepts and associated measures. Consider the following, from a “Correction” published by Kronholm et al. (2010b): “In the introduction of Kronholm et al. (2010a), we discuss what properties a differentiation measure, like F_ST, should or was assumed to have. Recent developments have shown that F_ST in fact does not have these properties.” The authors cite and thank Jost, who we also believe has helped identify and publicize some of the conceptual difficulties and problematic assumptions associated with these measures.

5. Ecological genetics has been more explicit about these distinctions (n.b., Lou Jost is an ecological geneticist), perhaps because it has to be clear about whether token numbers or token frequencies, within or across types, for genes or for species, are being measured. For example, a textbook in the field dedicates a whole chapter to “Genetic diversity and differentiation” (Lowe et al. 2004, Ch. 3). Some of the distinctions drawn are broadly similar to those we prefer, e.g., they note that “Genetic diversity measures estimate the amount of variation that is found in a population, while genetic differentiation measures describe how this variation is partitioned among populations.” However, we suggest a narrower definition of differentiation than the one Lowe et al. actually employ, and believe it is important to more carefully distinguish measures of heterozygosity from other measures of variation.

6. This point is made forcefully by Jost (2008).

7. This kind of example is used by Jost (2008) and others to show why measures like G _ST should not be interpreted as measures of diversity. The equations for calculating the Shannon Entropy measure and G _ST are below, and it is a worthwhile exercise to compute the heterozygosity partitions for a variety of circumstances in order to get a sense of how the measures in fact behave. Note that for loci with relatively low overall diversity—which were the kind that Lewontin’s 1972 analysis in fact focused on—measures like the Shannon Entropy measure and G _ST do not usually produce such misleading results, and so this should not be taken to imply that the estimates of within- and between-population diversity (in the sense we describe below) in humans are necessarily very different than that implied by Lewontin’s analysis. Nevertheless, the analysis used is inappropriate if one is trying to measure diversity.

8. While Jost’s D has several features that make it a convenient example of a diversity measure, our use of it as an example should not be taken as an endorsement of it as the only or the best measure and associated partition. For discussion, see e.g., Hoffmann and Hofmann 2008, Gabriele et al. 2010, Whitlock 2011, and cites therein.

9. Nei’s standard distance, for example, assumes that populations diverge because of genetic mutations and drift, whereas Reynold’s distance assumes only drift and purposefully excludes mutation. See Libiger et al. (2009).

10. The SMEO-P model—i.e., “set-up, mathematically manipulate, explain, objectify—pluralize”—developed in Winther (2006) is useful here.

11. These debates surrounding the ontology of socially identified races take place against a background where basic life prospects (income, life expectancy, morbidity, standards of living, and quality of life more generally) are strongly correlated with socially identified race. The question: “What accounts for these differences?” lurks behind these debates, but the implications of that question are not always obvious, nor is it obvious what the answers to that question the various positions in the debate suggest. But at least some of the dangers in biological explorations of human population structure that, e.g., Kitcher (2007) points towards, and that Kaplan (2010) highlights, are wrapped up in these issues.

12. We ignore here the complexities of developmental biology and non-genetic heritable resources.

13. In various forms, this position, or one very much like it, has been defended by Appiah and Gutmann (1996), Hacking (2005), Haslanger (2010), Omi and Winant (1989), and others.

Matthew Kopec, Michael J. Wade, and the editors of this special issue of Biological Theory, Kim Sterelny, Massimo Pigliucci, and Werner Callebaut, kindly provided feedback on this paper. Kaplan would like to thank the participants at the KLI “The Meaning of ‘Theory’ in Biology” workshop for their comments, and the Oregon State University Center for the Humanities, whose support helped make this work possible. Winther thanks Ian Hacking, Carlos López-Beltrán, and Amir Najmi for ongoing discussions on these topics. Kim Sneppen’s support of Winther’s guest research stay at the Biocomplexity Center, Niels Bohr Institute, University of Copenhagen, is also gratefully acknowledged. This paper, written fully jointly, with authors listed alphabetically, is the result of an ongoing collaboration that emerged from an e-mail exchange between the authors that began with Kaplan reading Winther (2011).

Authors and Affiliations

1. Philosophy Department, Oregon State University, Corvallis, OR, USA

   Jonathan Michael Kaplan

2. Department of Philosophy, University of California, Santa Cruz, CA, USA

   Rasmus Grønfeldt Winther

Corresponding author

Correspondence to Jonathan Michael Kaplan.

Reprints and permissions