Sarich Got It Wrong

Updated: May 30th 2012

            Before I start criticizing a piece of work of the U.C. Berkeley physical anthropologist, Vince Sarich, I want to make it clear that I owe a significant part of my career to Professor Sarich.  As an undergraduate at Berkeley I had Sarich as part of a team teaching my introduction to physical anthropology course and more importantly, Sarich taught the upper division course in human variation that I took and which helped to make me want to study genetics and physiology and modern human populations.  I must have learned a lot—I earned an A+—and someone must have thought I knew something as I was one of the undergraduates asked to write a letter in support of Sarich’s promotion and tenure in the anthropology department at U.C.  I was able to write a very positive letter, given my basic lack of understanding of higher education at that point in my academic career, because Sarich was an interesting, knowledgeable, and passionate teacher.  It wasn’t until I was approaching my own tenure and promotion that I began to appreciate how tough it must have been for Sarich to persist in his pioneering molecular anthropology work in spite of criticism from almost all of the major players in the discipline at that time.  So when I decided to add his book (Sarich and Miele, 2005) on race to my required reading list in my class on Race, Ethnicity, and Human Variation as an example of bad racial science, it was not an easy decision.  Many of the aspects of this book have been critiqued in various book reviews, but some of the most egregious errors have gone unnoted and I want to take issue with one of these in this commentary.
            Chapter 7 of Sarich and Miele is supposed to be the pièce de résistance, cutting to the core those who deny the biological reality of race.  One of the hardest issues for the racialists like Sarich is the calculation of the apportionment of human variation by individual, population, and race, going back to the work of Richard Lewontin in 1972.  He was the first to make such a calculation using data on 17 blood groups from a number of populations around the globe.  He calculated that 85.4% of the variability was attributable to inter-individual differences within populations, 8.3% was attributable to differences between populations within races, and 6.3% was attributable to continental racial groups[i].  Interestingly, Sarich glosses this finding as 15% between races, 85% within races in making this one of the key nostrums that he hopes to demolish in Chapter 7.  In the body of the chapter, Sarich refers to the 15% as between populations, which is also incorrect, but closer to Lewontin’s figures.  Technically, if race were not being considered, only individuals and populations, we would apportion the variability 85.4% within populations, 14.6% between.  But that leads to the first problem with Sarich’s discussion of this apportionment of genetic variation having to do with his definition (or lack thereof) of race.
          Sarich begins his discussion by acknowledging the truth of this calculation and the fact that many studies have confirmed a similar within-between ratio in the now nearly 40 years since Lewontin’s original study.  Many of the more recent pieces are analyzing DNA sequence information in a variety of populations, using sophisticated computer routines to calculate the variance statistics.  In order to drive this point home, when I assign chapter 7, I also have the students read the article by Brown and Armelagos (2001) reviewing these calculations and discussing their ramifications.  Most racialists continue to cite the 85-15 ratio (cf., Edwards, 2003) but most studies, like Lewontin’s original study, find races to account for 5 – 10% of the variability, and populations to account for 5 – 10% of the variability.  In order to claim 15% as the between race proportion of genetic variation, one has to define a race as a population, and that is not what most people mean by race—the common definition, and that used by virtually all scientists, is a relatively small number of geographic groups, frequently corresponding to continental divisions (something like what Lewontin used).  However, when Sarich and Miele finally get around to defining race, not until page 207 in the chapter on race and behavior, they assert:

There is a substantial amount of agreement in the field on a working definition of the term "race" (when the author bothers to define the term, of course), as in our discussion of race and law earlier. Races are populations, or groups of populations [emphasis added], within a species, that are separated geographically from other such populations or groups of populations and distinguishable from them on the basis of heritable features (Sarich and Miele, 2005:207).

          This definition seems reasonable, especially to the lay public not familiar with human population biology.  First, let’s do away with the qualifying phrase, “or groups of populations,” since to maintain the 85-15 number, race must exactly equal population, not groups of populations, which would reduce the proportion of variation attributable to races.  Secondly, the deceptive term in their definition is population.  Most lay readers think of population as a national level phenomenon or perhaps a continental phenomenon, but in this biological context, population can only mean breeding population or Mendelian population—the unit of study of population biologists or population geneticists.  This is a much smaller unit than the public conception of population.  Until the last several hundred years, breeding groups of humans have consisted of at most a few hundred individuals.  With urban concentrations and high mobility, that figure has increased in the developed world to several thousands of individuals that each person could statistically define as a potential mate.  If race equals population, we are talking about several millions of races today—not a definition that anyone is really comfortable with, although the number doesn’t appear to bother Sarich and Miele, since in their subsequent arguments, they want to be able to create a race wherever they find any kind of measureable difference.  In any event, I think as soon as you point out that equating race with population creates millions of races, most reasonable people would say that is not what they mean by race.  But remember, if we are to accept the 15-85 ratio, we must accept race as population or reduce the 15 and increase the 85.
          Sarich continues his discussion of the ratio with an anecdote about a discussion where Henry Harpending apparently raised an issue about intra-individual genetic variation.  Sarich goes on to assert that mothers and fathers are not especially closely related and therefore, since we have 23 pairs of chromosomes that differ in their parentage, half of the within population variation should be due to the variability within individuals, meaning that the within population variation is only 42.5% and when compared to 15% between populations, that is over one-third of the variance between populations (remember populations are not races).  There are two separate problems with this logic.  First the mathematical issues.  Since Y chromosomes and mtDNA do not come in double copies, they would reduce by some theoretically quantifiable amount the ½ of the within population variation that can be attributed to intra-individual variation.  There is also the problem that the intra-individual variation is only close to ½ of the within population variability as long as the parents of the population are unrelated.  If my parents were completely unrelated, and their parents were completely unrelated, and their parents parents were completely unrelated, going on back into history, by the time we got back a thousand years into history, I would have over 1 trillion unrelated ancestors, and Sarich would have over 1 trillion unrelated ancestors, and the reader would have over 1 trillion unrelated ancestors.  The problem with this is that those 1 trillion ancestors are substantially more individuals than have ever lived since the beginning of our species.  Spencer Wells explains this phenomenon in his anecdote about the Albatross where he explains why we can trace ancestry with genetic markers in his book The Journey of Man (Wells, 2003:45), while denying the biological validity of race.  While the math of geometric increase is correct, the fallacy is that our ancestors are unrelated.  Over time, in small, relatively isolated populations, our ancestors have come to be substantially related to one another, and to share a substantial amount of genetic information.  What this means is that instead of intra-individual variation accounting for ½ of the within population variation, it would account for substantially less.  I would guess less than 10% in most cases, but it would have to be calculated for each population based on the history of the group.  The net result is that Sarich’s assumption that intra-individual variation would account for half of the 85% within population variation is significantly inflated, so the 15% to 42.5% within to between population comparison that Sarich wants to make is vastly incorrect.
          Bind1109101048031.jpgThe second category of logical flaw in the reasoning is in apportionment of variation.  This is a hierarchical partitioning of the variation, not like the use of two independent effects in a statistical analysis of variance.  In a hierarchical partitioning like this, each lower level is fully contained by the variation explained by the succeeding level.  That is, intra-individual variation is a fraction of the inter-individual (within population) variation, which is a fraction of the within race variation.  So instead of reducing the within population variation by half (or much less than half as suggested above), the within population variability still accounts for about 85% of the variation.  This is just the way variation is partitioned when dealing with hierarchical systems like this.  If we were considering variation in skin color and we had individuals and households and communities that we were comparing, the individual variation would be subsumed in the households and the between household variation would be a part of the community level variation.  So no matter what Sarich calculates as the reduction in population variation based on intra-individual differences, it would not reduce the within population variation, it merely gives us a more fine-grained way of looking at it.
          Sarich also tells of an e-mail from Henry Harpending telling him that no one had ever published the recalculation of genetic variation that Sarich proposes.  I suspect that Henry said that because with his math and statistical sophistication he would be painfully aware of all of the flaws that I mention above and he would realize that such a nonsensical solution could never be published in a journal article where it would be subjected to peer-review criticism.  It could only be published in a book where it hides (4 pages out of 262) and where the editors likely sent the book for review to sympathetic reviewers.  As an aside, I would point out that to my knowledge this naïve calculation has not appeared anywhere else, even in the racialist journals that support the kind of race based assertions Sarich and Miele are making in their book (and where the book was reviewed very favorably).

References Cited:
Sarich V, Miele F. 2005. Race: The Reality of Human Differences. Boulder, CO:Westview Press.
Lewontin R. 1972. The apportionment of human diversity. Evolutionary Biology,6:381–398.
Wells S. 2003. The Journey of Man: A Genetic Odyssey.  Princeton:Princeton University Press.
Brown RA, Armelagos GJ. 2001. Apportionment of Racial Diversity: A Review. Evolutionary Anthropology, 10:34-40.
Edwards AWF. 2003. Human genetic diversity: Lewontin’s fallacy. BioEssays 25:798–801.

[i] Lewontin used a six race system that does not really correspond to any other classification: Caucasians, Black Africans, Mongoloids, South Asian Aborigines, Amerinds, and Oceanians.

Submit item           Edit item