Wednesday, August 29, 2007

Sexual Selection and Human Phenotypic Variation

Does sexual selection explain many phenotypic differences among human populations? This hypothesis was first put forward by Charles Darwin and has been given a new twist by Henry Harpending. The following is a question raised about its falsifiability on the hbd mailing list and my reply.


I've been thinking about Prof. Henry Harpending's paper "Human Diversity and it's History" (see online copy at You probably are familiar with it's ideas already but I will recap and you can thus check whether I understand it correctly.

Essentially Harpending argues that human phenotype differences have their origin in sexual selection and that this process is inherently conservative. That's 'conservative' in contrast to 'normal' gene flow processes. So external phenotype characteristics are more likely to stay in place versus "neutral" or "unseen" (and thus not sexually selected) characteristics. Thus phenotypes are more likely to reflect the actual human migration history, than analysis of genotypes or other neutral measures, e g blood types etc.

This hypothesis, if true, means that the last generation or so of 20th century physical anthropologists (eg Hooton, Coon, Howells, Birdsell etc) who took metrical analysis of phenotype differences to the max were "barking up the right tree", and a subsequent generation who focused on measuring phenotypical neutral traits (esp blood etc) may have been "barking up the wrong tree".

I'm slowly coming to my point so forgive me for dragging it out.

As best as I can tell the Harpending hypothesis here doesn't outline any prospective falsification tests. I was wondering if you had heard of, or had any ideas that could be used to test this elegant hypothesis?

I'm interested in this as Harpending's hypothesis would seem to me to be in a position to bolster Joseph Birdsell's trihybrid hypothesis of the historical origins of the Australian aboriginals, although of course, if correct, it would have universal applicability.

Tim Gillin


There are two tests:

1. Look at human populations with intense male-male competition for mates. This kind of mate competition only partly selects for physical traits preferred by the sex in short supply (females). It also selects for physical traits that help males intimidate or fight off other males (e.g., increased body size, higher bone density, larger muscle mass, higher testosterone levels, etc.).

2. Look at human populations with intense female-female competition for mates. This kind of mate competition selects more for those physical traits preferred by the sex in short supply (males). Such traits will stimulate mate-choice algorithms or any mental algorithm that monitors the visual environment. The selected traits thus tend to be vividly colored. At high intensities of mate competition, "color polymorphisms" will develop: novel colors will have a slight edge over less novel ones, so that vividly colored phenotypes will not only proliferate but also diversify. When any one phenotype becomes too common, the selective pressure shifts to others that are rarer and more novel.

In most species, the first scenario is much more common. Sexual selection is usually about too many males competing for too few females. This is because reproduction predisposes females to invest more in their offspring, particularly during pregnancy and early infant care. During these times of life, females are unavailable for reproduction and drop out of the "mate market." Unless males can match female reproductive investment, they can best serve their reproductive interests by inseminating other females. So, at any given moment, too many males will be competing for too few females.

Unlike the situation in most species, human males have the potential to match female reproductive investment, in part because their offspring are dependent for a longer time and in part because humans have colonized temperate and arctic environments where women are less able to provide for themselves through food gathering. This 'paternal investment' is less important among the agricultural peoples of sub-Saharan Africa and New Guinea, where year-round agriculture enables women to be self-sufficient. Women even become net providers of food. The costs of polygyny thus become negative and men best serve their reproductive interests by acquiring as many wives as possible.

So we can test the Harpending hypothesis by comparing low polygyny/high paternal investment populations with high polygyny/low paternal investment populations. This kind of comparison was done when Winkler and Christiansen (1993) studied two Namibian peoples, the !Kung (hunter-gatherers and weakly polygynous) and the Kavango (agriculturalists and highly polygynous). The latter were found to have markedly higher levels of both total testosterone and DHT, as well as a much more robust physique. The authors suggested that this hormonal difference may account for the !Kung’s neotenous appearance, i.e., sparse body hair, small stature, pedomorphic morphology, and light yellowish skin.


Winkler, E-M., & Christiansen, K. (1993). Sex hormone levels and body hair growth in !Kung San and Kavango men from Namibia. American Journal of Physical Anthropology, 92, 155-164.


tim said...


Thanks for taking the time to examine the issue of testing sexual selection . It's commonly cited as a key process by authors like Richard Dawkins, for instance, for explaining human diversity but apart from interesting analogies to other species, they just leave it at that. Your comments have helped fill in a few 'missing links' for me!!

I'm interested in the Harpending paper above, especially page 8, as a means of explaining some of the cases of alleged 'parallel evolution' among diverse human populations. Harpending talks about gene flow homogenising 'neutral' (i.e. unseen) characteristics over time and sexual selection acting to preserve some key traits.

In Cavalli-Sforza's classic there is a great table illustrating the homogenising impact of genetic flow over time.(Table 1.17.1, 'The History and Geography of Human Genes'). The power of this force would seem to me to put the onus in debate on explaining why any local human differences remain at all despite human residency of most of the planet for 50-100,000 years.

Anyhow, be that as it may, Harpending discusses the cases of the Kalahari Bushmen and the neighbouring so-called 'black bushmen' in his article. He also touches upon the case of the Vedda and the South-east Asian 'negritos' in the last couple of paragraphs.

The last case is particularly interesting. There are the well known negrito people of the Andaman Islands as well as remnant 'negrito-oid' peoples in Malaysia, the Philippines and elsewhere in SE Asia.

My understanding is that apart from the defining external traits of these "negrito-oid" people they have little else in common in with each other. For their "Harpending neutral" traits they have more in common with their non-negritoid neighbours, than with other negritoids a few hundred miles away.

In the past, when physical anthropologists relied mainly on detailed metrical studies of phenotype features, partly because they had few other tools, these similarities were explained as evidence of ancient migrations and dispersal, with 'remnant' groups left as isolates.

As the means to study 'neutral' traits became more readily available, (blood groups, even crania) the pendulum of opinion among physical anthropologist swang away from these early migration based explanations. The former remnants now had their dissimilarities from their immediate neighbours explained as a product of local forces (local environmental selection, genetic drift.. take your pick.)

The problem here, or so it seems to me, is that these local explanations (genetic drift, local environment) are very different and have been made rather ad hoc. I'd say the explanations have been used more for support than illumination. They don't seem to have, for example, anything like major population genetics studies behind them. It is easy to relabel the 'too hard basket' as 'genetic drift' and declare 'case closed'.

Little thought seems to be given to those remaining and very real similarities between the variant populations. Or the astronomical odds against their similarities (...and matching sets of similarities) being the product of parallel evolution or (worse yet) genetic drift.

To make matters worse, there is a philosophic hump too. It seems to me that there is an analogy between this debate and the debate for and against 'diffusionism' in cultural anthropology. In the 19th century 'diffusionism' rode high with anthropologists positing all pyramid building civilisations had Egyptian roots ...that kind of thing. With the eclipse of European colonialism, cultural diffusionism (which certainly happens sometimes) has fallen on hard times and 'indigenous innovation', essentially parallel or convergent cultural evolution, is now the preferred explanation. This debate, however relevant it may be for discussing contemporary and near contemporary civilisations (last 3000yrs or so) has unfortunately cast a shadow over a physical anthropology debate that is rooted much deeper in time.

I think Harpending's hypothesis may help us out of this mess. If successful it would probably revive some of the observations and findings made by the last generation of old time physical anthropologists.

Exactly how to test all this is another question. To use the pygmy example, my guess is that we are still probably a long way from a full genetic mapping of the phenomenon. Presumably if the various pygmy populations around the world all had distinct local origins we would find "many paths to pygmyism" in the genome, even if there were 'functional similarities'. But if pygmyism arose only once or twice, an explanation that I think is superior in an Ockham's razor sense, my guess is that we'd find identical genetic structures at the relevant locii between otherwise diverse populations.

Am I making sense here? Am I missing something?

Peter Frost said...


Yes, you make a lot of sense. Just a few points:

1. I'm skeptical of the position that gene flow is so strong among human populations that significant inter-population differences are necessarily due to differences in selection. If so, we would not have cases of convergent evolution, i.e., two different human populations using different alleles to achieve the same effect. If our species were truly in a state of effective panmixia, any population could borrow useful alleles from any other population, anywhere in the world.

2. Many neutral markers, notably blood groups, are useless for tracing population affinities, especially if you're going back more than a dozen generations. They shift back and forth too easily, creating false affinities with unrelated populations. On the basis of blood groups, I have a greater affinity with some apes than I do with some humans. (Well, some of my critics wouldn't be surprised).

3. I believe that the different Negrito groups of south and southeast Asia share a common ancestry, but this common ancestry goes back to a very ancient demographic expansion that spread out of Africa and along the south Asian coastline some 50,000 years ago.

4. When we study human physical variation, we have to ask ourselves what we wish to study. Do we want to know the degree of relatedness or unrelatedness between different populations? Or do we want to know the degree of functional similarity or dissimilarity between different populations? The two questions are very different, yet it seems to be assumed that the more unrelated two populations are the more they will be functionally dissimilar.

Take the different Negrito groups of south and southeast Asia. They look very similar and yet they have lived apart for some 50,000 years. How come? Because they have lived in similar environments and have been exposed to similar selective pressures.

Now take Europeans and sub-Saharan Africans. They have been apart for about the same length of time, if not less. Yet they look very different. How come? Because they have been exposed to very different selective pressures.

tim said...

Thanks for the fascinating reply Peter.

I suspect all of the Oceanic negrito groups have a shared ancestry as well. It would seem to me to be too much of a coincidence. Still there do seem to be many anthropologists and others who are untroubled by what I imagine must be a coincidence of cosmic proportions.

I also suspect that the 'negritos' maintenance of pygmyism may be through a mechanism such as you describe. But my suspicions of course are not proof.

So it would seem to me that the appropriate falsification test for this idea would be an analysis of the relevant set of alleles associated with the set of characteristics that are usually listed in the physical description of negrito people, in particular, height, hair etc.

Presumably if the same set of alleles were found at the same locations for representatives subjects from the various 'negrito' populations, then there would be strong evidence of shared ancestry. If on the other hand the sets were found to be dissimilar, we'd have strong evidence of convergent or parallel evolution.

Is my reasoning on or off the mark here? Has human genetic mapping yet reached the level where this kind of detailed analysis would even be possible? My layman's understanding is that that addressing this kind of problem is still some way off.


Peter Frost said...


If the alleles have adaptive value, any similarities among those populations might be due to convergent evolution, i.e., unrelated populations becoming similar because of common selective pressures.

MtDNA is a better bet for determining population affinities. It has a high degree of selective neutrality and is less likely to produce false affinities (because more than one loci are examined).

This kind of genetic comparison has actually been done. Take a look at: