"Not so fast, says one anthropologist"

Hawks sightings in the news.

I've been in the midst of a grant proposal -- yes, I actually do write those from time to time! Yes, you can support the site by giving my grant proposals glowing reviews...

Anyway, there hasn't been much time for me to follow up on that "Skull study deals death blow to multiregional evolution" story that's been going around this week. But I've written a few notes:

Charles Roseman makes the essential point in Michael Balter's article about the study:

Charles Roseman, an anthropologist at the University of Illinois, Urbana-Champaign, says that he is not convinced that the Nature authors have adequately tested the Out of Africa model versus its multiregional rival. The researchers assumed that the multiregional model requires that modern humans arose more than once. "Proponents of the multiregional model have been very clear for some time that their models do not posit multiple origins, as suggested in the paper," Roseman says.

The thing that irritates me is that the phrase "multiple origins" is right there in the abstract: (emphasis added)

The origin and patterns of dispersal of anatomically modern humans are the focus of considerable debate [1, 2, 3]. Global genetic analyses have argued for one single origin, placed somewhere in Africa [4, 5, 6, 7]. This scenario implies a rapid expansion, with a series of bottlenecks of small amplitude, which would have led to the observed smooth loss of genetic diversity with increasing distance from Africa. Analyses of cranial data, on the other hand, have given mixed results [8, 9, 10, 11, 12], and have been argued to support multiple origins of modern humans [2, 9, 12]. Using a large data set of skull measurements and an analytical framework equivalent to that used for genetic data, we show that the loss in genetic diversity has been mirrored by a loss in phenotypic variability. We find evidence for an African origin, placed somewhere in the central/southern part of the continent, which harbours the highest intra-population diversity in phenotypic measurements. We failed to find evidence for a second origin, and we confirm these results on a large genetic data set. Distance from Africa accounts for an average 19-25% of heritable variation in craniometric measurements - a remarkably strong effect for phenotypic measurements known to be under selection.

Really, they could have gotten this right with very little effort. It would have taken away the headline-grabbing part; but I can hardly believe that this kind of thing is headline-grabbing any more. I mean, Hanihara's craniometric database is awesome, sure, but they haven't done anything with the data that Bill Howells didn't do 35 years ago. This is a simple regression of variance against distance, with a second multiple regression including "climate," meaning a variable consisting of mean annual temperature and precipitation. It is a lot less sophisticated than the analyses that Roseman (2004) did, or Katerina Harvati and Tim Weaver (2006).

OK, so here are some of my notes about the paper (Slashdot wanted to know...):

1. There was a theory of "multiple origins" of modern humans. It was called polygenism, it was rejected by Darwin and was utterly discredited more than 50 years ago. One of the cited papers above for the idea of "multiple origins" [ref. 9] is one of my papers, titled "Multiregional, not multiple origins." Now, I know that some people might write a paper with the words, "not multiple origins", in the title, and secretly, in their heart of hearts, have meant to send the message that "multiple origins" is the way to go. But that wasn't what we had in mind! How could we have been any clearer?

Now, you might think, well this is just a semantic quibble. So they shouldn't have written "multiple origins" -- just ignore that. They did test multiple evolution, didn't they?

No. They conclude that a single origin is the best explanation, because adding a second origin somewhere else doesn't improve the explanatory power of their model. This "second" origin is assumed to be from a long-isolated population -- otherwise, they wouldn't say this:

[W]e cannot distinguish between single and multiple exoduses from Africa, because both scenarios would lead to a major cline from Africa (347).

Of course, "multiple exoduses" is precisely the prediction of multiregional evolution.

2. The main result of the study is that there is a trend toward lower within-population phenotypic variance as populations are farther from Africa. This trend does provide a good match to the cline of reducing genetic variance outside of Africa. However, the genetic cline in diversity itself is utterly uninformative about the geographic origin of recent humans. The diversity cline may just as easily be explained by population size (larger long-term within Africa), migration (biased in one direction), or selection. These are not obscure facts, I can't really believe that anyone conversant in modern human origins and genetics isn't aware of them.

Well, it's a short paper -- maybe that's why these problems slipped past Nature's reviewers...

3. A lot of people know about human genetic variation, and they are aware that a series of sequential founder effects can yield a cline of reducing genetic diversity. But here's the problem: what works for allele frequencies ain't necessarily true for quantitative features.

The phenotypic variance of craniometric characters is somewhere between 25 and 75 percent environmental. So from the outset, there is a large component of variance that isn't explained by allele frequencies. More problematic, although phenotypic variance is the sum of genetic and environmental variances, reducing the genetic variance does not tend to reduce the phenotypic variance by an equal amount. This is because a reduction in variation in the genetic background tends to increase a third component of the variance, the genotype-environment interaction variance.

Consider a field of hybrid corn. All the plants are genetically uniform, and this elimination of genetic variance tends to vastly decrease the phenotypic variance. At least, as long as the environment is also uniform. But introduce variation in the environment -- low spots in the field with standing water, patches toward the edge with greater pest damage, and so on -- and the variance in the genetically uniform field may be disproportionately great compared to the effects of the same factors in a genetically diverse field.

All this is to say that if you want to use phenotypic variance to measure genetic variance, then you have to ensure that the environmental variance is equal. Of course, that's a problem for human crania, since people manifestly live in different environments, arguably with greater variability (at least with respect to diet and climate) in Africa.

4. So, variability in these samples declines further from Africa. What does that mean?

Well, if I wanted to answer that question, then I would look first at exactly which measurements show the trend.

As a whole, except for three measures that actually increase in variability with distance from Africa, the rest show a quantitatively slight trend toward lower variation out of Africa. For most it is not significant (with R2 < 0.03), but for a few it reaches statistical significance even with a slight negative correlation (R2 < 10%). Since all of the comparisons include New World populations, I would want to see the data scatter for these characters to see if most of the slight negative correlation of variability with distance was explained by these far-flung populations that probably really did experience a strong bottleneck within the last 15,000 years.

That leaves five measurements with strong negative correlations between distance from Africa and variability: Basion-prosthion length, upper facial breadth (fmt-fmt), nasion-prosthion height, nasal height, and zygomaxillary subtense (a measure of facial projection).

It seems to me that the recent evolution of the face is mostly a consequence of selection on the dentition, which has evolved rapidly in the last 15,000 years throughout the world. That has little, if anything, to do with modern human origins.

But it has a lot to do with my grant application.


Manica A, Amos W, Balloux F, Hanihara T. 2007. The effect of ancient population bottlenecks on human phenotypic variation. Nature 448:346-348. doi:10.1038/nature05951