Friday 23 May 2014

Microcephalin makes a comeback


As far back as 2006 Bruce Lahn noted that microcephalin was linked to lower intelligence. His University then counselled him that he would be better off working on other matters. Danielle Posthuma took up the challenge and, in a triumph of empiricism, failed to replicate the result. That’s right. Failure to replicate is a success for knowledge. I smiled when I heard her present the results in Amsterdam in 2007, because if such research had been frowned upon world wide, rather than just in the US, then we would have imagined that the microcephalin effect was true, but suppressed by political pressure. Odd, how jumpy people are when they don’t want to know that intelligence differences are due to the genes (but fearfully believe it anyway), whereas us poor souls think such hypotheses depend on the evidence.

Now we have the sequel, Microcephalin II.  Michael Woodley (passim) will be well known to readers of this blog and Heiner Rindermann; Stratford, Ed Bell and Davide Piffer will be joining him in the halls of fame. They have investigated the pattern of population level correlates of MCPH1 and ASPM frequency counts and IQ. They find evidence for a substantial mediation effect stemming from health-status variables, which hints at a possible new role for MCPH1 in particular as an immune system boosting gene.

Woodley, M. A., Rindermann, H., Bell, E., Stratford, J. & Piffer, D.
(2014). The relationship between Microcephalin, ASPM and intelligence: A
reconsideration. Intelligence, 44, 51-63.


Despite the fact that the recently evolved Microcephalin and the related
Abnormal Spindle-like Microcaphaly Associated (ASPM) alleles do not
appear to be associated with IQ at the individual differences level, the
frequencies of Microcephalin have been found to correlate strongly with
IQ at the cross-country level. In this study, the association between
these two alleles and intelligence is examined using a sample of 59
populations. A bivariate correlation between
Microcephalin and population average IQ of r = .790 (p ≤ .01) was found,
and a multiple regression analysis in which the Human Development Index,
Disability Adjusted Life Years (DALY) lost due to Infectious diseases,
DALY Nutritional deficiencies, and Würm glaciation
temperature means were included revealed that Microcephalin remained a
good predictor of IQ. Path analysis, with both direct and indirect paths
from Microcephalin to intelligence, showed good model fit. These
multivariate analyses revealed strong and robust associations between
DALYs and Microcephalin, indicating that the former partially mediates
the association between the latter and IQ. A second smaller
correlational analysis involving ten country-level estimates of the
frequencies of these two alleles collected from the 1000 genomes
database replicated this pattern of results. To account for the findings
of this study, we review evidence that these alleles are expressed in
the immune system. Microcephalin is strongly associated with DNA repair,
which indicates a special role for this allele in the intrinsic
anti-viral immune response. Enhanced immune functioning may have
advantaged both hunter–gatherer and agrarian societies coping with the
heightened disease burden that resulted from population growth and
exposure to zoonotic diseases, making it more likely that such growth
and concomitant increases in intelligence could occur.

Get the whole paper here:


  1. Fair enough. This is sound indirect evidence for an evolutionary link between a gene and population level intelligence. However, you seem to be forgetful of the DIRECT evidence provided by Piffer in his 2013 Mankind Quarterly paper, followed by two papers published on Open Behavioral Genetics ( which found clustering of 12 IQ increasing alleles across populations and strong correlation (0.9) with country IQ.

    1. Piffer's sixth author on that paper, Einstein.

  2. Piffer not forgotten, just not mentioned in that post, but in this one

  3. That's just an abstract copied and pasted from LCI conference proceedings, with no input from you.

  4. Peter Frost commented on that paper, and I happen to agree:

    "I have mixed feelings about this study. Looking at the world distribution of this allele (see above map), I can see right away a much higher prevalence in Eurasia and the Americas than in sub-Saharan Africa. That kind of geographic distribution would inevitably correlate with IQ. And it would also correlate with the prevalence of infectious diseases.

    Unfortunately, such correlations can be spurious. There are all kinds of differences between sub-Saharan Africa and the rest of the world. One could show, for instance, that per capita consumption of yams correlates inversely with IQ. But yams don't make you stupid.

    More seriously, one could attribute the geographic range of this allele to a founder effect that occurred when modern humans began to spread out of Africa to other continents. In that case, it could be junk DNA with no adaptive value at all. There is of course a bit of a margin between its estimated time of origin (circa 37,000 BP) and the Out of Africa event (circa 50,000 BP), but that difference could be put down to errors in estimating either date."

    Not that either he or I think there's nothing to this, but this could be well, correlational, with the cause of the association operating quite a few links away.

    1. I think that the problem of false positive certainly bedevils any search for individual genes related to intelligence, and to group differences in intelligence.

    2. Michael A. Woodley30 May 2014 at 16:03

      To say that the study is somehow deficient because it is 'correlational' is just plain statistically illiterate (this happens a lot unfortunately). Sure, correlation does not equal causation, but find me just one single instance of a causal relationship where there is no correlation (just one would suffice). At any rate, the study isn’t even correlational; it is explicitly mediational. Apparent associations between two variables can be mediated by other variables, which in turn cause the association (that includes yams and IQ by the way – try latitude, or temperature). The pattern of causal inference discussed in the paper is based on the direction of the effect of certain mediating variables (in this case infectious disease burden), as determined via multivariate analyses (most explicitly in the case of the Structural Equations Models on page 57). The robustness of this mediation is furthermore demonstrated via the replication of the same pattern of effects in two entirely different datasets. Granted, there is variance in the model which has not been accounted for, however given that the multivariate regressions pick up between 83 and 88% of the total variance in the criterion variables (hint: the relevant coefficients are listed next to R^2 in the tables) without falling afoul of serious multicollinearity, I am hard pressed to think what might have been missed by way of major mediators. If people are convinced by Just-Not-So stories of geography, founder effects, hunter-gatherers or other factors in generating the observed pattern of mediation then I invite them to put these to the test and prove me wrong (all data are appended to the publication making this a potentially very simple task for anyone with a half-way decent stats-package). Kutner et al. (2005) is an excellent place to begin learning about applied linear modelling incidentally.


      Kutner, M., Nachtsheim, C., Neter, J., & Li, W. (2005). Applied linear statistical models (5th ed.). Irwin, California: McGraw-Hill.

  5. Right, but when the same pattern occurs for 14 IQ increasing or 56 height increasing alleles (as shown by Piffer), arguing for random drift becomes really silly.