Jim Flynn once observed that no-one was funding research into the genetics of racial differences in intelligence because they feared they would find something.
Here is my psychologist’s summary of where we are as regards the genetics of intelligence in general: 10%.
That is to say, by poking about in the genetic code researchers can find patterns in the genomes of the samples of discovery (n=100,000+) which, when tested on other independent samples (n=25,000+) account for almost 10% of the variance in intelligence. However, they don’t always have IQs available on the people in those large samples, so they use the weak proxy of years of education. However, the Sniekers et al. (2017) paper has real IQs and perhaps as a consequence has found novel genes associated with intelligence. Those authors say:
The strongest, positive genetic correlation was with educational attainment (rg = 0.70, s.e.m. = 0.02, P = 2.5 × 10−287). Moderate, positive genetic correlations were observed with smoking cessation, intracranial volume, head circumference in infancy, autism spectrum disorder and height. Moderate negative genetic correlations were observed with Alzheimer’s disease, depressive symptoms, having ever smoked, schizophrenia, neuroticism, waist-to-hip ratio, body mass index and waist circumference.
(The link to the paper is given below).
The various methods genetic researchers use are best explained by them and not me, but here are a few steps which I believe I understand. SNPs are personal variations in the code which make us unique. Call them exceptional quirks of coding, which would explain a great deal of why some people are so different from others, often in ways which do not seem the least bit productive to man or beast. However, if variants confer survival advantages they become more frequent in the population with each successive generation, as calculated by the breeder’s equation, and may eventually reach the entire population, in which case they are said to have achieved fixation: they are a fixed feature of the code.
In order to be sure that any feature of code is really the cause of any physical feature or behaviour, geneticists have to guard against false positives. Given that they are conducting multiple comparisons, they set their significance levels very high, compared to social psychology at least. Levels of p<.00000005 are usual. Steve Hsu argues that samples of at least n= 1,000,000 will be required to achieve stable results. Be that as it may, over the past decade the demonstrated variance in intelligence accounted for by the genetic code has risen from 0% to 1% and then by irregular steps to almost 10%. As a psychologist watching other psychologists working alongside geneticists, I doff my hat to them for their achievements. Their progress is exciting because the genetic code is causal.
Into this vast battleground of immense international armies of researchers, often several hundred to a published paper, steps the lone figure of Davide Piffer, who argues thus: if I select only those genes that are revealed as being associated with intelligence in virtually all the different published studies (call them the Perennial Reliables) then I can make up a preliminary genetic score for group intelligence. Since the genetic studies of individual intelligence have been done on Europeans, he uses those few genes to create a predictive score which is set to IQ 100, the Greenwich Mean Intelligence. The next step is so simple as to have baffled me the first time I heard Piffer present his results in May 2014. All Piffer did was to look up standard databases to see how frequent those particular intelligence genes were in non-European genetic groups. He was able to show that by this simple technique you could generate a predicted group IQ which matched the observed genetic group IQ pretty closely. For example, the Chinese had more of these genes and were brighter than Europeans, Africans had fewer and were less bright than Europeans. QED.
Although this was a fascinating result, I was still worried about false positives. If intelligence is caused by many genes of very small effect, how could so few genes create an almost perfect match with the results of international intelligence testing? Even though this related to group results, and would be unlikely to give anything other than an unreliable prediction for individuals, I feared that a simple error was lurking somewhere. (One should always fear that, even if the result seems a good fit with the data). It is for those reason that I eschewed the temptation to declare in 2014 that Piffer had solved the problem, or at least provided a substantial first step.
By 2015 Piffer had refined his results, and was getting a closer match with observed group differences. Later 2016 I started getting news that Professor Risch had made fun of Piffer’s approach, comparing it to PISS. When I got the transcript of his talk I was astounded to hear him say that when he applied Piffer’s approach to the data he got the same results. I had expected an exposition as to why the match with the group results was flawed, and instead got a replication. This made me take a further look at Piffer’s results. Here is the relevant excerpt from Prof Risch’s speech:
I then noted that Prof Risch’s supposed refutation was based on testing the group prediction on a test sample of n=2: the genomes of Craig Ventnor and James Watson. As Piffer observed “Try publishing a genetics paper with an n of 2”. However, there may have been other reasons for considering that the approach was flawed which were not given in the paper. At the moment, as an outsider to this subject, I count it as a partial replication.
Now that the Sniekers et al. (2017) paper mentioned above has identified novel intelligence genes
Piffer has done a further updating of his paper, which I link below. He finds that his original formula can now be strengthened and has stronger power to predict the observed group differences in intelligence. He explains to me that besides calculating average frequencies he also factor-analyzed the alleles (the individual variations in the code) to sort out the polygenic selection signal from the noise. The signal is that bit of the code which has been selected for in subsequent generations because it confers an advantage, in this case quickness in learning which increases the likelihood of surviving more challenging and changeable environments.
The first link is to Piffer’s preprints for the educational attainment scores.
The second link is to his blog post his updated paper
The third link is for the more technically minded a link to the RPubs).
Piffer is working to get this work published in a single paper. To be sure, he is not claiming to be able to predict any individual’s intellectual level with his technique, but to be able to predict the group averages is an achievement in its own right. Even more reassuringly to me, he has shown that a random collection of SNPs does not produce the observed group intelligence differences. This gets around my initial concern that his particular collection of SNPs could have been a fluke, just picking up some other aspects of racial difference.
Here is his composite factor score, from the “new genes same results” paper, with groups ranked by ability:
I looked up the 18 intelligence GWAS SNPs and the 9 EA quasi-replicated SNPs and could find 4 in ALFRED. Factor analysis was run on them, producing a very interesting factor. For ease of interpretation, I report results ranked from highest to lowest:
Continent Population Factor
EastAsia Tujia 1.507
East Asia Mongolian 1.358
EastAsia Daur 1.246
EastAsia Yi 1.19
EastAsia Koreans 1.127
EastAsia Miao 1.078
EastAsia Japanese 1.018
EastAsia Dai 0.987
EastAsia Hezhe 0.98
EastAsia Han 0.936
EastAsia Lahu 0.877
EastAsia Tu 0.828
EastAsia Xibe 0.802
Europe Orcadian 0.753
EastAsia She 0.737
EastAsia Uyghur 0.566
Asia Hazara 0.506
Asia Kalash 0.475
Asia Oroqen 0.445
Europe Italians_N 0.437
Europe Italians_C 0.404
SE Asia Cambodians, Khmer 0.34
Siberia Yakut 0.311
Europe Adygei 0.257
Asia Druze 0.254
Europe French 0.217
Asia Burusho 0.151
EastAsia Naxi 0.113
Europe Russians 0.073
Asia Balochi 0.055
Asia Palestinian -0.071
Europe Basque -0.088
Asia Bedouin -0.156
Europe Sardinian -0.225
Asia Brahui -0.334
Asia Pashtun -0.426
Asia Sindhi -0.438
Oceania Melanesian, Nasioi -0.533
Oceania Papuan New Guinean -0.569
Africa Mozabite -0.768
Africa Mandenka -1.153
Africa Yoruba -1.27
NorthAmerica Maya, Yucatan -1.3
NorthAmerica Pima, Mexico -1.312
SouthAmerica Amerindians -1.366
Africa Biaka -1.369
Africa Bantu Kenya -1.381
SouthAmerica Surui -1.382
Africa Mbuti -1.415
Africa Bantu SA -1.454
Africa San -1.488
SouthAmerica Karitiana -1.53
So, is the problem of the genetics of racial differences in intelligence now sorted out? No, not yet. The paper is a proof of concept. It appears that you can get a surprisingly good prediction of group differences in intelligence by the use of this simple technique. As other papers continue to find new variants which code for intelligence in European populations, these new bits of code can be added to Piffer’s predictive equation. It is for others to test it, and to knock holes in it.
Prediction: we will need very many more SNPs before we can attempt predictions of individual IQs across different races at better than a correlation of r=0.7