The Unz Review - Mobile
A Collection of Interesting, Important, and Controversial Perspectives Largely Excluded from the American Mainstream Media
Email This Page to Someone

 Remember My Information

 TeasersJames Thompson Blogview

Bookmark Toggle AllToCAdd to LibraryRemove from Library • BShow CommentNext New CommentNext New Reply
🔊 Listen RSS

young gifted black

In the whole world you know
There are billion boys and girls
Who are young, gifted and black,
And that’s a fact!

Though she is not a psychometrician, I was reminded of Nina Simone’s song by a recent paper on identifying gifted children, which found that an IQ test was better than the standard teacher referral systems at detecting bright black and hispanic kids. Good news I thought, and yet another vindication of intelligence testing.

However, before talking about that study, I want to heave half a brick at the concept of “giftedness” as regards cognitive ability. My irritation is that the concept is applied in a way which suggests that there is another avenue to being bright than that of being bright. “Giftedness” researchers tend to cluster in a separate reservation, arguing that they can detect this gift in a way particular to them, and that it is different from being intelligent. Doubtful.

This study also reminded me of my first months testing the intelligence of children referred to a child psychiatry clinic, and coming across a young black boy with an IQ of 140. On starting my job I had decided, slightly against the usual custom at that time, to give all children and their parents a summary and explanation of their test results. I suggested to his mother that the boy’s father should ring up, and when he did, very happy and excited, we planned extra reading and events for him. My records of that period must be long lost, but it would be interesting to know what became of him.
Now to the study.

Universal screening increases the representation of low-income and minority students in gifted education. David Carda and Laura Giuliano. PNAS vol. 113 no. 48 David Card, 13678–13683, doi: 10.1073/pnas.1605043113

The authors state the problem thus:

In 2012, 7.6% of White K−12 students participated in gifted and talented programs nationwide, compared with only 3.6% of Blacks, 4.6% of Hispanics, and 1.8% of English learners.

In Florida, an IQ of 130 is required for gifted status. That immediately suggests that the above figures are too high, since only 2.2% would be expected among white students and 0.13% among black students. Odd. Perhaps they are merely “promising” rather than actually “gifted” students. It turns out there is a Plan B stream comprised of English language learners and financially poor students who are allowed in at IQ 116. That laxer benchmark would allow in 14.3% of white students and 1.9% of black students.

In one test district a group non-verbal intelligence test was used to select bright children, and then face to face testing was conducted to confirm “gifted” status. Cut-offs were 130 and 115.

As to the results, the authors say:

Our analysis yields three main conclusions. First, the introduction of the screening program led to a large increase in the fraction of students classified as gifted. Second, the newly identified gifted students were disproportionately poor, Black, and Hispanic, and less likely to have parents whose primary language was English. They were also concentrated at schools with high shares of poor and minority students and low numbers of gifted students before the program. Thus, the experiences of the District confirm that a universal screening program can significantly broaden the diversity of students in gifted programs. Third, the distribution of IQ scores for the newly identified students was similar to the distribution for those identified under the old system, particularly among students who qualified under the Plan B eligibility standard. The newly identified group included many students with IQs well above the minimum eligibility threshold, implying that even high-ability students from disadvantaged groups were being overlooked under the traditional referral system.

Note that whereas before there were two entry routes there are now 4, or possibly even 6, each with an error term. That is because at each stage there will be errors of commission and omission. Some kids will be unjustly turned down, and possibly some accepted through lucky guesses. The technique of using intelligence tests will identify bright children, which is precisely what intelligence test were designed to achieve. So, IQ testing has produced an excellent result, with a better discriminative power than the referral system, and far fairer. Although it is nothing strictly to do with the selection method, the twin cut-offs are a messy complication. Half were judged by 130 and half by 115 cut-offs. So, they want the gifted, and the half-gifted, on the assumption that the latter would be gifted but for adventitious disadvantages. Let us look at the results.

This is the test used. It is non-verbal, which is a big advantage when discussing racial and cultural differences, because it is assumed that verbal tests are more subject to those influences.


Here are two key tables, Table 1 showing the basic facts about the samples from which the selection was made, Table 4 showing the characteristics of the selected “gifted” children.

Naglieri universal screening

Table 1 is complicated. It gives the demographics of the students, which is fine, but then reveals that half of them are deemed eligible for the lower entry standard Plan B. So, the simple explanation is that this group had fewer gifted children, but that IQ tests were better at detecting them, albeit to the lower standard required of them.

Naglieri scores for A and B

Table 4 is also complicated. For the full sample of 78,065 students (see bottom line) the average IQ is roughly 130, for the 37.947 on Plan B the IQ is 124. Scholastic achievement scores are much higher in the full sample than the Plan B sample. So, Plan A students are probably at a mean of 136.

In fact, one of the unremarked features of the data is that, as the proportion of white children in the school population falls from 61% in 2004 to 43% in 2007, achievement drops from 1.39 to 1.22, and even more in the Plan B group.

There is more detail which could be looked at, but I draw two main points from this paper.

• Category: Science • Tags: Blacks, IQ, Psychometrics 
Some immigrants don’t contribute much: locals blamed.
🔊 Listen RSS


Commenting on the findings shown on the Government website, the Prime Minister said: “What this audit shows is there isn’t anywhere to hide. That’s not just for Government, it’s for society as a whole. Britain has come a long way in promoting equality and opportunity but what the data we’ve published today shows is that we still have a way to go if we are going to truly have a country that works for everyone.” She added: Organisations will be forced to “explain or change” over the report’s findings, admitting the “findings will be uncomfortable” for public services.

I had a look at this much-trailed UK Government website on racial differences. It follows the well-known pattern, of treating race as a “now you see it, now you don’t” variable. That is, you are allowed to see differences in outcomes, but not differences in ability or character. For the opinion-forming, chattering classes, the tone was set by the required-listening Radio 4 news program, The World at One, in which the invited guests did not dream of mentioning ability, but competed to say how quickly the Government should institute policies to overcome the revealed disparities.

For example of “now you see it, now you don’t”, as regards education, there are many tables on scholastic differences between races, but absolutely nothing on any intelligence differences, despite the Cognitive Abilities Test being given to all schoolchildren, providing verbal, quantitative and non-verbal test scores. Hamlet without the Prince. I have asked the officials where I can find them on the site. In fact, some of the data can be dug up elsewhere, and I show it later on.

For example, as regards crime and punishment, there are many tables on racial differences in stop-and-search, arrest rates and so on, but absolutely nothing on racial classification of perpetrators as described by victims and witnesses. One cannot evaluate arrest rates without knowing who needs to be arrested. If victims and witnesses say the assailant was Chinese, the Police should be searching for Chinese persons. Indeed, the effectiveness of the Police in that instance would depend on them stopping and searching many Chinese people.

For example, there are many tables on standards of housing, but nothing I can find on savings rates, which is relevant to later wealth, quality of housing, and particularly of home ownership.

Do you need to know any more? The website is a question-begging selection which insinuates bias without providing fair benchmarks. Look at the collection of findings to see if I have missed anything, or judged it too harshly:

Despite the short-comings of the official website, I assume you expect some comment from me, so here are a few points.

Government interventions are going to have to be early, very early. By 21 weeks differences in head circumference are apparent. Not 21 weeks of life, 21 weeks of gestation.

Buck et al. (2015) “significant differences in HC (head circumference) were detected at 21 weeks (in descending order): Whites, Asians, Hispanics, and Blacks (all pairwise comparisons were highly significant except between Asians and Hispanic groups).” Mothers had been selected as being in good health, and were dropped from the study if there were any complications, which means that the strongest of the presumed social and environmental variables which might affect the developing foetuses have been reduced. Since environmental variables have an ineffable X factor, they can never be totally removed, environmentalists assert.

In the first few days of life racial differences are apparent in motor development and tolerance of restriction.

Somewhere between 3 and 4 years of age, tests detect racial differences in intelligence between black and white children.

By 7 years of life, the differences are stark. Here, from the website, are the percentages of children, by racial group, who can do Maths to the expected standard, and also to the higher standard. Maths can be defined logically, and so has an inherent measure of difficulty. Because of possible manipulation of basic pass rates, I have ranked the racial groups by the higher standard. To give you an idea of how feeble the “expected” standard is, note how long it holds up while the percentage of “higher standard” students collapses. The expected standard shows a 3-fold range, the higher standard a 20-fold range. “Expected” is a cop-out, “higher standard” the real thing which will determine employability.

Maths at expected and higher standard age 7

If teachers were trying to treat children badly according to their race, is this demonstrated outcome likely? In terms of racial purity, Black Africans are the real thing, being indubitably both Black and African, and so if teachers were strongly racist against Black Africans they would treat them the worst. But the (presumably mostly white) teachers allow them to excel as much as the local White British, all of these groups getting 18% of their number into the higher achieving category. Black Caribbeans, though Black and much more likely to have lived in Britain for a generation, if not two, and talking with local accents which should ingratiate them even with racism-inclined teachers, do less well, with only 12% excelling. Also, why are teachers so in love with the Chinese and Indians, clearly visible as genetic intruders? Why do they turn against White Irish Travellers (2% excel) having previously loved White Irish (20% excel). Mysterious are the ways of stereotypes. In the UK, teachers are not showing racial solidarity, or not in a consistent pattern, as required by racial supremacy. It is almost as if teachers did their best with all their students, regardless of their race. The results would be consistent with the observation that education in all well-organized countries contributes no more than 10% of the variance in scholastic outcomes, 90% being determined by the students themselves.

Immigrant scholastic results in each host country are mostly predictable from country of origin scholastic results, but there is variation according to immigration histories, such as whether skilled or unskilled labour was required at any particular time. UK Indians and Africans may not be representative of those groups in their home countries, and Indian province data shows enormous variation in scholastic ability. We cannot be sure that such regional, national and tribal differences exist in Africa, but further research may reveal cognitive elites.

Some groups are bimodal in terms of socio-economic status, such as Black Africans. The social profile of African immigrants is probably bimodal. They have almost as many parents in the professional ranks as the UK average, but also a very large number of unemployed persons. It is an odd distribution, suggestive of at least two different sources of immigrants as regards social status.

🔊 Listen RSS

Mankind Quarterly vol 58 no 1

Sex differences are in the news. A male Google employee reviewed some of the literature on the topic in the context of his workplace practices, and got sacked. A book questioning the role of testosterone in sex differences, and more generally the veracity of innate biological sex differences, got the Royal Society Science Book prize, though it was not reviewed by Royal Society Fellows expert in that area of knowledge. More generally, there are frequent news items about the lack of women in STEM subjects, in technology jobs and in corporate boardrooms, and these discussions often blame a glass ceiling of misogyny impeding women’s progress. Meanwhile, with rather less publicity, Prof Richard Lynn has revisited his 1994 paper in the light of recent research, and invited critics to take his finding apart.

As Editor Gerhard Meisenberg comments:

In this issue of Mankind Quarterly, Richard Lynn presents a data-rich summary of his developmental theory, followed by 10 comments by scholars working in the field and a reply to the comments. Many of the commentators add pieces of empirical evidence to the puzzle of sex differences, while others propose theoretical alternatives or refinements to the developmental theory. Taken together, the target article and comments offer a fairly representative overview of the current status of research on cognitive sex differences and the theoretical approaches used by different researchers in the field.

Here are the papers:

Sex Differences in Intelligence: The Developmental Theory. Richard Lynn
Male and Female Balance Sheet. James R. Flynn
Counting is not Measuring: Comment on Richard Lynn’s Developmental Theory of Sex Differences in Intelligence. Roberto Colom
Common Paradoxes in the Study of Sex Differences in Intelligence. Helmuth Nyborg
Cognitive Sex Differences: Evolution and History. David Becker and Heiner Rindermann
The Male Brain, Testosterone and Sex Differences in Professional Achievement. Edward Dutton
Sex Differences in Intelligence: Developmental Origin Yes, Jensen Effect No. Gerhard Meisenberg
Sex Differences in Self-Estimated Intelligence, Competitiveness and Risk-Taking. Adrian Furnham
Sex Differences in Intelligence: A Genetics Perspective. Davide Piffer
Presumption and Prejudice: Quotas May Solve Some Problems,but Create Many More.
Guy Madison
Sex Differences in Cognitively Demanding Games: Poker, Backgammon and Mahjong.
Heitor B.F. Fernandes
Sex Differences in the Performance of Professional Go Players. Mingrui Wang
Sex Differences in Intelligence: Reply to Comments. Richard Lynn

Subscriptions to Mankind Quarterly here:

Prof Lynn begins with the following observation:

It is a paradox that males have a larger average brain size than females, that brain size is positively associated with intelligence, and yet numerous experts have asserted that there is no sex difference in intelligence. This paper presents the developmental theory of sex differences in intelligence as a solution to this problem. This states that boys and girls have about the same IQ up to the age of 15 years but from the age of 16 the average IQ of males becomes higher than that of females with an advantage increasing to approximately 4 IQ points in adulthood.

Lynn goes on to show that most experts in the field assert that there are no sex differences in intelligence, or that such differences that exist cancel each other out. He then goes on to consider the obvious anomaly, that since brain size is related to intelligence, and men have larger brains than women, they ought to be more intelligent.

Pakkenberg and Gundersen (1997) reported that men have an average of four billion more neurons than women, a difference of 16 percent. Further data showing that men have more neurons than women, have been given by Pelvig et al. (2008).

Lynn then explains how he made his prediction about higher male intelligence:

To calculate the magnitude of the higher adult male IQ that would be predicted from the larger male brain size I took Ankney’s figure of the male-female difference in brain size expressed in standard deviation units of 0.78d and Willerman et al.’s (1991) estimate of the correlation between brain size and intelligence of 0.35. These figures would give adult males a higher average IQ of 0.78 multiplied by 0.35 = .27d = 4.0 IQ points. In my 1994 paper I presented data showing adult male advantages of 1.7 IQ points on verbal ability, 2.1 IQ points on verbal and non-verbal reasoning ability, and 7.5 IQ points on spatial, giving an average male advantage among adults of 3.8 IQ points and thus very close to the predicted advantage of 4.0 IQ points. I published further data for this male advantage in Lynn (1998, 1999). The male advantages given by Meisenberg (2009) given in Table 1 of 0.42d for whites and 0.30d for blacks are reasonably consistent with these results.

Eysenck accepted my thesis that men have a 4 points higher IQ than women and calculated that this advantage combined with the greater male variance of a standard deviation of 15 for men and 14 for women would produce 55 men and 5 women per 10,000 with an IQ of 160 and above, a ratio of 10:1. The same point has been made more recently by Nyborg (2015, p. 51), who presents data for a male advantage of 3.9 IQ points among American white 17 year olds and calculates that this advantage gives men a ratio of 5:1 to women at an IQ of 145 (approximately one per 300 males).

You may remember that I played around with these figures showing the male/female ratios which resulted from different assumptions about male/female differences in intelligence, and male/female differences in standard deviations.

Back to Lynn.

Table 1 shows how, as male brains become bigger their intelligence advantage grows bigger.

Lynn development of male advantage

In fact, even on the children’s version of the Wechsler (ages 6 to 16) there is a male advantage, and clear sex differences in the four indexes of ability, girls showing a processing speed advantage. As Lynn wryly observes, everyone has been wrong that there are no sex differences on the Wechsler which is a broad test of abilities, administered face to face and thus able to monitor engagement and effort. Furthermore, he has been wrong that sex differences do not show up till 16. They are present before that.

Lynn development wisc diffs

On the adult Wechsler, over 33 studies show that there is a clear pattern of male advantage, equivalent to 3.8 IQ points.

Lynn development table 5

The more detailed nature of adult sex differences in ability are shown in Table 6

• Category: Science • Tags: Gender, IQ 
🔊 Listen RSS

Biome mean annual temperature

If people were crops, where would they be best planted?

Like many people, I have read some books which have led me astray. They were plausible, and although I could see errors in them, I continued reading so as to learn new things. I am willing to accept that authors can be wrong about some things, yet right about others. It is a matter of degree. Baloney detection is not that difficult, and some books have to be cast aside, lest they clutter one’s mind with nonsense.

As to “Guns, Germs and Steel” I could see that the author was wrong about intelligence, deliberately wrong, but I grimly understood that many public figures feel it politic to lie about such things, and I was willing to keep reading in order to learn about geography and flora and fauna, about which I know too little. Cavalli-Sforza also found it politic to dismiss intelligence in the early pages of “Genes, Peoples and Languages”, and then give his results anyway in the rest of his book. Perhaps he knew that most people read only the first 25 pages. The same is true of many psychology researchers who like to denigrate intelligence, particularly when it comes to group differences. Their secrets are safe with me. I have no wish to see them bothered, just because they muddy the waters and leave dots unconnected. Live and let live.

Geography has somewhat lapsed in popularity as a discipline. Perhaps this is due to a conceit that humans have mastered geography, so late their bounded home, and can now rise above it, into a totally secure, well insulated, air-conditioned world. Such over-confidence may account for people who live in flood zones being surprised when they are flooded.

But geography is a common thread in the great debate about what makes some nations richer, better organized and more agreeable than others. In his tome, Jared Diamond blamed geography for these differences, arguing that people were all alike, but had to deal with different circumstances. (In fact, without any supportive evidence, he announced that the residents of Papua New Guinea were the brightest, perhaps seeking virtue through perversity). The contrary position is a longer-term view, namely that over generations (at least 16 of them?) people become adapted to their circumstances, and change in their character and ability through natural selection.

On this point, researchers have usually looked at latitude as an indicator of geographic influences. Distance from the Equator is a good predictor of outcomes. Can one do better than this, and include other relevant measures to get a best-fit between human types and their regions of origin? In that last question lies a complication. Originally in ships, and now more swiftly and cheaply in wide-body jets, people move around. In fact, looking back over history, migrations are hardly new, though they took far longer to achieve prior to modern transport. It should be possible to trace back genetic groups to their geographies of origin. Cavalli-Sforza required his subjects to show settled residence traceable back to 1492, which is a mere 21 generations ago, if we estimate 4 generations per century. Of course, with genetic methods we should be able to do better. However, the work to be considered below does not take up that particular issue. Rather, it seeks to create a typology of biomes which may be related to intelligence. It is a work in progress, initially a conference presentation (see link below) with a set of explanatory notes which set out some of the arguments, and the matters which still need to be resolved. It explores some options, and may give us a better approach to the general discussion of which environments favour the development of intelligence.

Using biome mapping and weighting to more precisely predict biogeographic differences in intelligence
Steven C. Hertler, Mateo Peñaherrera-Aguirre

Latitude and mean annual temperature powerfully predict the biogeographic distribution of intelligence. As single metrics, latitude and mean annual temperature have only one another as competitors. Of course, they are highly inter-correlated, with obvious causal connections. Mean annual temperature may in fact be superior to latitude because of its more explicitly composite nature. This is especially true if mean annual temperature is measured in a sophisticated fashion, across multiple measurement points which are then amalgamated. If this is done, mean annual temperature can implicitly account for oceanic warming trends and high altitude steppes and mountain ranges. However, there are other physical ecological components for which mean annual temperature fails to account, or does so insufficiently. One example is moisture or hydrology. Countries in Saharan and sub-Saharan Africa, for instance, do not greatly vary in mean annual temperature, even while having contrasting amounts of available water. Another example is soil quality. Inceptisols (early soil formations) and other soil types can be found across great stretches of latitude, and so are present at a variety of mean annual temperatures. Climatic factors such as hydrology and soil type that remain outside the reach of the predictive powers of mean annual temperature, nonetheless, are of great import to human cognitive evolution.

To improve upon a meta-indicator of climate, in this case latitude, Figueredo and colleagues included a Temperate Broad Leaf Deciduous Forest Factor. This turned out to be a powerful composite predictor variable because Temperate Broadleaf Deciduous Forests only exist within certain parameters; specifically between 40 and 60 degrees north latitude, within a particular band of temperature values, bounded in the north by permafrost, and in the south by competing coniferous tress. Their presence also denotes moderate moisture and rich brown alfisols (typically under a hardwood forest cover) which may be particularly productive of early agricultural yields, while discouraging helminth endo-parasites (worms). The presence of temperate broadleaf deciduous forest biomes seems particularly valuable in accounting for some nations with high outlying intelligence. Notwithstanding these observations, temperate forests are not present in the majority of regions or nations. Accordingly, the present study uses a global biome map made available by the World Wildlife Federation to extend the Temperate Broadleaf Deciduous Forest Factor into a broadly applicable biome classification system that can compass the full range of selective regimes. We present ordinarily ranked biomes via a hypothesized relationship to cognitive evolution. To the extent that the present distribution of cross-national intelligence is analysed, it is analysed in light of migration, population heterogeneity and predicted migration routes. The heuristic value of such a method is compelling. Physical ecology is to some extent better predicted while community ecology is directly measured.

At this point you might chose to look at the conference slides, and then come back to the notes.

Soils can be ranked by differences in fertility, and these vary considerably in different parts of the world.

Biome soils

Hertler cites the following passages from Walter & Breckle (1999), which are illustrative of the readings contributing to his biome model.

• Category: Science • Tags: Genetic Diversity, Geography, IQ 
🔊 Listen RSS

Royal Society - High Res -21

The Royal Society is the world’s oldest scientific society, and is held in high regard. To be a Fellow of that society is a great accomplishment. I am glad to have friends who have achieved this status, including one of the few couples who are both Fellows.

So, it is a considerable surprise to learn that the Royal Society has awarded their science book prize to Cordelia Fine’s “Testosterone Rex”. Awarding such a prize strongly implies that the science being described is reasonably well established, and that the prize-winning books can be trusted to give a fair and balanced account of their research fields. The prize implies a quality standard. The Society says:

Judges praised Fine’s powerful book, Testosterone Rex, for its eye-opening, forensic look at gender stereotypes and its urgent call for change.

Testosterone Rex: Unmaking the Myths of Our Gendered Minds by psychologist and author Cordelia Fine is the 30th anniversary winner of the ‘Booker Prize of science writing’, the prestigious Royal Society Insight Investment Science Books Prize.

In Testosterone Rex, Fine uses the latest scientific evidence to challenge – and ultimately overturn – dominant views on both masculinity and femininity, calling for readers to rethink their differences whatever their sex.

Why do I doubt the wisdom of this award? Well, in my case it goes back years to debates about sex differences between Simon Baron-Cohen and Cordelia Fine. My assessment was that Baron-Cohen was publishing interesting studies, and that Fine was making criticisms without attempting to carry out replications. A research finding is not invalidated by a contrary hypothesis, since that merely outlines something which might be the explanation, and itself needs to be demonstrated. I judged that Baron-Cohen was a researcher and Fine was a mixture of critic and polemicist. As you may remember, Baron-Cohen confirmed to me recently that his work on neonate visual preferences (newborn boys prefer a mechanical mobile, newborn girls prefer a human face) has yet to be replicated and invalidated, and thus still stands 16 years later. Furthermore, on matters of detail, Fine’s criticisms of that study ignore the counterbalanced presentation actually used as a method.

It is a matter of judgment as to whether the debaters are being fair to the relevant literature. I felt that Fine was selective in her arguments, and lost confidence that she was reporting research findings in a balanced way.

I had always assumed that the Royal Society Science Book prize was awarded by the Fellows, and that they were consulted about the basic science as described in each of the books. It seems not.

Professor Richard Fortey FRS was joined on the judging panel by: award-winning novelist and games writer, Naomi Alderman; writer and presenter of BBC Radio 4’s All in the Mind, Claudia Hammond; Channel 4’s Topical Specialist Factual Commissioner, Shaminder Nahal and former Royal Society University Research Fellow, Sam Gilbert.

These are the basis facts I have been able to gather about the judging panel.

Prof Richard Fortey, FRS is a British palaeontologist, geologist by training, who served as President of the Geological Society of London, with a primary research interest in trilobites. He is the author of popular science books on a range of subjects including geology, palaeontology, evolution and natural history.

Naomi Alderman is a novelist, author and game designer. At Oxford she read Philosophy, Politics and Economics.

Claudia Hammond is a broadcaster, writer and part-time psychology lecturer at Boston University’s London base where she lectures in health and social psychology. She has written three popular books.
Shaminder Nahal is a television journalist, Deputy Editor of a news program.

Dr Sam Gilbert, Institute of Cognitive Science, UCL. Has a strong research record in cognitive neuro-psychology.

By my reckoning, Prof Richard Fortey and Dr Sam Gilbert certainly, and Claudia Hammond probably, would have been perfectly capable of evaluating the arguments in the winning book, and looking at the critical reviews of Fine’s work over the years, at least going back to 2010. They could not have been out-voted. More recent reviews might have helped them:

In my view, two points stand out in these reviews. 1) Sex differences may exist for reasons other than testosterone 2) Dr Fine sometimes gives a selective view of a research paper in her text, and a more balanced view in a footnote, which might mislead a casual reader.

The stated judgment of the committee was that the winner had overturned dominant views on masculinity and femininity. Prof Fortey said:

“A cracking critique of the ‘Men are from Mars, Women are from Venus’ hypothesis, Cordelia Fine takes to pieces much of the science on which ‘fundamental’ gender differences are predicated. Graced with precisely focused humour, the author makes a good case that men and women are far more alike than many would claim. Feminist? Possibly. Humanist? Certainly. A compellingly good read.”

Given that there will be a general perception that a prize-winning science book is full of prize winning science, I think it would be prudent for the panel to be composed of Fellows, or if that is not possible, to at least ask Fellows in the relevant disciplines to give all short-listed books a proper evaluation before they are given prizes on the basis of being a “good read”.

• Category: Science • Tags: Academia, Feminism, Political Correctness 
Optimal prediction to the rescue.
🔊 Listen RSS

Hsu predicted and actual height

The “missing heritability” problem: current genetic analysis cannot explain as much variance as that suggested by population heritability estimates. This has been a cue for “Down with twin studies” arguments, in which those of dramatic inclinations have chosen to imagine that heritability estimates were thereby disproved. Not so. I was never particularly worried about this argument, regarding it as only a matter of time before the genetic code was cracked sufficiently to bridge the gap.

Another problem about breaking the genetic code is that some important human characteristics, like height and intelligence, are controlled by many genes of small effect. As regards height, this is in fact a problem of proportionality: tall people are usually taller not just because they have longer legs, though they do, but that they are generally longer and thus taller as a consequence. Building a taller body involves a large set of changes. Indeed, perhaps as many at 20,000 SNPs are required, each of them doing only a little. Equally, for intelligence as many as 10,000 SNPs may be involved. However, if many SNPs are required for an important trait, each doing very little, it is hard to prove or disprove their involvement. Rather than just identifying significant SNPs, showing that a technique can account for a good proportion of the overall variance is important. Prediction matters.

Now a paper comes along which claims to have hoovered up the SNP heritability variance for height, and to have done so by using machine learning, namely the LASSO or compressed sensing technique. It also gets 9% of the variance for scholastic attainment, close to the 10% I had previously mentioned as the current upper limit.

Accurate Genomic Prediction of Human Height. Louis Lello, Steven G. Avery, Laurent Tellier, Ana I. Vazquez, Gustavo de los Campos, and Stephen D.H. Hsu. bioRxiv preprint first posted online Sep. 18, 2017.
The abstract:

We construct genomic predictors for heritable and extremely complex human quantitative traits (height, heel bone density, and educational attainment) using modern methods in high dimensional statistics (i.e., machine learning). Replication tests show that these predictors capture, respectively, ∼40, 20, and 9 percent of total variance for the three traits. For example, predicted heights correlate∼0.65 with actual height; actual heights of most individuals in validation samples are within a few cm of the prediction. The variance captured for height is comparable to the estimated SNP heritability from GCTA (GREML) analysis, and seems to be close to its asymptotic value (i.e., as sample size goes to infinity), suggesting that we have captured most of the heritability for the SNPs used. Thus, our results resolve the common SNP portion of the “missing heritability” problem – i.e., the gap between prediction R-squared and SNP heritability. The∼20k activated SNPs in our height predictor reveal the genetic architecture of human height, at least for common SNPs. Our primary dataset is the UK Biobank cohort, comprised of almost 500k individual genotypes with multiple phenotypes. We also use other datasets and SNPs found in earlier GWAS for out-of-sample validation of our results.

The introduction sets out the problems clearly, and distinguishes the SNP hunting techniques from genomic prediction which, “based on whole genome regression methods, seek to construct the most accurate predictor of phenotype, tolerates possible inclusion of a small fraction of false-positive SNPs in the predictor set. The SNP heritability of the molecular markers used to build the predictor, can be interpreted as an upper bound to the variance that could be captured by the predictor”.

The authors have used the UK Biobank database with nearly 500,000 genotypes. The paper has, quite necessarily, very technical supplementary appendices, but the underlying approach is to use large samples of the data to train the learning procedure, and then test the results on samples of 5,000 genotypes which had been held apart for that purpose. In my primitive terms, the sample of discovery is used to generate the best predictor, warts and all, and that is tested on the sample of proof. I like this, because it is pragmatic, not burdened by too many prior assumptions about genes, uses all the data to advantage, and is willing to include weak signals.

Fig 3 shown above reveals a good fit with the data for height.

As the authors say in their discussion

Until recently most work with large genomic datasets has focused on finding associations between markers (e.g., SNPs) and phenotype. In contrast, we focused on optimal prediction of phenotype from available data. We show that much of the expected heritability from common SNPs can be captured, even for complex traits affected by thousands of variants. Recent studies using data from the interim release of the UKBB reported prediction correlations of about 0.5 for human height using roughly 100K individuals in the training[19]. These studies forecast further improvement of prediction accuracy with increased sample size, which have been confirmed here.

We are optimistic that, given enough data and high-quality phenotypes, results similar to those for height might be obtained for other quantitative traits, such as cognitive ability or specific disease risk. There are numerous disease conditions with heritability in the 0.5 range, such as Alzheimer’s, Type I Diabetes, Obesity, Ovarian Cancer, Schizophrenia, etc. Even if the heritable risk for these conditions is controlled by thousands of genetic variants, our work suggests that effective predictors might be obtainable (i.e., comparable to the height predictor in Figure (4)). This would allow identification of individuals at high risk from genotypes alone. The public health benefits are potentially enormous.

We can roughly estimate the amount of case-control data required to capture most of the variance in disease risk. For a quantitative trait (e.g., height) with h2∼0.5, our simulations predict that the phase transition in LASSO performance occurs at n∼30s where n is the number of individuals in the sample and s is the sparsity of the trait (i.e., number of variants with non-zero effect sizes). For case-control data, we find n∼100s (where n means number of cases with equal number controls) is sufficient. Thus, using our methods, analysis of∼100k cases together with a similar number of controls might allow good prediction of highly heritable disease risk, even if the genetic architecture is complex and depends on a thousand or more genetic variants

In summary, this is exciting stuff. It would appear that, given large samples and meeting signal sparsity requirements, compressed sensing may help track down predictive formulas for many traits and conditions. The benefits are enormous, as are that greatest benefit, a gain in understanding.

• Category: Science • Tags: Genetics of Height, Genomics, Height, Heritability 
🔊 Listen RSS

scandi murder mystery

Longitudinal studies are a mixed blessing for any up-and-coming researcher. You have to wait so long that the world seems to pass you by, and when you finally publish, your results are often drowned out by more fashionable controversies. Worse, you might die before you actually get any interesting results. Yet, if you can last the course, you get the last laugh, because the long view of human development is the most informative, since each person is their historical control. The cross-sectional dross is revealed to be a mess of confounding variables. The true strong light is longitude.

Below is a short video about the Study of Mathematically Precocious Youth, now 45 years old. It is a university production, with an upbeat American tone. If that proves too much for you, watch a Scandinavian murder mystery instead.

🔊 Listen RSS

compressed sensing

In my last post “Even more genes for intelligence”, I alluded to the mysterious Hsu Boundary, and I encourage you to use this phrase as often as possible. Why should other researchers have a monopoly of jargon? The phrase should help you impress friends, and also to curtail tedious conversations with persons who have limited understanding of sampling theory, themselves the biggest sample of all.

The “Hsu boundary” is Steve Hsu’s estimate that a sample size of roughly 1 million people may be required to reliably identify the genetic signals of intelligence. However, that has to be 1 million real persons, with individual data points, on which the best available techniques can be applied, not aggregated samples which are then subjected to a meta-analysis.

The reason for this is that the genetic code is a very long message. Even when summarized according to agreed principles, it can generate multiple comparisons, and is a rich soil for false positives. In reaction to that, significance levels are correspondingly raised to demanding levels, but that may rule out some real signals. A sample of at least 1 million, Steve calculated, would be required to get around this problem. Once gathered, then more advanced methods, beyond linear regression, could be applied to the data.

Aggregated samples, put together by international collaborative projects, cannot always take the level of analysis down to the individual patient. They are doing meta-analysis, aggregating together data from many sources. They share summary statistics i.e., the statistical evidence from linear regression in favour of association of a specific SNP with the phenotype. This has the advantage of making it easier to pool data, but it is not the most effective method for building a predictor. Hsu does not believe they will cross any special threshold from summary statistics on ~1M samples. They will, however, obtain better and better results as power increases. They will find patterns which have tighter confidence limits, and as such they will be identifying stronger signals.

On a wider note, it may be available somewhere, but we need an accessible central register of the samples used in all studies, particularly for those studies that then go on to aggregate them for larger sample meta-analysis. This would allow us to understand overlaps between different meta-analyses.

A complexity that we have discussed before is that internationally aggregated samples on intelligence will probably have been measured with different tests. For once, the theory of general intelligence assists us here, in that a comparable g can be extracted from a broad range testing procedures, putting all subjects onto the same g scale. An additional complexity is that for many samples no psychometric test scores are available, but scholastic tests are far more commonly obtainable. Scholastic attainment is very important, but it is not perfectly correlated with intelligence.
In a major study, Ian Deary and colleagues found a correlation of .8 between cognitive ability at 11 years and national examinations at age 16.

Intelligence and educational achievement. / Deary, Ian J.; Strand, Steve; Smith, Pauline; Fernandes, Cres. Intelligence, Vol. 35, No. 1, 2007, p. 13-21.

Excellent, but probably as high as can be achieved, and international scholastic levels will vary considerably, thus making the aggregation of subjects in different national school systems somewhat error prone. An even less powerful measure of intelligence is “years of education”. This is subject to many artefacts, typically that it is a reasonable measure when the extra years are only open to brighter students, but less so when nations are seeking to boost the abilities of all students by requiring them to stay in school longer.

Back to the analysis of genetic data. If you have all the individual data in one place, and have a reliable and valid measures of mental ability, you can use more sophisticated machine learning techniques, where Hsu predicts a threshold at ~million or so genomes (could be 2 million; not that precise). Summary statistics + linear regression has advantage that it can be applied through meta-analysis without sharing samples – you can pool lots of data without altering the original ethical requirements, since individual data are not shared.

What are these more sophisticated machine learning techniques? Compressed Sensing is the front runner, a signal processing paradigm which has an algorithm which captures all the locations with some effect on intelligence, so long as there are not too many of them relative to the sample size. The more advanced technique where Hsu predicts a boundary is called Compressed Sensing:

At the reasonable level of heritability of roughly .5 and a high probability threshold required for a real hit, then:

For heritability h2 = 0.5 and p ~ 1E06 SNPs, the value of C log p is ~ 30. For example, a trait which is controlled by s = 10k loci would require a sample size of n ~ 300k individuals to determine the (linear) genetic architecture.

We show using CS methods and theory that all loci of nonzero effect can be identified (selected) using an efficient algorithm, provided that they are sufficiently few in number (sparse) relative to sample size. For heritability h2 = 1, there is a sharp phase transition to complete selection as the sample size is increased. For heritability values less than one, complete selection can still occur although the transition is smoothed. The transition boundary is only weakly dependent on the total number of genotyped markers. The crossing of a transition boundary provides an objective means to determine when true effects are being recovered. For h2 = 0.5, we find that a sample size that is thirty times the number of nonzero loci is sufficient for good recovery.

So, this approach identifies a real boundary. So long as the important signals are sparse (usually the case) then a third of a million individuals suffice.

Finally, we appear to have come to the true HSU boundary, a phase transition in which selection of signals becomes easier. Is it like moving from the troposphere to the stratosphere? Perhaps it is more like the familiar natural phase transition or phase boundary shown at a very precise threshold (e.g., 100 degrees Celsius) where the basic organization of atoms and molecules can change drastically (e.g., H2O changes from a liquid to a vapor).

Similarly, the behaviour of an optimization algorithm involving a million variables can change suddenly as the amount of data available increases. We see this behavior in the case of Compressed Sensing applied to genomes, and it allows us to predict that something interesting will happen with complex traits like cognitive ability at a sample size of the order of a million individuals.

Machine learning is now providing new methods of data analysis, and this may eventually simplify the search for the genes which underpin intelligence.

• Category: Science • Tags: I.Q. genomics, IQ 
🔊 Listen RSS

Intelligent tissues

The intelligence gene hunters have been stepping up their activities, and keep coming back with more trophies. Danielle Posthuma and colleagues are at it again, studying very large samples and finding further novel genes which load on brain tissues. I hope someone somewhere is keeping track of the overall picture, perhaps in a control room with multiple screens, like the NASA control centre of old, tracking the orbit of each SNP as it hoves into sight.

This is all very good, but it makes life difficult for mere commentators. When starting to write my comments I chose the working title “More genes for intelligence”. When trying to save it my Word program told me, somewhat severely, that I already had a document of that name. So, perhaps this note will have to read “Even more genes for intelligence.

This was the position in May:

Briefly, in a study of 78,308 individuals Sniekers et al. said:

We identify 336 associated SNPs (METAL P < 5 × 10−8) in 18 genomic loci, of which 15 are new. Around half of the SNPs are located inside a gene, implicating 22 genes, of which 11 are new findings. Gene-based analyses identified an additional 30 genes (MAGMA P < 2.73 × 10−6), of which all but one had not been implicated previously. We show that the identified genes are predominantly expressed in brain tissue, and pathway analysis indicates the involvement of genes regulating cell development (MAGMA competitive P = 3.5 × 10−6). Despite the well-known difference in twin-based heritability for intelligence in childhood (0.45) and adulthood (0.80), we show substantial genetic correlation (rg = 0.89, LD score regression P = 5.4 × 10−29). These findings provide new insight into the genetic architecture of intelligence.

So, where are we in September?

GWAS meta-analysis (N=279,930) identifies new genes and functional links to intelligence. Savage et al. say:

Intelligence is highly heritable and a major determinant of human health and well-being. Recent genome-wide meta-analyses have identified 24 genomic loci linked to intelligence, but much about its genetic underpinnings remains to be discovered. Here, we present the largest genetic association study of intelligence to date (N=279,930), identifying 206 genomic loci (191 novel) and implicating 1,041 genes (963 novel) via positional mapping, expression quantitative trait locus (eQTL) mapping, chromatin interaction mapping, and gene-based association analysis. We find enrichment of genetic effects in conserved and coding regions and identify 89 nonsynonymous exonic variants. Associated genes are strongly expressed in the brain and specifically in striatal medium spiny neurons and cortical and hippocampal pyramidal neurons. Gene-set analyses implicate pathways related to neurogenesis, neuron differentiation and synaptic structure. We confirm previous strong genetic correlations with several neuropsychiatric disorders, and Mendelian Randomization results suggest protective effects of intelligence for Alzheimer’s dementia and ADHD, and bidirectional causation with strong pleiotropy for schizophrenia. These results are a major step forward in understanding the neurobiology of intelligence as well as genetically associated neuro-psychiatric traits.

What is all this about?

As you know, I try to understand these procedures by analogy with code breaking. The actual DNA code comes all wrapped up, so it needs to be broken apart, shattered into pieces and then assembled again to reveal its underlying sequence. This involves some assumptions, but the mapping of the code is all about finding where everything is located, ideally precisely where it is in the sequence of base pairs. However, this approach loses the packaging information, so researchers pay attention to genetic linkage, the tendency of DNA sequences that are close together on a chromosome to be inherited together. Call this the packaging information. Guilt by association. Of course, this code is very complicated, but at least it has been road tested for millennia. This may account for many sections being conserved, on the sensible basis one does not tamper with the instructions on which life depends, even though some of those instruction may be redundant.

There are many techniques being used, and this paper describes the results of each in supplementary sections. Positional mapping means that genetic variants are linked to a gene when they are physically located inside that gene. Expression quantitative trait locus mapping (eQTL) links a genetic variant to a gene when that variant changes the expression of that gene. The variant is not necessarily located inside the gene. Chromatin interaction mapping links genetic variants to genes by looking at the 3D organization of chromosomes, which may allow remote genetic variants to influence a gene when they become close through DNA folding. Exonic variants are the parts which code for proteins, whereas introns do not. (Consider introns to be intervening sequences and exons to be expressed sequences). One day someone will write a user’s manual for all this research, though it will have to be updated every few months.

In gene-set analysis using the GWGAS results, six Gene Ontology gene-sets were significantly associated with intelligence: neurogenesis (Beta=0.153), neuron differentiation (Beta=0.178), central nervous system neuron differentiation (Beta=0.398), regulation of nervous system development (Beta=0.187), positive regulation of nervous system development (Beta=0.242), and regulation of synapse structure or activity(Beta=0.153). Conditional analysis indicated that there were three independent associations, for the neurogenesis, central nervous system neuron differentiation, and regulation of synapse structure or activity processes, which together accounted for the associations of the other three sets. Linking gene-based P-values to tissue-specific gene-sets, we observed strong associations across various brain areas (as shown in the figure), most strongly with the cortex (P=5.12×10-9), and specifically frontal cortex (P=4.94×10-9). In brain single-cell expression gene-set analyses, we found significant associations of striatal medium spiny neurons (P=1.47×10-13) and pyramidal neurons in the CA1 hippocampal (P=4×10-11) and cortical somatosensory regions (P=3×10-9).

Using polygenic score prediction we show that the current results explain up to 5.4% of the variance in four independent samples.

Our results also suggested a protective effect of intelligence on ADHD (OR=0.46) and Alzheimer’s disease (OR=0.66). In line with a positive genetic correlation, we observed that intelligence was associated with higher risk of autism (OR=1.47). There was evidence of a bidirectional association between intelligence and schizophrenia including a strong protective effect of intelligence on schizophrenia (OR=0.58), and a relatively smaller reverse effect (bxy= −0.195), with additional evidence for pleiotropy.

• Category: Science • Tags: Genomics, IQ 
🔊 Listen RSS

Longevity survival days

In describing the Kaplan-Meier survival graphs in “Vita Brevis, Dignitatis Inutilis” I correctly described the findings in the Iveson et al. paper, but then went a step too far in equating a decrease in mortality risk with an identical increase in lifespan. I said:

The good news is that a standard-deviation increase in IQ score is associated with a 24% decrease in mortality risk. So, at IQ 115 lifespan is 24% longer than average.

The second sentence should have read “So, at IQ 115 your chance of getting to a particular age, say 79, is 24% better than average”.

Calculating the actual lifespan increase is somewhat more complicated. That is mostly because the living are still living, and their deaths are not yet precisely foretold. Nonetheless, the curves give the best estimates of survival, and show that chances of survival of brighter persons are better than average.

Regarding the findings in their paper, the authors tell me:

We usually find that people are satisfied with knowing the Hazard Ratio per Standard Deviation of intelligence. However, you could look at the horizonal (x axis; time) distance for a given probability of being alive for -1 Standard Deviation, mean, and +1 Standard Deviation of intelligence.

To make your own estimates, look at Fig 1 and decide on a probability level, or an age to be achieved. Then read the result off the other axis.

For example, if you accept that a .8 probability is a fair bet for life planning purposes, say as regards pensions, retirement expenditure generally, or starting new ventures then, very roughly and judged by eye:

IQ 115 live 26,500 days which is 72.6 years
IQ 100 live 24,500 days which is 67.1 years
IQ 85 live 23,000 days which is 63 years

So, the “cost” of being of average intelligence rather than 1 standard deviation above average is 5.5 years.

As a rough rule of thumb, those of IQ 115 live 10 years longer than those of IQ 85. Since the latter group will have much lower wages than those of IQ 115 it is natural to assume that social class of origin determines the difference in lifespan, and that wealth is what causes “inequalities” in health outcomes. In fact social class only slightly reduces the effect of intelligence on lifespan.

In summary, it appears that those born with “system integrity” have better brains and better bodies, an advantage which is theirs to cherish or to lose.

• Category: Science 
No Items Found
James Thompson
About James Thompson

James Thompson has lectured in Psychology at the University of London all his working life. His first publication and conference presentation was a critique of Jensen’s 1969 paper, with Arthur Jensen in the audience. He also taught Arthur how to use an English public telephone. Many topics have taken up his attention since then, but mostly he comments on intelligence research.