Height, Flynn Effect, and shared environment

Some interesting examples of shared environmental effects on height.

Atlantic: ... a database of 2,236 British soldiers who served in World War I, and then they looked up their birth records. The soldiers were relatively representative of the male population as a whole—about two-thirds of the 1890 British male birth cohort enlisted. It turns out that subtle differences in their heights hinted at their origins:

Those from white-collar backgrounds were taller: This follows the theory that wealth buys better food and living conditions, and thus greater height in adulthood. The men who hailed from the top two social classes stood a half-inch taller, on average.

The more kids there were in a household, the shorter they were: Not only because there was less food to go around, but also because it made it more likely that there were more people in each bedroom. “Crowding can help spread respiratory and gastrointestinal infections,” Hatton said. “People sneezing on each other, that sort of thing.” Each additional sibling cost the men an eighth of an inch, and having more than one person per bedroom shaved off a quarter-inch.

Children of literate mothers were taller: When mothers couldn’t read, they were less likely to know about the importance of a balanced diet or clean cutlery. The researchers measured the percentage of women by region who were only able to sign their marriage certificates with an X, rather than their name. People from areas with a high percentage of illiterate mothers were a quarter-inch shorter.

People from industrial districts were shorter than those from agricultural areas: Regardless of income, the Dickensian living conditions of 19th century British cities suppressed height by about nine-tenths of an inch.

Average European male height increased 11 centimeters between 1870 and 1970, about a centimeter per decade, or 1.5 SD in a century. Seems suspiciously similar to the Flynn Effect! In addition, over the same period of time average years of schooling went up significantly: e.g., UK 1870 4 years to 1930 7 years (see Appendix A of this paper). Most individuals born 100 years ago experienced significant deprivation by modern standards.

My own view is that there is nothing particularly mysterious about the Flynn Effect: living conditions, nutrition, and availability of education have all improved drastically in the last 150 years. So g scores should have as well. The Flynn Effect can be consistent with high heritability for non-deprived individuals in modern environments.

See also Swedish height in the 20th century and Flynn on the Flynn Effect.

The north-south gradient in average height found in Europe (see figure) may be a consequence of differential selection pressures that vary by region.

Flynn on the Flynn effect

Via GNXP, this lecture by James Flynn. The lecture is long but worth reading in full. His discussion of factor loading in the context of the decathalon, 100m speed, hurdling, etc. is excellent and very relevant to the subject of intelligence and its various cognitive components. The Flynn effect refers to a significant increase in raw IQ scores over the last 100 years or so -- equivalent of 30 points or more, in some cases. This raises a number of thorny issues. Were our ancestors idiots? Is IQ really that malleable under environmental influence (contrary to recent twin studies)? Is the principal component identified as "g" actually time dependent? See Flynn's answers below. A couple of comments: (i) Dramatic gains are seen only in certain areas of intelligence (see table from Sailer discussion), which are plausibly the areas in which modern life provides much more stimulation. (ii) Average people 100 years ago were massively deprived by modern standards -- much more than we could ever reproduce in a modern twin study. US GDP per capita is 10x higher now, and average years of schooling has increased dramatically. It's not surprising that a child who only received (say) 6 years of formal schooling would be far behind someone with 12. Modern twin studies only include kids raised in a much smaller range of environments -- no twin in any recent study had less than the legally mandated US high school education. With a smaller range of environmental effects, the genetic component plays a larger role, leading to the high (.5-.7 or so) heritability result for IQ (similar to height). (Flynn notes: In the America of 1900, adults had an average of about 7 years of schooling, a median of 6.5 years, and 25 percent had completed 4 years or less.) (iii) The analogy with height is quite appropriate. While taller parents tend to have taller children (i.e., height is heritable), we've seen siginificant gains in average height as nutrition and diet have improved. (iv) I think Flynn would agree that variance in adult IQ must have been much larger in the past. People like Newton or Thomas Jefferson obviously had tremendously more exposure to ideas and abstract thinking than a kid on the farm with little or no education, few books, no TV and no radio. The Flynn effect does not imply that the great geniuses of the past were necessarily inferior to those of today.

Lecture by Professor James Flynn at The Psychometrics Centre, Cambridge Assessment, University of Cambridge, 15th December 2006 Naming the paradoxes (1) The factor analysis paradox: Factor analysis shows a first principal component called "g" or general intelligence that seems to bind performance on the various WISC subtests together. However, IQ gains over time show score gains on the WISC subtests occurring independently of one another. How can intelligence be both one and many? (2) The intelligence paradox: If huge IQ gains are intelligence gains, why are we not stuck by the extraordinary subtlety of our children's conversation? Why do we not have to make allowances for the limitations of our parents? A difference of some 18 points in the average IQ over two generations ought to be highly visible. (3) The MR paradox: In 1900, the average IQ scored against current norms was somewhere between 50 and 70. If IQ gains are in any sense real, we are driven to the absurd conclusion that a majority of our ancestors were mentally retarded. (4) The identical twins paradox: Twin studies show that genes dominate individual differences in IQ and that environmental effects are feeble. IQ gains are so great as to signal the existence of environmental factors of enormous potency. How can environment be both so feeble and so potent? The solutions in shorthand (1) The WISC subtests measure a variety of cognitive skills that are functionally independent and responsive to changes in social priorities over time. The inter-correlations that engender "g" are binding only when comparing individuals within a static social context. (2) Asking whether IQ gains are intelligence gains is the wrong question because it implies all or nothing cognitive progress. The 20th century has seen some cognitive skills make great gains, while others have been in the doldrums. To assess cognitive trends, we must dissect "intelligence" into solving mathematical problems, interpreting the great works of literature, finding on-the-spot solutions, assimilating the scientific world view, critical acumen, and wisdom. (3) Our ancestors in 1900 were not mentally retarded. Their intelligence was anchored in everyday reality. We differ from them in that we can use abstractions and logic and the hypothetical to attack the formal problems that arise when science liberates thought from concrete referents. Since 1950, we have become more ingenious in going beyond previously learned rules to solve problems on the spot. (4) At a given time, genetic differences between individuals (within a cohort) are dominant but only because they have hitched powerful environmental factors to their star. Trends over time (between cohorts) liberate environmental factors from the sway of genes and once unleashed, they can have a powerful cumulative effect.

Psychometrics links

Retired physicist Bob Williams writes: "... I have studied psychometrics and intelligence research for about 17 years and have gotten to know almost all of the current researchers through my membership in the International Society for Intelligence Research. I attend the conferences and discuss the recent studies with the people who have done them."

He sent me the following interesting links:

Summary of research on heritability of intelligence

Diminishing influence of g at high end of ability

Summary of research on evolution of human intelligence


From Williams' notes on ISIR 2007: a discussion with James Flynn:

The interviewer was David Lubinski (Vanderbilt). He started the interview by asking Flynn to tell us about his life. From that point on, Lubinski only managed to get in a dozen or so words. Once the camera started, Flynn launched into an enthusiastic discussion of himself and his ideas. He made a favorable impression from the start by telling us that he grew up in a Catholic family and has become a serious atheist. He also immediately told us of his well known (socialist) political orientation and did not ever attempt to disconnect his political notions from his thoughts about intelligence and the secular rise that bears his name. Flynn speaks clearly and with a loud voice, which was well suited for the interview and later became something of a tool for overpowering people with whom he disagreed. In spite of this, I found him to be very likable, smart, and complicated. He occasionally took positions that were diametrically opposed and admitted that he had done so.

Flynn mentioned his first article (1984), showing a 14 point gain over 30 years (I may have gotten the span wrong, but I think this is correct). In 1987 he published a paper that showed the largest gains were on the Raven’s tests. He has found little, if any, gain in vocabulary. It is well known that FE (Flynn Effect) gains have been greatest in tests of abstract reasoning and least in tests that relate to scholastic items. This contrast immediately suggests that the FE gains have not been g loaded, since g loaded gains would necessarily boost all cognitive tasks. Flynn has not discussed this. I asked him what portion of the gains were g loaded and he went into his history discussion, then finally said “I don’t know.”

Flynn has approached the FE as a matter to be discussed qualitatively, without the support of carefully constructed research studies. Throughout his talk (this was an interview, but it turned into a speech), he described intelligence as a factor with historical variability. His debate with Charles Murray is available on the web and contains the same illustrations that he used in the “interview.” His bottom line is that test items have become easier over time, not because people are more intelligent, but because the items being tested have taken on a more central role in their lives. This makes sense with some test items (vocabulary, for example, but that is one area where he has found little increase), but is not at all apparent with respect to other test items, some of which are unrelated to daily experiences (just look at the list of subtests used in the W-J III).

He mentioned reaction time tests and simply discarded them as not important. This is not so easy to accept, given that a battery of RT tests can produce a very highly g loaded discrimination, yet chronometric measures have not been shown to change over decades. Meanwhile, Flynn was positive about the future of brain imaging and seemed willing to accept future results, even if they support causation that is opposite of his preference. He expects that such research will ultimately find a physiological seat of g. I fully agree, but also find this comment to be in conflict with his arguments from the Dickens-Flynn paper (Heritability Estimates Versus Large Environmental Effects: The IQ Paradox Resolved).

Interestingly, Flynn does not believe that the FE will narrow the B-W IQ gap. Some environmentalists have argued that it will, but no evidence has surfaced to show that after decades of rise, the FE has actually narrowed the gap.

Flynn admitted to having been fired twice and explained that those events prompted him to move to New Zealand. He claims to have defended Larry Summers, James Watson, and Chris Brand. ...In these areas, Flynn shows himself to be both honest and inconsistent. He is inherently likable and extraverted, although I remain of the opinion that he is less interested in discussion than in lecturing. I talked to him for a while at a reception and found him to be pleasant, polite, and sharp.

During his commentary, Flynn said that he favored affirmative action. In fact, he showed a strong interest in finding ways to help blacks overcome various obstacles. When Gottfredson questioned him about affirmative action, he said he is not a fanatic on the subject and admitted that it is unfair to whites. He said that, when employers want to hire a black worker, they are generally unconcerned with whether the employee is outstanding or average.

He attributed at least some of the FE gains to decreasing family size and noted that it cannot go much lower. When I had a chance to talk to him, I asked what portion of the FE gains he believes are g loaded. After repeating his historical perspective, he said “I don’t know.”

From notes on ISIR 2008

Jan te Nijenhuis

The Flynn effect in Korea

• United States FE rate, 3 points per decade.
• Scandinavia, recent slight reversal (IQ decrease).
• Estonia, recent gains.
• Japan, for those born from 1940 to 1965, 7.7 points per decade.
• Korea and Japan, same gains.
• Korea experienced an explosion in education and height of 1.5 to 2.0 SD.

The study involved two formats. One was to use single Raven’s test comparisons for two groups for 1952 and 1982. The other was to test a single group with two tests that were normed 30 years apart. The Koreans gained at about the same rate as did the Japanese, but delayed by 20 to 30 years. The gains have not yet leveled off or reversed.

Some useful references on this topic:

The Structure and Measurement of Intelligence

Genius: the natural history of creativity,

both by Hans Eysenck.

The g-factor, by Arthur Jensen

IQ regression from Terman data

A reader sent me a link to some data on regression from Eysenck's book The Structure and Measurement of Intelligence. (I actually own this book but had forgotten about these results!) Eysenck quotes from the Terman study:

1528 California children with IQs of 140 or higher were followed into adulthood in order to assess the importance of IQ in adult success and adjustment. The mean IQ of those that married and had children was 152; that of their spouses, 125. The mean IQ of this whole group of parents was 138.5. The mean IQ of 1571 of their children was 133.2, a little less than the parents and showing some regression to the mean.

These results show very little regression -- less than would be predicted from the estimates given in my earlier post. However, I suspect a correction needs to be applied as the Terman study was done before the Flynn Effect was understood. If the same version of the Stanford Binet was used on the parents and children, without Flynn re-norming, then the children were not as exceptional as implied by the 133 average given above. If we take the Flynn Effect in mid-20th century America to be about 10 points per generation (this is just a rough guess), then the 133 should be corrected to about 123, and is pretty consistent with the estimate I gave of: 100 + .6(38.5) = 123, which assumes narrow sense heritability h^2 = .6.

The discussion by Eysenck uses higher heritability numbers (note the non-negligible V_CE = variance due to common environment). If the Terman study used re-normed tests then their data would be evidence for much less regression than in my estimates.

Are we getting smarter? Why?

It is well known that raw scores on IQ tests have been increasing at a rate of what would be about 3 IQ points per decade (the so-called Flynn effect). This means that, were the result of the test not rescaled so that the average is 100 by definition, the average IQ would have risen to 130 over the last century - i.e., the average person today scores better than all but 2 percent or so of the population in 1900. Malcolm Gladwell discusses this effect and its possible causes in this week's New Yorker, reviewing a new book claiming that modern society, with its fast-paced multimedia entertainment (including video games, computers, TV, etc.) actually improves our cognitive skills.

Gladwell: "Twenty years ago, a political philosopher named James Flynn uncovered a curious fact. Americans—at least, as measured by I.Q. tests—were getting smarter. This fact had been obscured for years, because the people who give I.Q. tests continually recalibrate the scoring system to keep the average at 100. But if you took out the recalibration, Flynn found, I.Q. scores showed a steady upward trajectory, rising by about three points per decade, which means that a person whose I.Q. placed him in the top ten per cent of the American population in 1920 would today fall in the bottom third. Some of that effect, no doubt, is a simple by-product of economic progress: in the surge of prosperity during the middle part of the last century, people in the West became better fed, better educated, and more familiar with things like I.Q. tests. But, even as that wave of change has subsided, test scores have continued to rise—not just in America but all over the developed world. What’s more, the increases have not been confined to children who go to enriched day-care centers and private schools. The middle part of the curve—the people who have supposedly been suffering from a deteriorating public-school system and a steady diet of lowest-common-denominator television and mindless pop music—has increased just as much. What on earth is happening? In the wonderfully entertaining “Everything Bad Is Good for You” (Riverhead; $23.95), Steven Johnson proposes that what is making us smarter is precisely what we thought was making us dumber: popular culture.

...As Johnson points out, television is very different now from what it was thirty years ago. It’s harder. A typical episode of “Starsky and Hutch,” in the nineteen-seventies, followed an essentially linear path: two characters, engaged in a single story line, moving toward a decisive conclusion. To watch an episode of “Dallas” today is to be stunned by its glacial pace—by the arduous attempts to establish social relationships, by the excruciating simplicity of the plotline, by how obvious it was. A single episode of “The Sopranos,” by contrast, might follow five narrative threads, involving a dozen characters who weave in and out of the plot. Modern television also requires the viewer to do a lot of what Johnson calls “filling in,” as in a “Seinfeld” episode that subtly parodies the Kennedy assassination conspiracists, or a typical “Simpsons” episode, which may contain numerous allusions to politics or cinema or pop culture. The extraordinary amount of money now being made in the television aftermarket—DVD sales and syndication—means that the creators of television shows now have an incentive to make programming that can sustain two or three or four viewings.

...It doesn’t seem right, of course, that watching “24” or playing a video game could be as important cognitively as reading a book. Isn’t the extraordinary success of the “Harry Potter” novels better news for the culture than the equivalent success of “Grand Theft Auto III”? Johnson’s response is to imagine what cultural critics might have said had video games been invented hundreds of years ago, and only recently had something called the book been marketed aggressively to children: (Johnson) Reading books chronically understimulates the senses. Unlike the longstanding tradition of gameplaying—which engages the child in a vivid, three-dimensional world filled with moving images and musical sound-scapes, navigated and controlled with complex muscular movements—books are simply a barren string of words on the page. . . .

Books are also tragically isolating. While games have for many years engaged the young in complex social relationships with their peers, building and exploring worlds together, books force the child to sequester him or herself in a quiet space, shut off from interaction with other children. . . .

But perhaps the most dangerous property of these books is the fact that they follow a fixed linear path. You can’t control their narratives in any fashion—you simply sit back and have the story dictated to you. . . . This risks instilling a general passivity in our children, making them feel as though they’re powerless to change their circumstances. Reading is not an active, participatory process; it’s a submissive one.

He’s joking, of course, but only in part. The point is that books and video games represent two very different kinds of learning. When you read a biology textbook, the content of what you read is what matters. Reading is a form of explicit learning. When you play a video game, the value is in how it makes you think. Video games are an example of collateral learning, which is no less important."

Slumdog brainpower

I found the comment below interesting. Currently I think the effective population of India for producing elite brainpower (i.e., comparing to a reference population with average IQ = 100 and access to first world education and training) is roughly 100 million. For China this number might be 300+ million, but we don't feel their presence in the West as strongly because of language barriers and because of the ability of the Chinese economy to absorb many of the high achievers domestically. (Despite all the Chinese scientists and engineers emigrating to the West there are just as many staying at home to develop a continent-sized economy with a space program, high speed trains, tech and defense industry, etc.) Both the China and India numbers should increase drastically in another generation, although see below for challenges confronting India.


See these related posts, or click the India label below.

More where those came from

IIT uber alles

To have and have not

Shanghai from an Indian perspective

Battle for brainpower

Background: I come from Southern India, high achiever, made a pile as they would say, so broadening my attention to study these subjects to understand the cause of our underdevelopment and what we can do about it. I only have experience with Southern India, where I live and travel extensively; I also get to travel the world extensively so I can give detailed comparisons. While I obviously do want to help my country, I am of the "Face Reality As It Is" persuasion, so I don't worry about political correctness of any discussion here.

First, the good news: thanks to better nutrition, I have witnessed IQ go up substantially in the past 30 years. The typical South Indian immigrant you see in the US today (and they do seem to come mostly from the South these days) comes from a vastly lower social scale than the one you saw 40 years ago. The typical immigrant has parents or grandparents who would not be able to function well in Western society, because they lack the cognitive skills. So the Flynn effect seems very visible to me, from my vantage point in South India.

Second, there are even more gains to be had. Malnutrition is slowly going away in the South, but in the North, it is still very very prevalent - we are talking more than 50% of kids malnourished.

Third, a pacific temperament and a traditional respect for authority keeps lower IQ populations from the pervasive social breakdown you see in the US in similarly situated communities. Families are staying intact; if that changes in India in any large scale, I would write India off, because collectively, we Indians are still too dumb to handle "sophisticated" mating behavior (though I think even the Swedish men stay together with their mates to support their kids, in spite of the promiscuous reputation of their society) - I hope our Hindu Gods literally save us on this one!

Now the bad news. Democracy empowers the dumb, probably true everywhere, but truer in India. The elected need to be skillful at manipulating the dumb, so they tend to be smarter than those electing them. Still, all this manipulating exacts a toll. One of the manipulating they do is jobs-for-dumb-constituents, which begets appallingly bad government, particularly at local levels.

This has all sort of bad effects. With such bad local governance, much of India looks like an ungoverned mess - your senses get physically assaulted in almost any urban space in India, disgusting levels of filth and squalor. This is the result of the utter inability to plan, explained by the IQ of those in charge at local levels.

This is more serious than even that. As India urbanizes, we risk going back on Flynn effect, because poor sanitation leads to *urban* malnutrition - not lack of food, but gastrointestinal illness among kids - which lowers IQ. So not having functional lower levels of government puts India's progress at risk. This, I attribute, directly to democracy.

Given these two opposing forces, which was is it going to go? I believe the good edges out the bad, but that would be a hard case to make when you see the filth.

Only good news I can gather is that London and Hong Kong and so on were extremely filthy at a point in their development too.

Accurate Genomic Prediction Of Human Height

I've been posting preprints on arXiv since its beginning ~25 years ago, and I like to share research results as soon as they are written up. Science functions best through open discussion of new results! After some internal deliberation, my research group decided to post our new paper on genomic prediction of human height on bioRxiv and arXiv.

But the preprint culture is nascent in many areas of science (e.g., biology), and it seems to me that some journals are not yet fully comfortable with the idea. I was pleasantly surprised to learn, just in the last day or two, that most journals now have official policies that allow online distribution of preprints prior to publication. (This has been the case in theoretical physics since before I entered the field!) Let's hope that progress continues.

The work presented below applies ideas from compressed sensing, L1 penalized regression, etc. to genomic prediction. We exploit the phase transition behavior of the LASSO algorithm to construct a good genomic predictor for human height. The results are significant for the following reasons:

We applied novel machine learning methods ("compressed sensing") to ~500k genomes from UK Biobank, resulting in an accurate predictor for human height which uses information from thousands of SNPs.

1. The actual heights of most individuals in our replication tests are within a few cm of their predicted height.

2. The variance captured by the predictor is similar to the estimated GCTA-GREML SNP heritability. Thus, our results resolve the missing heritability problem for common SNPs.

3. Out-of-sample validation on ARIC individuals (a US cohort) shows the predictor works on that population as well. The SNPs activated in the predictor overlap with previous GWAS hits from GIANT.

The scatterplot figure below gives an immediate feel for the accuracy of the predictor.

Accurate Genomic Prediction Of Human Height
(bioRxiv)

Louis Lello, Steven G. Avery, Laurent Tellier, Ana I. Vazquez, Gustavo de los Campos, and Stephen D.H. Hsu

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.

This figure compares predicted and actual height on a validation set of 2000 individuals not used in training: males + females, actual heights (vertical axis) uncorrected for gender. For training we z-score by gender and age (due to Flynn Effect for height). We have also tested validity on a population of US individuals (i.e., out of sample; not from UKBB).

This figure illustrates the phase transition behavior at fixed sample size n and varying penalization lambda.

These are the SNPs activated in the predictor -- about 20k in total, uniformly distributed across all chromosomes; vertical axis is effect size of minor allele:

The big picture implication is that heritable complex traits controlled by thousands of genetic loci can, with enough data and analysis, be predicted from DNA. I expect that with good genotype | phenotype data from a million individuals we could achieve similar success with cognitive ability. We've also analyzed the sample size requirements for disease risk prediction, and they are similar (i.e., ~100 times sparsity of the effects vector; so ~100k cases + controls for a condition affected by ~1000 loci).


Note Added: Further comments in response to various questions about the paper.

1) We have tested the predictor on other ethnic groups and there is an (expected) decrease in correlation that is roughly proportional to the "genetic distance" between the test population and the white/British training population. This is likely due to different LD structure (SNP correlations) in different populations. A SNP which tags the true causal genetic variation in the Euro population may not be a good tag in, e.g., the Chinese population. We may report more on this in the future. Note, despite the reduction in power our predictor still captures more height variance than any other existing model for S. Asians, Chinese, Africans, etc.

2) We did not explore the biology of the activated SNPs because that is not our expertise. GWAS hits found by SSGAC, GIANT, etc. have already been connected to biological processes such as neuronal growth, bone development, etc. Plenty of follow up work remains to be done on the SNPs we discovered.

3) Our initial reduction of candidate SNPs to the top 50k or 100k is simply to save computational resources. The L1 algorithms can handle much larger values of p, but keeping all of those SNPs in the calculation is extremely expensive in CPU time, memory, etc. We tested computational cost vs benefit in improved prediction from including more (>100k) candidate SNPs in the initial cut but found it unfavorable. (Note, we also had a reasonable prior that ~10k SNPs would capture most of the predictive power.)

4) We will have more to say about nonlinear effects, additional out-of-sample tests, other phenotypes, etc. in future work.

5) Perhaps most importantly, we have a useful theoretical framework (compressed sensing) within which to think about complex trait prediction. We can make quantitative estimates for the sample size required to "solve" a particular trait.

I leave you with some remarks from Francis Crick:

Crick had to adjust from the "elegance and deep simplicity" of physics to the "elaborate chemical mechanisms that natural selection had evolved over billions of years." He described this transition as, "almost as if one had to be born again." According to Crick, the experience of learning physics had taught him something important — hubris — and the conviction that since physics was already a success, great advances should also be possible in other sciences such as biology. Crick felt that this attitude encouraged him to be more daring than typical biologists who tended to concern themselves with the daunting problems of biology and not the past successes of physics.

Sven Magnus Øen Carlsen

I am curious about what serious chess players think of Elo scores across generations -- is there a Flynn Effect? Current world champ Magnus Carlsen has the highest score of all time. It seems possible that due to computer training today's players are much better than those of the past -- we'll never know how giants like Bobby Fischer would have developed using today's methods. By Elo score Fischer is only 14th on the all-time list!

I thought interest in chess would decline after Deep Blue defeated Kasparov. But apparently the game is still thriving.

The excerpts below are from an old interview in Der Spiegel:

Carlsen: ... my success mainly has to do with the fact that I had the opportunity to learn more, more quickly. It has become easier to get hold of information. The players from the Soviet Union used to be at a huge advantage; in Moscow they had access to vast archives, with countless games carefully recorded on index cards. Nowadays anyone can buy this data on DVD for 150 euros; one disk holds 4.5 million games. There are also more books than there used to be. And then of course I started working with a computer earlier than Vladimir Kramnik or Viswanathan Anand.

SPIEGEL: When exactly?

Carlsen: I was eleven or twelve. I used the computer to prepare for tournaments, and I played on the Internet. Nowadays, children start using a computer at an even earlier age; they are already learning the rules on screen. In that sense I am already old-fashioned. Technological progress leads to younger and younger top players, everywhere in the world.

SPIEGEL: Is being young an advantage in modern chess?

Carlsen: As a young player you have a lot of energy, a lot of strength, you are very motivated. But young players are often not good at defending a position; they cannot cope well when fate turns against them. The fact is simply that experience is a central issue. One of the most important things in chess is pattern recognition: the ability to recognise typical themes and images on the board, characteristics of a position and their consequences. To a certain degree you can learn that while training, but there is nothing like playing routine. I have always made sure to get that. I am only 19, but I have certainly already played a thousand games in the classic style.

This is also interesting:

SPIEGEL: Mr Carlsen, what is your IQ?

Carlsen: I have no idea. I wouldn’t want to know it anyway. It might turn out to be a nasty surprise.

SPIEGEL: Why? You are 19 years old and ranked the number one chess player in the world. You must be incredibly clever.

Carlsen: And that’s precisely what would be terrible. Of course it is important for a chess player to be able to concentrate well, but being too intelligent can also be a burden. It can get in your way. I am convinced that the reason the Englishman John Nunn never became world champion is that he is too clever for that.

SPIEGEL: How that?

Carlsen: At the age of 15, Nunn started studying mathematics in Oxford; he was the youngest student in the last 500 years, and at 23 he did a PhD in algebraic topology. He has so incredibly much in his head. Simply too much. His enormous powers of understanding and his constant thirst for knowledge distracted him from chess.

SPIEGEL: Things are different in your case?

Carlsen: Right. I am a totally normal guy. My father is considerably more intelligent than I am.

Der Spiegel carefully tested Kasparov's IQ at 135, as discussed here. See also The Laskers and the Go master.

Heritability 2.0

Current best estimates of the heritability of g come from twin and adoption studies. The table below, from the recent paper: Molecular Psychiatry (2010) 15, 1112, gives you an idea of the general consistency of results from a number of twins studies with large statistics.



But now that we have inexpensive genotyping, we can study heritability of a quantitative trait by looking at unrelated (or only distantly related) individuals, and asking to what extent similarity in genotype is correlated with similarity in phenotype. A simple way to think about this is to imagine that we have a sample of N people for whom both phenotype (measured g score) and genotype (e.g., SNP profile) are known. Form all possible pairs and plot magnitude of difference in g score against genetic distance between the individuals in the pair. The g score difference should (on average) decrease as the genetic distance goes to zero (at which point the pair are MZ twins; but we avoid the confound of shared prenatal environment). Even if we have no identical twins in the sample, and even if none of the people in the sample are closely related to each other, we can extrapolate to zero genetic distance to obtain an estimate of heritability. An analysis along these lines (more technically, a global fit across all SNPs of total heritability) for height yields a result which is consistent with the narrow sense heritability estimate from twin and adoption studies. The results for g have not yet been published, but rumor has it that they also support earlier estimates such as those given above.

With this new technique one can average over a much larger range of environments. (As we know, heritability is only defined with respect to a specific range of environments.) One can form the sample from individuals who grew up in very deprived or very privileged families. One can even compare cohorts born in very different eras -- as long as DNA samples are available. For example, we can compare people born 80 years ago (i.e., who are still living and whose early adult IQs are known from military or school records) to each other, or even to people born as little as 15 or 20 years ago (assuming raw scores or cross norms are available). This kind of analysis may reveal something interesting about the Flynn Effect.

Update: The paper I referred to is out. It yields a lower bound on narrow sense heritability which is consistent with the earlier twin and adoption studies.

Genome-wide association studies establish that human intelligence is highly heritable and polygenic.

Abstract
General intelligence is an important human quantitative trait that accounts for much of the variation in diverse cognitive abilities. Individual differences in intelligence are strongly associated with many important life outcomes, including educational and occupational attainments, income, health and lifespan. Data from twin and family studies are consistent with a high heritability of intelligence, but this inference has been controversial. We conducted a genome-wide analysis of 3511 unrelated adults with data on 549,692 single nucleotide polymorphisms (SNPs) and detailed phenotypes on cognitive traits. We estimate that 40% of the variation in crystallized-type intelligence and 51% of the variation in fluid-type intelligence between individuals is accounted for by linkage disequilibrium between genotyped common SNP markers and unknown causal variants. These estimates provide lower bounds for the narrow-sense heritability of the traits. We partitioned genetic variation on individual chromosomes and found that, on average, longer chromosomes explain more variation. Finally, using just SNP data we predicted ∼1% of the variance of crystallized and fluid cognitive phenotypes in an independent sample (P=0.009 and 0.028, respectively). Our results unequivocally confirm that a substantial proportion of individual differences in human intelligence is due to genetic variation, and are consistent with many genes of small effects underlying the additive genetic influences on intelligence.

Annals of psychometry: adult cognitive skills by country

You may have read (NYTimes; enjoy the 500 expert comments!) about the recent OECD study of adult skills, which showed Americans lagging behind most other advanced countries. The outcomes more or less recapitulate the PISA results, which are produced by the same organization, led by Andreas Schleicher (Without data, you are just another person with an opinion; Schleicher was trained in physics before moving into educational assessment). Studying adults rather than students makes generational differences apparent: young Koreans scored near the top in the world, whereas older Koreans, who grew up in a much poorer country, do less well. This was also true for Finns, but countries like the UK, US and Australia showed almost no difference between the young and old. It would appear that the Flynn Effect has abated for the current adult population in those countries.

The most striking findings in the report to someone already familiar with this kind of country level data, are that:

(I) Job growth in areas requiring superior ability is high, whereas there is almost no growth in other sectors (see figure below and figure 1.3 in the report). The biggest decline in jobs is in manufacturing-related areas, which are characterized by low, but not the lowest, scores. The number of menial jobs, characterized by the lowest scores, held their own. In other words, in developed countries jobs for factory workers declined while maids, gardeners and servants held their own... a grim future of income inequality based on cognitive ability?

(Difference between mean scores of skilled and "elementary" occupations is about 1 population SD -- Table A3.19. This is about the same as mean score difference between college graduates and those that did not reach upper secondary education! Table A3.9 (l))

(II) Scores on numeracy and literacy showed high correlations: 0.8 -- 0.9 in every country. This is much higher than, for instance, M and V correlation on the SAT.

Here are national score ranges for math skills:

This figure shows that Japanese high school graduates have better literacy skills than Italian college graduates. There is plenty more data like this. Interestingly the standard deviations in scores for Japan and Korea are on the low end compared to other countries.

Roughly every fifth Finn and Japanese reads at high levels (Level 4 or 5 on the Survey of Adult Skills). This means, for example, that they can perform multiple-step operations to integrate, interpret, or synthesise information from complex or lengthy texts that involve conditional and/or competing information; and they can make complex inferences and appropriately apply background knowledge as well as interpret or evaluate subtle truth claims or arguments. They are also good at numbers: they can analyse and engage in complex reasoning about quantities and data, statistics and chance, spatial relationships, change, proportions and formulae; perform tasks involving multiple steps and select appropriate problem-solving strategies and processes; and understand arguments and communicate well-reasoned explanations for answers or choices.

How does someone who can't do these things obtain a good college degree? If only 20% of Finns and Japanese can do these things (the largest fraction among all nations), what fraction of US students are ready for serious college work?

Swedish height in the 20th century

Average height of Swedish military conscripts during the 20th century. Looks like an increase of roughly 1 cm per decade or about 1.5 SD in one century. Not unlike the Flynn Effect over the same period.

It was important to generate this curve so that we could properly z score heights of individuals from different birth cohorts. Thanks to my collaborator Laurent Tellier.

Higher education and human capital

What good is higher education? The conventional view is that, in addition to producing a well-informed citizenry, it builds important human capital and raises national productivity. But what is the evidence for these assertions? In policy debates we are typically presented with faulty logic: workers in desirable, high value-added jobs (e.g., at Google or Biogen) tend to have lots of education. Therefore, if we want Americans to have such jobs we had better expand access to higher education. The counter argument, that returns to society as a whole from education diminish as access increases beyond the cognitive elite, is given below by a well-known curmudgeon and psychometric realist:

Brutal, just brutal: ...There is no magic point at which a genuine college-level education becomes an option, but anything below an IQ of 110 is problematic. If you want to do well, you should have an IQ of 115 or higher. Put another way, it makes sense for only about 15% of the population, 25% if one stretches it, to get a college education. And yet more than 45% of recent high school graduates enroll in four-year colleges. Adjust that percentage to account for high-school dropouts, and more than 40% of all persons in their late teens are trying to go to a four-year college--enough people to absorb everyone down through an IQ of 104.

Note the claim is not that benefits from higher education are zero for the average student, but merely that they diminish significantly as we expand access. At some point we need to consider whether the marginal cost exceeds the marginal benefit. No amount of schooling will turn an average student into a materials engineer, tax lawyer or derivatives trader.

I'm afraid these kinds of thoughts lurk in the minds of most professors these days -- I've heard them discussed many times. Why can't my students write? Why can't my students do simple math? Does the bottom half of the class really absorb anything from my lectures? Is science just too difficult for some people? If I showed you some of the emails I receive from students in my physics 101 course, you would cry at the lack of mastery of grammar and spelling, let alone physics.

Below I excerpt some depressing results from researchers at Stanford and Yale, which support the sorting and signalling model of higher ed, rather than the human capital building model.

Education and Verbal Ability over Time: Evidence from Three Multi-Time Sources

Nie, Golde and Butler

Abstract: During the 20th century, there was an unprecedented expansion in the level of educational attainment in America. Using three separate measures, this paper investigates whether there was a concurrent increase in verbal ability and skills. Changes in verbal ability in the general population as well as changes in the verbal ability of graduates of different levels of education are investigated. An additional investigation of how changes in the differences between males' and females' educational attainment are associated with changes in differences between their respective verbal abilities follows. The main finding is that there is little evidence that the large increase in educational attainment has resulted in an increase in any of the measures of verbal abilities and skills.

From the paper:

The results from using these three different measures of verbal ability and skills all show the same striking patterns: (1) there is no increase in scores in the overall population over time; (2) as the number of people obtaining a certain level of education increased, the verbal ability of those terminating with that degree has decreased. ...

Comment for the psychometric cognoscenti: where is the Flynn effect here? I see no overall increase in verbal IQ.






See also this less technical exposition:

Nie and Golde: ...Our initial hypothesis was that if amount of schooling causally affects any outcome, it would be verbal ability. The vast expansion of the American education system over the course of the 20th century served as our test bed. We expected that the huge increase in educational attainment in the U.S. across the decades would be accompanied by a substantial improvement in verbal abilities. To our initial amazement, we found no evidence for such improvement.

We started our investigation by showing that there is, indeed, a strong correlation between education and verbal ability. The data on which our analyses are based came from the General Social Survey, a program of in-person interviews that has been conducted regularly since 1972 by the National Opinion Research Center at the University of Chicago. While the samples were nationally representative, to avoid complications caused by changing demographics and questions about the validity of such tests with minority and immigrant populations, we included only the native-born, white American population 30 to 65 years of age, using information collected over the last 35 years of parallel surveys. (We used only those 30 years or older to ensure that we were dealing only with people who had completed their education; we stopped at age 65, lest we contaminate the analysis by differential mortality rates.)

Education levels and scores on a vocabulary test given to subjects are indeed correlated (see Figure 1). Over the three-plus decades studied, those with more education got better vocabulary scores, and vice versa.

Those results, however, do not necessarily imply that education causes increased verbal ability. If education did increase verbal ability, we would expect increasing levels of education over time to bring about measurably higher levels of verbal ability. During the 20th century, there was an unprecedented expansion in the levels of educational attainment in the U.S. The average American born between 1910 and 1914 received a bit more than 10 years of education. The average American born between 1970 and 1974 received 14 years of education. In 60 years, the "average American" went from being a high school dropout to having two years of college — a remarkable increase. The increase in education is across the board. A person born between 1910 and 1914 who obtained some postgraduate education was in the top 6 percent of his or her cohort in terms of education. By the 1970s, nearly 16 percent of the birth cohort had some postgraduate education. The percentage of college graduates or beyond has almost quadrupled over the same period, from just over 10 percent to almost 40 percent.

But, as Figure 2 shows, even though education has increased considerably through the decades, and even though education is correlated with verbal ability, verbal ability has stayed practically constant over time. The lack of change in the average vocabulary score of Americans, despite the large increase in the population's average years of schooling, is an intriguing finding. ...