Tuesday, 30 December 2014

Five Gold Rings (inherited)

On the Fifth day of Christmas tradition demands that you send your true love Five Gold Rings. Even with a small circle of readers, this could prove expensive. The next best thing is to send five special findings regarding genetics and intelligence differences, as an end of year gift.

(1) The heritability of intelligence increases from about 20% in infancy to perhaps 80% in later adulthood. This seems to be due to a process of genetic magnification, perhaps through the creation of personal micro-environments.

(2) Intelligence captures genetic effects on diverse cognitive and learning abilities, which correlate phenotypically about 0.30 on average but correlate genetically
about 0.60 or higher.  (Genetic correlations estimate the extent to which genetic effects on one trait are correlated with genetic effects on another trait independently of the heritabilities of the two traits. They can be thought about roughly as the probability that genes associated with one trait are also associated with the other trait. Genetic correlations are derived from the genetic analysis of covariance between traits using the same quantitative genetic methods used to analyse variance.) Multivariate genetic research—both from twin studies and GCTA—suggests that most of the genetic action is general across diverse cognitive abilities rather than specific to each ability. Intelligence is a good target for gene-hunting
because it indexes these generalist genes.

(3) Assortative mating is greater for intelligence (spouse correlations ~0.40) than for other behavioural traits such as personality and psychopathology (~0.10) or physical traits such as height and weight (~0.20). Assortative mating pumps additive genetic variance into the population every generation, contributing to the high narrow heritability (additive genetic variance) of intelligence. Verbal intelligence shows greater assortative mating (~0.50) than nonverbal intelligence (~0.30), perhaps because it is easier to gauge someone’s verbal ability such as vocabulary than their nonverbal intelligence such as spatial ability. Assortative mating for intelligence is caused by initial selection of a mate (assortment) rather than by couples becoming more similar to each other after living together (convergence). In part, spouses select each other for intelligence on the basis of education—spouses correlate about 0.60 for years of education19—which correlates about 0.45 with intelligence. Assortative mating may be greater than it is for intelligence for a few other traits such as social attitudes, smoking and drinking, although these traits might be affected by
convergence. Assortative mating increases additive genetic variance in that the offspring differ more from the average than they would if mating were random. The increase in additive genetic variance can be substantial because its effects accumulate generation after generation until an equilibrium is reached. For example, if the heritability of intelligence with random mating were 0.40, the additive genetic variance of intelligence would increase by one-quarter at equilibrium given assortative mating of 0.40, Falconer and MacKay (1996) equation 5, Table 10.6, p. 176.

(4) Unlike psychiatric disorders, intelligence is normally distributed with a positive end of exceptional performance that is a model for ‘positive genetics’. Mind you, although the authors do not say so, some people are probably extremely positive in their behaviours: calm and kind and helpful. It is simply that no-one has particularly looked for them, because they are not a nuisance and do not call attention to themselves.

(5) Intelligence is associated with education and social class and broadens the causal perspectives on how these three inter-correlated variables contribute to social mobility, and health, illness and mortality differences. Strange as it may seem, social class and education are also heritable.

The authors propose that three particular sets of findings are so common in genetic research that they could almost be considered laws: all traits are heritable, all traits show environmental effects, and the heritability of traits is brought about by many genes of small effects. Remembering those three observation will put you well ahead of most commentators when discussing the genetics of intelligence.

The authors describe the new technique of Genome-wide Complex Trait Analysis thus:

However, instead of using genetic similarity from groups differing markedly in genetic similarity such as monozygotic and dizygotic twins, GCTA uses genetic similarity for each pair of unrelated individuals based on that pair’s overall similarity across hundreds of thousands of single nucleotide polymorphisms (SNPs) for thousands of individuals; each pair’s genetic similarity is then used to predict their phenotypic similarity. Even remotely related pairs of individuals (genetic similarity greater than 0.025, which represents fifth-degree relatives) are excluded so that chance genetic similarity is used as a random effect in a linear mixed model. The power of the method comes from comparing not just two groups like monozygotic and dizygotic twins, but from the millions of pair-by-pair comparisons in samples of thousands of individuals. In contrast to the twin design, which only requires a few hundred pairs of twins to estimate moderate heritability, GCTA requires samples of thousands of individuals because the method attempts to extract a small signal of genetic similarity from the noise of hundreds of thousands of SNPs.

In my non-genetic language, this is like hunting through thousands of coded messages in order to find similarities between messages by looking at clumps of individual variations within the common code. Its great power, to me, is that it does not depend the specific instance of twins (seen as unrepresentative by critics of genetics) but on relative similarity on a very broad dimension of genetic relatedness.

Have a look at this excellent expert review: Genetics and intelligence differences: five special findings. R Plomin and IJ Deary.  Molecular Psychiatry (2014), 1–11


The review is written by the two persons in the world best able to explain modern research in genetics and intelligence. It is an excellent summary of the current state of play in the genetics of intelligence. Although most science research is incremental, and scientific “breakthroughs” happen mostly in news rooms when there is nothing much else to print, it might be that 2015 will provide a significant increase in understanding how a genome builds the problem solving brain. Although no one can accurately determine the size of the task, and it is generally thought a very complex problem, sample sizes are getting much bigger, genomic chips cheaper and more powerful, and the raw data processing power needed to interpret the results is increasing steadily.

Happy New Year!

Sunday, 21 December 2014

Christmas Carols 2014

As is the habit of my tribe, along the valley to the 13th Century Church of the Knights Hospitaller of Jerusalem we went again, in that Doomsday village that was home to Waleran the Hunter in Saxon times and, after the Norman conquest, Payne de Turberville (mentioned with transliteration by Thomas Hardy in Tess of the d’Urbevilles) for traditional Christmas carols and lessons. All the lesson were well read, with pride of place to an enthusiastic 8 year old boy, who looked as if he enjoyed spreading good news. “He practised several times yesterday” his proud father confided afterwards, latter adding that he lived in Winchester, which was the only town not in Doomsday, because it was written there “and if you wanted to know how many houses it had, you looked out of the window”.

In keeping with tradition, the opening soprano solo of “Once in Royal David’s City” was sung with the singer hiding modestly behind the South transept. Last year’s wantonness of having the singer face the audience had been an aberration to be avoided, it appears.

The Choir were 10 strong, and sang “The Lord at first did Adam make” and then later the canticle “Nunc Dimittis” in sparkling unaccompanied plainsong: 

Lord, now lettest thou thy servant depart in peace : according to thy word.
For mine eyes have seen : thy salvation,
Which thou hast prepared : before the face of all people;
To be a light to lighten the Gentiles : and to be the glory of thy people Israel.

Then four of them took up their instruments again (violin, flute, oboe, and cello) and the service resumed its Carol progress.

In strict conformity with stochastic tradition, the organist adventitiously omitted the culminating last verse of one carol, leading to the usual surprise and pre-orgasmic disappointment among those who had saved their breath to the last and had intended to launch into a transcendental descant.

At the end the priest intoned: Lastly let us remember before God all those who rejoice with us, but upon another shore and in a greater light, that multitude which no man can number, whose hope was in the Word made flesh, and with whom we for evermore are one.

An image of my parents came unbidden to my mind. They knew their psychology, the ancient scribes: how well soothing words play to ever-present anxieties, and the certainties of incanted repetition and confident blithe optimism.

Then the contemporary Eucharist: red wine and mince pies, conversations with choristers; the new Lords of the Manor who arrived on Friday paying their respects to all and sundry, and being introduced to the former Lords of the Manor who still return for the Christmas services;  some forty five souls chattering in the tiny church, with a charitable truce even among estranged neighbours, and then out down the winding path past the chest tombs into a grey and wintry day.

Depart in peace.

Merry Christmas.

Saturday, 20 December 2014

Memory span

For some reason, the Powers that Be imagine you will be out shopping this Saturday, little comprehending that you have eschewed such conspicuous consumption for the pleasures of staying at home and reading about digit span. In one of Pliny the Younger’s letters (you will know the precise reference, dear reader) he writes an aside after one of his more philosophical reflections, saying that only he and the esteemed recipient of his missive would be interested in such things, since at the moment of writing the town was deserted because everyone was at the races. Such are the sacrifices and soul hardening consequences of disinterested intellectual curiosity.

At ISIR2014 Gilles Gignac wondered whether memory span was the ugly duckling of intellectual assessment. This is familiar ground to readers of this blog, so I will only summarise the main points.

Although memory span subtests have been included in well-known intellectual assessment batteries for over 100 years, these subtests have been (and continue to be) regarded as the worst indicators of cognitive ability. For example, Wechsler (1939; 1958) contented that beyond an absolute minimum, there were no
intellectual benefits to a greater memory span. By contrast, in the area of cognitive science, individual differences in memory span, and working memory in particular, have been regarded with great regard. In fact, working memory has been suggested to be the fundamental basis of intellectual functioning. In this talk, the large gap between these two positions will be attempted to be narrowed, based on the key results associated with three studies.
Method: The first and second studies were based on the WAIS-IV normative sample data (N = 2,200). The second study was based on archival data dating from 1923 to 2008 and amounted to a total approximate N = 7,000. In the first two investigations, the analyses consisted principally of a combination of linear and nonlinear bifactor modeling. The third investigation involved the plotting of Digit Span Forward and Digit Span Backward normative sample means across time.
Results and conclusions: Memory Span, as measured by Longest Digit Span Forward, Longest Digit Span Backward, and Longest Digit Span Sequencing, evidenced nearly average loadings on the general factor (g), suggesting that they are decent indicators of g. In the second investigation, the association between memory span and FSIQ (as well as g) was observed to be nearly completely linear, such that every extra level of memory span corresponded to an increase of approximately 4 FSIQ points. Finally, based on the archival investigation, both Digit Span Forward and Digit Span Backward test scores were found to be completely resistant to the Flynn Effect across 85 years of data.
Discussion: This is the first investigation to examine and demonstrate that all three memory span subtests within the WAIS-IV are moderately good indicators of g. Furthermore, the fact that the association between the three memory span measures and g is largely linear underscores their assessment value at all levels of ability. Finally, memory span test scores are possibly the only test scores not to evidence any susceptibility to the Flynn effect, again, highlighting their attractiveness as indicators of cognitive ability.


Conference News: another author has re-analysed the results, and comes to a somewhat different conclusion about Flynn effects.  I am waiting for that paper to be accepted, and have forewarned Gignac that it is coming, in the hope of getting an advance copy of his eventual reply. You will be the first to hear the news, so long as you keep tuned to this channel.

Wednesday, 17 December 2014

Are women easily discouraged?


I cannot pry into your domestic arrangements, not without a larger budget anyway, but I wonder if you would agree that women are easily discouraged? Personally, I have not found it to be the case. On the contrary, to me women seem to have a well-developed ability to triumph in argument by the vigorous exposition of their cause, and a photographic recollection of such errors and transgressions committed by any man who would dare oppose them. This may be a unique observation. You will tell me. Methodologically, I understand that your perspective may differ if you are a woman or a man, but I trust you will follow the usual calm and detached empirical procedures to ensure an objective evaluation of this contentious matter.

However, my personal impressions may be due to restrictions in sample size. On the basis of studying larger samples of women with social psychology experiments on priming some researchers assert that if you communicate to a woman that she is about to take a test on which women do badly, then she will do badly on that test. That is, worse than she would have done if you had not communicated that negative expectation. This is called “stereotype threat” and is said to be the reason that women and some racial groups do badly on some tasks.

First, let us see if this is true. Paulette Flore works with Jelte Wicherts, who likes nothing more than taking a chainsaw to other people’s statistics, often with good effect. Will stereotype threat survive the investigations of this dynamic duo? Paulette gave an enthusiastic presentation on this subject at ISIR2014 in Graz.

Does stereotype threat influence test performance of girls in stereotyped domains? A metaanalysis. Flore, P.C., Wicherts, J.M.    

P.C.Flore@tilburguniversity.edu; J.M.Wicherts@tilburguniversity.edu


The effect of gender stereotype threat on math, science or spatial skills tests performance of girls has been studied in numerous experiments. Although theory predicts that girls in stereotype threat conditions will underperform compared to girls in control conditions, outcomes of the experiments strongly vary. In order to
understand the effect of stereotype threat in school-aged groups, we conducted a meta-analysis of stereotype threat experiments on test performance with schoolgirls as participants. In addition we considered the following theoretically relevant moderators: test difficulty, presence of boys, gender equality within countries, and the type of control group used in the experiments. We also studied the possibility of publication bias.

We carried out an extensive (grey) literature search. Selected papers included a sample of girls in their study, had a sample with mean age younger than 18 years, used a (quasi)-experimental design, administered the stereotype threat between-subjects, and used a math, science or spatial skills test as dependent variable. We used Hedges’ g as the effect size and fitted random and mixed effects meta-regressions with restricted maximum likelihood estimation. We carried out several tests for publication bias and excess of significant results.




We obtained 26 relevant papers or unpublished reports, out of which we selected 47
independent effect sizes for the analysis (total N= 3760). The estimated average effect size was small, but significant (Hedges’ g = -0.22). None of the theoretical moderators significantly predicted the stereotype threat effect. The tests for publication bias were all significant. Furthermore, the test of the excess of significance results indicated there were more significant studies included in the meta-analysis than to be expected based on the observed power. A consequence of publication bias, or an excess of significant results, is that the average estimated effect size is probably inflated.

Although the theory of stereotype threat has been well established, based on our results we question the robustness of the effect for the population of school-aged girls. To obtain a more reliable estimate of the average effect of stereotype threat we need new cross culturally administered experiments with sufficient power, which would enable us to disentangle the bias from the actual effects of stereotype threat.

Why did papers which showed an apparent effect of stereotype threat have a better chance of getting published? Of course, papers reporting any sort of effect are more attractive than papers which fail to find an effect, though it is important to know about both.  Do editors and reviewers want to find that women are underperforming because of an unfair performance expectation? Have they thought that through?

Let us do a thought experiment. Say you are an employer. (I know this is unlikely, and that probably like so many of us you are now or have been in the past a hireling of the State or their cosmetically enhanced front organisations and running dogs). You are told that if you employ a woman (or an ethnic minority person) they are likely to underperform when given a task on which they are told they will do badly. I am not an employer, but I have listened to employers, and the last thing they want is to hire someone who is “high maintenance”. Imagine you were an employer and one day, coming across a rival product being offered for sale, you rushed into your own office and said to your employees: “Our competitors have come up with a brilliant innovation. We have to use our brains to work out how to improve our own offering, and we have to do so quickly so as to outwit them with our intelligent counter-offer”. At this point you notice all the women are in tears, and all the ethnic minority people are mute with anxiety. From the employers’ point of view, hiring such fragile minds is a dreadful selection error. Never employ people who collapse under pressure and, most certainly, never let them fly an aeroplane in a thunderstorm.

It many be too much to expect, even after this thorough dismissal by Flore and Wicherts, that the notion of stereotype threat will not be pursued further, but brace yourself for further publications purporting to show that girls simply can’t take the pressure of negative expectations, and need to be guided through their fragility by sensitive condescension.

Tuesday, 16 December 2014

Are you a nuisance?

I can only guess, but I assume that my readers sometimes reflect on their achievements, and within the bounds of modesty assume they have contributed, however mildly, to the societies in which they live. Frontiers of science, alleviation of suffering, careful driving, scrupulous payment of invoices: that sort of thing. Of course, such self-assessments are often delusional. Memory can be selective, and the occasional important publication shines in recollection, while the large pile of unfinished projects, disgruntled colleagues and abortive grand designs fade into oblivion. Pereunt et imputantur.

What if we were to take an objective measure? Track a thousand newborns, and keep a close account of the profit and loss ledger. At this point you may feel a trifle uneasy. Who are we to judge these matters? What price the jocular remark of a mute inglorious Milton? How could one possibly assess the wit of someone who lacks a Twitter account? Furthermore, you may recoil at the possible results of such an enquiry. If some individuals turn out to be a nuisance and a high cost to society, what then? Should they be exiled to some other land whether the natives are even more generous and gullible, or should we intervene as best we can to make them into productive citizens? These are not trivial matters, and the researchers were at pains to highlight the moral choices which arise from a clear headed evaluation of costs and benefits. In particular, their discussion pre-supposes a compassionate society, with redistributive taxation providing educational, health and welfare benefits. The question barely arises outside a welfare states. In such less kindly states, if people are a nuisance they are simply a nuisance, but not a direct cost, since no one will be paying them any benefits.

Terrie E. Moffitt & Avshalom Caspi used the ISIR 2014 conference to test reactions from assembled researchers about the findings so far, and about the issues which arise from them. They presented their data on the Dunedin study, a four-decade longitudinal study of a birth cohort of 1000 New Zealanders. They examined risk factors in childhood and measures of social, health, and economic costs in adulthood.

Adult social and economic outcomes fit the Pareto principle: 20% of the cohort accounted for approximately 80% of every outcome: the cohort’s months of social welfare benefits, years of absent-father childrearing, pack-years of cigarette smoking, hospital admissions, pharmacy prescription fills, criminal court convictions, and injury-related insurance claims. Moreover, high-cost individuals with one problem outcome tended to also have multiple problem outcomes. An ultra-high-cost sub-segment of the cohort was identified who accounted for 80% of multiple problems.

I can remember my interest in the Pareto principle when I first came across it, but I now see it as part of what I call “the comparative percentages muddle”. For example, would you be outraged to hear that 90% of national acne is owned by 10% of the population? A moment’s thought will show you that some people have acne and others do not. Comparative percentage obscures a skewed distribution. Teenagers tend to have acne, and some young adults have acne which reaches chronic levels. Then take the more usual diatribe: the top 1% own 10% or 30% or whatever of the national wealth. In the same way that it seemed odd that some some small percentage “own” all the acne it seem iniquitous that another small percentage owns a large percentage of the wealth. The comparative percentage muddle is based on the untested assumption that 1% of the population should own no more than 1% of the wealth. Comparative percentage shares are a clumsy (and perhaps intentionally misleading) way of showing distributions. For example, in a country where every citizen is paid the same wage for 40 years it will still be the case that older workers will have  more savings than the young because they will have had more years in which to build up savings, even in a country where you cannot pass on your wealth to your children. Savings accumulate over the life course, so without age correction the comparative percentage wealth statistic is misleading. Add in compound interest on savings, and add in a mild wage differential for more educated workers and the whole thing becomes a muddle in search of indignation.

The authors know all this, and realise that the beguiling Pareto observation is a post-hoc description, which of itself predicts nothing. In this case it simply asserts: there are some troublesome people, and they will account for most social problems. The critical question is: which kids will grow up to be responsible for a disproportionate amount of trouble (and can anything be done to make them behave better)?

The authors say: Risk factors measured in childhood that characterized this ultra-high-cost group were: low family socio-economic status, child maltreatment, low self-control, and low IQ. Effect sizes were very large. Predictive analyses showed that together, SES, maltreatment, self-control, and IQ measured in the first decade of life were able to predict 80% of the individuals who are using 80% of multiple costly services. We developed an index of the integrity of a child’s brain at age three years. This age-3 brain-integrity index was a strong predictor of the cohort members who four decades later became members of the ultra-high-cost population segment.
Implications: Much research has shown that childhood risk ‘X’ can predict poor adult outcome ‘Y’, but modest effect sizes discourage translation of findings into targeted childhood interventions. This study illustrates that the vast bulk of a nation’s social services, crime control, and health-care are expended on a relatively small population segment. During early childhood, this population segment is characterized by a small set of risk factors: low SES, child maltreatment, low self-control, low IQ, and poor brain integrity. Reducing these factors may bring surprisingly good return on investment.

The comments from the audience were that it would be an error to describe the neurological examination as an “index of the integrity of the child’s brain”. Brains are assumed to be present. Better to say that an examination of behaviour, skills and neurological responses shows that many of the troublesome children can be detected at that age.

The assessment is interesting. It includes the Peabody Picture Vocabulary Test, which is simple and a good predictor. A word is spoken and the child has only to point to the one of four pictures which best describes the word. It has been doing good work since 1959 and is an excellent example of the power of intelligence measures: simple to do but profound in their implications. They also tested language and motor development, and simplest of all, what the child’s behaviour had been like during the 90 minute session.

There were varying views as to whether interventions in early childhood would be effective. I think the Abecedarian project achieved useful results in increasing ability somewhat (by about 4 IQ points on average), but not all researchers are convinced about that. Training parents in how to manage a very difficult and demanding child could be very useful, but that remains to be proved at the scale required, though King’s College has done good pioneering work on this.

The core of the argument is a social one, and goes to the heart of policy making. The authors calculate that about 45% of the population are “low cost users”. In other words, they draw very little on community resources, yet pay most of the taxes that provide those services to others. The authors have identified some ultra high cost users who are a net drain on resources. Compassionate societies pay their bills, including the highest bill, which is often being the target of their bad behaviours. Whilst they remain willing to do so welfare states will survive, and might even invest yet more resources in the hope of bringing about improvements in parenting. If they were to decide to invest all of their resources in their own children rather than divert their earnings to other people’s children the redistributive State would collapse.

Sunday, 14 December 2014

#IQ2014 True but taboo

Excellent talk from Susan Pinker on the difficulties of talking to the press about intelligence. She was able to very quickly explain that scientists often have great difficulty understanding that their interest in whether something is true or not does not meet the basic requirements of dealing with the public’s strong emotions regarding taboo subjects. Researchers cannot understand that a good story (true or not) always trumps their most earnest recital of reliable facts. The wisdom of crowds crowds out even the best statistics. You cannot expect to connect with an audience who are in the throes of righteous indignation by reading them a list of met-analytic effect sizes.

Pinker looked back at Hume’s theory of moral sentiments: Reason alone cannot be a motive to the will, but rather is the “slave of the passions”. Moral distinctions are not derived from reason. Moral distinctions are derived from the moral sentiments: feelings of approval (esteem, praise) and disapproval (blame) felt by spectators who contemplate a character trait or action.

By way of analogy, Pinker argues that the elephant in the room is not individual and group differences in intelligence, but the immense emotions which surround the subject, and reason is no more than the little boy riding the elephant.

Why should any member of the public listen to a treatise on intelligence testing when they are overcome with repugnance at, for example, the use of IQ tests to determine who can be subjected to capital punishment in America, and hold you responsible like a corrupt doctor attending a torture session? Unless emotions are understood, communication about facts is unlikely. Where there have been historical injuries these should be talked about, or it will seem that the researcher knows but does not care that abuses took place.

The next points apply even to non-taboo subjects. Experts are cursed with knowledge, often too much and of the wrong sort (too detailed, too complex, of marginal interest to most people). Jargon is a real problem, as is the lack of connection between a finding and everyday life, which may make researcher look utterly cold and heartless (particularly if it involves explaining to the journalist that they cannot boost their IQ).

It is essential to know your audience. Lecturing about sex differences, Pinker fell foul of a particular German audience who assumed that all science was a social construction of oppressive patriarchs. Forewarned, she would have discussed this view at the very beginning, and explained that she was describing the world as it is, not as it ought to be. She imagined that it would be clear to the audience, as an independent-minded professional woman who combined clinical psychology practice with science journalism, that she was not a stooge or a push-over as regards male domination.

Pinker’s one plea was “prepare yourself for talking to a journalist as you would for a job interview”. Understand your audience, and their interests and background before saying a word. Aim to make three points. If you want to say more, only do so if you can branch your additional remark off those three points. Never say anything “off the record”. There is no such thing. Watch your step, take care, and remember “the journalist always wins”. If you attack back (other than correcting an obvious major error) you will look petulant, and also give further publicity to whatever the journalist’s implied about you, for example that you want to use IQ tests to slaughter somebody, or clone a master race.

Finally, in defence of journalists, Pinker pointed out that they had to get the equivalent of a research grant every single day. They have little job security, and have to get something printed somewhere so as to survive till the following day. Their 24 hours is as big as your 24 month research project.

Finally: tell a tell a story. For example, what is the chance that a woman science journalist who was invited to give a lecture about a woman science journalist who died when a lorry hit her would herself be hit by a lorry a month before the lecture?

Damn, damn, damn. I should have started with that story. Anyway, let me tell you what happened next.

There was a lot of good stuff in this talk, and I hope she will get it printed somewhere so why not ask her to do so?  info@thelavinagency.com

Saturday, 13 December 2014

Competences of immigrants and natives

J. Biosoc. Sci., page 1 of 28, 6 Cambridge University Press, 2014





Above is the link to the paper. No comments from me, but look forward to comments from you.

#IQ2014 second day so far


Numerical ability is the gateway into STEM success finds E.Stern at ETH Zurich, looking at engineering and maths/physics students, average IQ roughly 120. Sample size still small, will report when up to 200 subjects.

Paulette Flore (P.C.Flore@tilburgunivesity.edu) looked at stereotype threat on girls’ test performance. She finds that the effect is at best slight, and probably due to publication bias. 26 relevant papers found total n=3760 and although a positive effect was initially found the Hedges’ g –0.22 effect size was very small. The power of most studies was low, below .4 with one exception at .8

Georg Kramer (georg.kramer@phst.at) finds that a general mental ability test plus a structured interview were a good predictor of academic achievement in college bachelor degree admission, predicting grade point average every year, and did better  in predicting final grades than the 3rd year grade point average.

#IQ2014 first day


This is a jumble, just to prove I arrive at the conference.

Ian Deary presented an extraordinary range of his work covering over 30 years showing how (rewinding) he came to write particular papers and then how the new generation of younger colleagues in his department “remixed” them with better techniques and data sets.

In words I will not repeat to sensitive readers he spoke of the power of massive samples, or better still, entire populations, to silence carping critics who claim that psychometricians always use small samples. I will try to get further detail from the very full presentation.

Gignac presented data on digit span, formerly the ugly duckling of intelligence assessment, and saved only by Jensen from being dropped from the Wechsler. Giofre put in a plea for sensitive scoring systems which incorporated partially correct scores when studying working memory.

In a very big meta-analysis Tim Bates showed that social class interacts with intelligence to some extent in US samples, but not in other parts of the world. It suggests that the much quoted Turkheimer (2003) is something of an outlier in the US funnel plot, but there is a US/rest of world difference, though hard to be sure why, possibly less supportive welfare environment for poor Americans.

Latvala showed that parent’s anti-social behaviour did not have behavioural effect on offspring. Effects seemed to be primarily genetic.

This morning we had a keynote address by Terrie Moffitt and Avshalom Caspi on how one identifies the children who will later turn out to be ultra-high cost individuals in society, using welfare, health service and getting into trouble with the law. With 81% efficiency such children can be spotted at age 3. However, whether anything can be done about it is unclear, though training in self-control seems to be a possible avenue of intervention.




Wednesday, 10 December 2014

Attending conferences: a test of intelligence?


The cognitive load imposed by attending a conference is considerable. One must book the conference, then the hotel, calculating whether you will need an extra night before or after the conference. Then you must book flights so as to arrive in good time before, and leave in good time afterwards. If necessary, you must book a car, taxi or a train to get from the airport to the conference hotel, allowing a time for the transfer. If these arrangement mean you stray into another day, then the dates must be coordinated again. The plane must not leave the airport before the car has been returned. It is almost as hard as digits backwards.

You must also ensure you have the necessary currency, passports, travel tickets, and the correct clothing. It is prudent to carry a universal power adapter, extra batteries for devices and a satnav for the car.

Is it any wonder that most people stay at home and read the abstracts afterwards? However, conferences sometimes engender the sensation that something is being achieved, that a contribution is being made to knowledge.

The notion of being a contributor to society runs deep with many citizens, and the distinction between contributors and non-contributors is both interesting and sometimes upsetting. The very young and the very old are understood to contribute less at those ages, though they will generally have contributed much over the bulk of their lifetimes. Setting aside age differences, what distinguishes contributors from non-contributors?

One of the keynote addresses at #IQ2014 looks at this contentious issue.

In “Early childhood origins of an ultra-high-cost segment of the population” Keynote Terrie E. Moffitt of Duke University and Avshalom Caspi of King’s College London
(terrie.moffitt@duke.edu; avshalom.caspi@duke.edu) say:

Worldwide, the population is aging and children are becoming rare. Nations increasingly view young people as a valuable resource for the economic and social wellbeing of whole societies. There is intense interest in early interventions to help all children achieve their potential. However, evidence is lacking to identify which
childhood risk factors to target, to yield the best return on investment.
Method: We used data from the Dunedin Study, a four-decade longitudinal study of a birth cohort of 1000 New Zealanders. We examined risk factors in childhood and measures of social, health, and economic costs in adulthood.
Results: Adult social and economic outcomes fit the Pareto principle: 20% of the cohort accounted for approximately 80% of every outcome: the cohort’s months of social welfare benefits, years of absent-father childrearing, pack-years of cigarette smoking, hospital admissions, pharmacy prescription fills, criminal court convictions, and injury-related insurance claims. Moreover, high-cost individuals with one problem outcome tended to also have multiple problem outcomes. An ultra-high-cost sub-segment of the cohort was identified who accounted for 80% of multiple problems.
Risk factors measured in childhood that characterized this ultra-high-cost group were: low family socio-economic status, child maltreatment, low self-control, and low IQ. Effect sizes were very large. Predictive analyses showed that together, SES, maltreatment, self-control, and IQ measured in the first decade of life were able to predict 80% of the individuals who are using 80% of multiple costly services. We developed an index of the integrity of a child’s brain at age three years. This age-3 brain-integrity index was a strong predictor of the cohort members who four decades later became members of the ultra-high-cost population segment.
Implications: Much research has shown that childhood risk ‘X’ can predict poor adult outcome ‘Y’, but modest effect sizes discourage translation of findings into targeted childhood interventions. This study illustrates that the vast bulk of a nation’s social services, crime control, and health-care are expended on a relatively small population segment.
During early childhood, this population segment is characterized by a small set of risk factors: low SES, child maltreatment, low self-control, low IQ, and poor brain integrity. Reducing these factors may bring surprisingly good return on investment.

I will give you more details from the conference, that is, if I manage to get myself there successfully.

Tuesday, 9 December 2014

Job opportunities in intelligence research


I don’t know how many jobs are ever offered directly to students who want to do research on intelligence, but currently they must be few and far between. However, here are some opportunities for students who already have a Master’s degree.

The Centre for Cognitive Ageing and Cognitive Epidemiology in Edinburgh is offering 3-year fully-funded PhD studentships commencing September 2015.
The studentships are in cognitive ageing/cognitive epidemiology or experimental and human neuroscience and the ageing brain. A number of projects spanning the Centre's 6 research groups are available:
- Cognitive Epidemiology
- Human cognitive ageing: Individual differences
- Human cognitive ageing: Human cognitive neuropsychology
- Mechanisms of cognitive ageing
- Genetics, genomics and epigenetics: Experimental, pathway and statistical analysis in brain ageing
- Human and animal brain imaging

Further information and an application form is available on the CCACE website: www.ccace.ed.ac.uk.
The deadline is Friday 30th January 2015; interviews will be held in March 2015. Candidates must already hold a Master's degree in a relevant field.

If your own department is looking for intelligence researchers, please send me the details so I can circulate them as well.

Listen to her play


To Holy Trinity Church last night, the Cathedral of the Arts and Crafts movement , to listen to the Carol Service, the celebration somewhat perturbed by the ministrations of a Sky TV crew such that the opening address was a safety announcement by an AFM (assistant floor manager), providing the most lowering introduction to Christmas imaginable, but the mood lifted quickly to the finely detailed rafters with the opening carol.

Then a feast of performers: Actors Robert Lindsay, Sally Phillips (Bridget Jones’ Diary), James Norton; the Freshfields Choir and CMS Chorale, that ensemble blessed with one soprano with a clarion descant. I suggested to her later that she get an agent, but her husband said she already sang just for him and he did not want his domestic arrangements perturbed, and who could blame him.

Then the astonishing Katherine Jenkins, OBE, whose slight form packs a mighty voice. She delivered a great Leonard Cohen Hallelujah; a beautiful Silent Night; and a Placide Cappeau/Adolphe-Charles Adam O Holy Night which would have gained approbation from anyone who ever sung in the Welsh valleys. An ex-teacher before her first record contract, she paid particular attention to the kids in the Colvestone Primary School Choir, beneficiaries of the The Prince’s Foundation for Children and the Arts http://www.childrenandarts.org.uk/

I mean no disrespect to the aforementioned great performers, but let me continue as I had intended.

Then the star: This lady rocks. Just off the plane with her violin that morning at 3 am from a Sony recording session in Poland, Jennifer Pike gave a commanding performance of transformative power. Her playing of Vaughan William’s The Lark Ascending was a masterpiece. Power, precision and infinite tenderness from the first note.

She closes her eyes while playing, she says, because she wants to see the form of the music in her mind, visualising where it must move to next. From time to time she looks at the faces in the audience, and seeing them changes her performance, making it different and more fun than the empty studio she was recording in the day before. She doesn’t consciously aim for a particular style, though Menuhin is always in her mind, not surprising given that she won the Menuhin International Violin Competition at the age of 12.

And so to bed.

Link for “Are bright people normal” paper



Last week I posted about normal brightness, but did not at that time have a link to the actual paper. This is a legitimate open source link. You can read it without a publisher visiting you at home and disembowelling you. Well, I cannot exclude the possibility that they will do so out of sheer devilment, but your grieving family will probably win the case against them.


Sunday, 7 December 2014

Cognitive gaps between immigrants and natives



Here is one of the posters from the ISIR conference which begins later this week. I will post up the full paper at a later date, but of course the idea of a poster is to get you to talk with the authors, ideally over coffee, but in this case more easily by commenting here on the blog.


Friday, 5 December 2014

Jobs: IQ tests versus interviews


In a place of constrained avian discourse I got into a debate about intelligence testing.  One suggestion made to me was:  Psychologists' naive interpretations of "intelligence" have been doing social harm for 80 years. Please stop it.

I replied: Ignoring intelligence measures is not without harm.

Another  questioner asked: I'd love an example of intelligence measures helping society in an equitable way.

I replied: Probably best example job selection, more equitable than interviews.

Anyway, that is enough background to understand the next step, which is about disclosing the material upon which one relies for a particular opinion. There are often testy exchanges about this. Some people get offended if told, in the middle of an argument, You should read the following papers. In my view academic debates should be punctuated by long moments of silence, broken only by pages turning as the combatants do the necessary reading. Show me your references is often a valid challenge. Not always, because there are some general points and personal observations which can be given without chapter and verse, but the ideal is that we should back up our opinions with relevant publications. The other reason for checking references is that memory is fallible, and although people may remember the general shape of the results, relevant details will have been forgotten.

I made my remark about job selection based on something I had read a decade ago in Ian Deary’s Intelligence: A very short introduction Oxford University Press, 2001. Here is the reference:

Schmidt, FL & JE Hunter (1998) The validity and utility of selection methods in personnel psychology: practical and theoretical implications of 85 years of research findings. Psychological Bulletin, 124, 262-74.



These authors had asked the simple question: is it worthwhile for an employer to select people for a job on the basis of, among other things, a test of general mental ability? The authors then looked at the correlation between the selection measures and eventual productivity, showing that intelligence test results were far better than interviews. Specifically, using test results will result in higher productivity, using an interview alone would lose three quarters of that gain. So, my main point about IQ tests being more fair than interviews can be validated by a good meta-analytic study. It was this paper which I had in mind when I made that remark. Incidentally, intelligence tests were originally seen as liberators, allowing bright working class children to rise through the system, despite not having been coached by private schools. Interviews were recognised to be rituals in which school ties and other social markers could be displayed so as to get unfair advantage.

There are other good predictors, though many take longer, and are thus more costly. For example, work sample tests are the best predictor, but since they involve a detailed evaluation of how each applicant does the sample job, they are time consuming, and difficult to arrange for some jobs. Highly structured job interviews do well (I had forgotten this) but unfortunately unstructured interviews are far more common, and far less valuable, so interviews are usually poor overall. Reference checks are not much help (perhaps because references are all much the same). Years of education is a very poor predictor, which is interesting considering it is often used as a surrogate measure of social advantage in educational research.

Compared with all these, intelligence testing can be used for all jobs, is a good predictor, and is relatively quick and cheap. The more complex the job, the better the prediction. Hunter and Schmidt found intelligence tests to be the best selection measure, and the 18 others only supplements. In terms of the multiple regression, after IQ the integrity test added 27% predictive power, a work sample or a structured interview another 24%. IQ plus integrity test is a good basic combination for an efficient result.

What did I not remember? That which I did not know when I first read the reference a decade ago. I did not know the high regard in which Schmidt and Hunter are held by methodologists. I recalled the highlights of their work as quoted in Deary’s book, rather than their names. The major consequence of their careful work was to establish a benchmark for excellence in meta-analysis.

In their 2004 paper General Mental Ability in the World of Work: Occupational Attainment and Job Performance Table 1 gives you the general ability scores of enlisted men by their civilian occupations. An instructive list which shows the intellectual demands of occupations, and also serves as a sad roll call of all the jobs that were part of an industrial society and are no more.


Their magnum opus is this fearsome tome:

Methods of Meta-Analysis: Correcting Error and Bias in Research Findings – 2004 by John E. Hunter (Author, Editor), Frank L. Schmidt (Author) Sage.

Make sure you give it as suggested reading to intelligence denigrationists.

Thursday, 4 December 2014

Advance warning: Austrian intelligence drops


This year’s International Society for Intelligence Research meeting 12-14 December will be in Austria, and it is anyone’s guess what the influx of intelligence researchers will do, however briefly, for the intellectual levels of that nation. At the very least the need to congregate in the same place and at the same time will test some basic skills, and embarrassing lapses will be gleefully published, assuming I make it to the correct place in time to record them.

There will be much to report on, and I hope to be able to do that on your behalf. On past performance it is almost impossible to cover a conference whilst also taking part. It is a once-a-year chance to meet researchers and find out what they are working on, to hear argument and counter argument, and to watch scholars of entirely contrary positions chat over coffee. Given the very high demands on my limited channel capacity, I will probably report twice: once very briefly from the conference, and then at a more length subsequently when authors have sent me their papers.

In a welcome nod to modernity the organisers have agreed a Twitter tag


I will try to get this noticed by giving you Twitter highlights, and hope you will assist me in this project so as to spread the word about the conference.

The program is shown below. The main topics are: Memory and cognitive bias; genetics; methods and psychometrics; intelligence and life outcomes; intellectual growth; neuroscience; education; creativity.


If some topics are of particular interest to you, please let me know and I will mark them for special attention.

You will see many familiar names (familiar to loyal readers anyway) and also new names of researchers who are presenting for the first time. If you have students who would like to find fame and fortune they should apply for the ISIR best graduate student presentation award given to the graduate student who gives the best presentation and includes a $500 cash award.

In the best traditions of transparency, I am sometimes on the assessment panel, and sometimes not, so a coffee and an Austrian pastry could be a good investment, or not.

Tuesday, 2 December 2014

Are bright people normal?


Every now and then I hear stories about us humans being far too dull to have done anything interesting in the past, and that the Pyramids or some such ancient government job creation scheme must have been the work of super-intelligent beings from outer space. Of course, this begs the question: by what process did these beings become super-intelligent?

However, in the way that, slightly to our surprise, we carry little bits of Neanderthal DNA and probably some early hominid adaptations to altitude and the like in our genomes, it might be the case that bright people carry bits of DNA of a sort not usually seen in common folk. For example, genius genes, genes for witty repartee and a gene for the intrinsic, uninstructed appreciation of Feynman diagrams.

In search of the miraculous, a group of researchers from the Plomin lab have analysed the Swedish Enlistment Battery test results of 3 million conscripts, from which they abstracted sibling and twins for analysis.

Nicholas G. Shakeshaft, Maciej Trzaskowski, Andrew McMillan, Eva Krapohl, Michael A. Simpson, Avi Reichenberg, Martin Cederlöf, Henrik Larsson, Paul Lichtenstein,Robert Plomin. Thinking positively: The genetics of high intelligence.
Intelligence 48 (2015) 123–132.

They say:

a key question is whether the genetic causes of high intelligence are qualitatively or quantitatively different from the normal distribution of intelligence. We report results from a sibling and twin study of high intelligence and its links with the normal distribution. We identified 360,000 sibling pairs and 9000 twin pairs from 3 million 18-year-old males with cognitive assessments administered as part of conscription to military service in Sweden between 1968 and 2010. We found that high intelligence is familial, heritable, and caused by the same genetic and environmental factors responsible for the normal distribution of intelligence. High intelligence is a good candidate for “positive genetics” — going beyond the negative effects of DNA sequence variation on disease and disorders to consider the positive end of the distribution of genetic effects.

So, normal.

But there is a little more to this than meets the eye. Very low intelligence is often caused by some “negative” genes which “interfere with the working of a naturally good brain, much as a bit of dirt may cause a first-rate chronometer to keep worse time than an ordinary watch ”. It is not impossible that a handful of “positive” booster genes are responsible for creating genius minds, even in Sweden.

In quantitative genetic studies (Nichols, 1984; Reichenberg et al., in preparation), a critical piece of evidence is that siblings of individuals with severe intellectual disability have an average IQ near 100, whereas siblings of those with mild intellectual disability have an average IQ of around 85, about one standard deviation below the population mean. In recent molecular genetic studies, rare non-inherited mutations appear to be a major source of severe intellectual disability
(Ellison, Rosenfeld, & Shaffer, 2013).

Naturally, given the big difference between familial and genetic retardation, then there might be a difference between normal wit and genius brightness. You will have noted the first part of the sentence is wrong, because familial retardation is also genetic, but “genetic” retardation is genetic plus genetic mutations. By the way, this also relates to presumed genetic differences in intelligence:


Another option is that genius is caused by some special hothouse training, in which young children are taught chess moves day and night, with calculus at lunchtime. Such kids would show genius despite their genetics: an environmental effect due to family circumstances.

In this study the top 5% of the sample were considered bright. The refined readers of this blog may be allowed a superior snigger (a snigger is laugh in a half-suppressed, typically scornful way).

The sample comprised 3 million 18-year-old Swedish males. From these, 363,905 families were identified containing at least two conscripted male siblings born in Sweden. From each family, we selected one twin pair if present (the youngest, if the
family contained more than one pair); if there were no twins, we selected the two male siblings closest to one another in age (the youngest, again, if two such pairs had the same age difference).

The authors tried two different techniques in their data analysis, and each raise interesting issues. For example, if you dichotomise population into “Very bright” vs everyone else, then the very bright group will have reduced variance, and variance is what you need when you are conducting analyses of variance.

The analysis which in my view has most power is DeFries–Fulker (DF) extremes analysis. Regression to the mean is a fascinating measure which has a strong genetic explanation and only a weak environmental explanation. Indeed, where two siblings are brought up in the same family, as is the case here, there is no really credible environmental explanation as to why one of a sibling pair should regress more to the mean than another. They have the same parents, the same house full or not full of books, and often go to the same schools.

(I will try to come back to this elsewhere, because genetic regression to the mean raises the question as to which mean you regress to: the national mean, or the mean for your genetic group?)

The dichotomous data – high intelligence versus the rest of the distribution – can be analysed by comparing the degree of concordance for MZ and DZ twins, and for non-twin siblings. Here, we used probandwise concordance: the proportion of  “affected” individuals (i.e., those with a stanine score of 9, in this case) who have a twin or sibling who is also affected. This method indicates morbidity risk, i.e., the probability that a sibling or co-twin of someone in the high-intelligence group will also be in that group

As shown in Fig. 3, MZ co-twins of those in the high intelligence group regress to the population mean to a much smaller extent than do DZ co-twins, suggesting genetic
influence. As discussed in Methods, DF extremes analysis uses continuous data, and can estimate the genetic and environmental factors influencing the difference in mean intelligence between the two intelligence groups (high intelligence vs. the
rest of the population), by quantifying the differential regression to the mean for MZ and DZ co-twins of probands.




Fig 3 shows the general population properly distributed around the Standardised mean of 0.0. The fraternal twins of bright children (shown in blue) are at IQ 0.95 which is almost 1 sd above the mean. The identical twins of bright children (shown in light green) are at 1.47 which is almost one and a half sd above the mean. Bright children (actually young adults) have a mean of 1.98, almost precisely 2 sd above the mean, which is a shade higher than expected, if I have understood the inclusion criterion correctly and they were selecting IQ 124+ the top 5%.

As explained earlier, the authors find that the results strongly support the continuity hypothesis. Bright people are normal, and at the upper end of a normal distribution. 

We found no support for the genetic Discontinuity Hypothesis that nonadditive genetic variance is greater for high intelligence, as suggested by the emergenesis hypothesis (Lykken, 1982, 2006).

So, dear reader, even if you are bright, you are very probably normal.

In terms of future research strategy, the authors say: selecting individuals of high intelligence might increase power for gene-hunting based on the simple hypothesis that high-intelligence individuals are enriched for intelligence-enhancing alleles and harbour few intelligence-depleting alleles. In other words, intellectual development can be disrupted by any and many mutations, including non-inherited (de novo) mutations, but high intelligence requires that everything works correctly. This hypothesis provided the rationale for a genome-wide case-control association study for cases with extremely high intelligence (IQ >150) compared to unselected control individuals (Spain et al., in preparation). However, in an initial report, this design does not appear to have found richer results either for identifying individual DNA variants, or for genomic approaches such as comparing the total number of rare variants (which generally have negative effects and might be expected to occur less frequently in the high-intelligence sample). Nonetheless, it is early days for the use of high-intelligence samples to increase power for gene-hunting.

Sometime soon this team will tell us if anything has come out of the detailed genomic studies of high intelligence being carried out in Beijing. Follow this blog.