Thursday, November 13, 2014

Genes and IQ - as touchy as I'd expected

Well, it’s clear that Dominic Cummings is not one to use one tweet when six will do. But his gist is clear: he does not like my article on genes in Prospect.

It is very clear that the issue of heritability of IQ and educational achievement, and environmental effects on these, is extremely contentious, and it’s not too hard to cherry-pick the data to support whatever conclusion you like. That said, the idea that “Social class remains the strongest predictor of educational achievement in the UK”, as this paper by a professor of education points out, seems fairly well established. I accept that I might have missed some acknowledgement of this in Cummings’ book-length document to Gove, but if so, I’ll need to have it pointed out to me. I find it very hard to see how anyone reading Cummings’ paper would have come away with that impression – rather, he argues rather strongly that genes, not socioeconomic status, are the central determinant.

But his claims are stronger than that. I have been rereading this particular passage to see if somehow I have got the wrong end of the stick:

“Raising school performance of poorer children is an inherently worthwhile thing to try to do but it would not necessarily lower parent-offspring correlations (nor change heritability estimates). When people look at the gaps between rich and poor children that already exist at a young age (3-5), they almost universally assume that these differences are because of environmental reasons (‘privileges of wealth’) and ignore genetics.”

So what is Cummings implying here, if not that the differences in school performance between rich and poor children might be, at least in large part, genetic? That the poor are, in other words, a genetic underclass as far as academic achievement is concerned – that they are poor presumably because they are not very bright? I am trying very hard to square this idea with the statement by Turkheimer and colleagues (2003 – paper here) that “in impoverished families, 60% of the variance in IQ is accounted for by the shared environment, and the contribution of genes is close to zero; in affluent families, the result is almost exactly the reverse.” (Yes, Cummings alludes briefly to Turkheimer’s work, but only by linking to a blog that mentions it and only in order to dismiss it as largely irrelevant to the debate.) Cummings does not say that we should give up on the poor simply because they are genetically disadvantaged in the IQ stakes – but comments like the one above surely give a message that neither better education nor less social disadvantage will make an awful lot of difference to academic outcomes.

Cummings advocates the potential value of “finding the genes responsible for cognitive abilities”. The entire point of my article is to challenge the assumption that there are “genes for” cognitive abilities. I point out that a large group of the world’s top researchers on genetics and cognition has just conducted a big study to identify such “genes”, and finds that most of the previous candidates are illusory. They find three genes linked to variations in IQ, and these account for differences of at most just 1.8 IQ points and are in the authors’ words “not useful for predicting any particular individual’s performance because the effect sizes are far too small”. Of course, it may be that the key “genes for” just haven’t yet been found. Or it may be - and Cummings has evidently failed to grasp this central point – that because a trait is somewhat inheritable does not imply that there are “genes for” this trait in any meaningful sense. That is why I find his comments misleading.

A little bit of reading comprehension here. Cummings complains about my remark that “So it’s not clear, pace Cummings, what this kind of study [i.e. the one mentioned in the paragraph above] adds to the conventional view that some kids are more academically able than others. It’s not clear why it should alter the goal of helping all children achieve what they can, to the best of their ability.” He says “I did not make the argument he implies – i.e. we should ‘alter the goal of helping all children’…”. Does “it should alter” refer to (A) the aforementioned academic study; (B) Dominic Cummings’ paper; (C) Michael Gove’s MP expenses claim? The point here is not that Cummings doesn’t want all children to achieve what they can – I genuinely believe he does want that – but that it is not at all clear from current findings on genes and IQ that that research has much to offer the formulation of educational policy.

I’m less concerned about Cummings’ accusations of “unprofessional journalism, riddled with errors”, since he doesn’t really explain what these errors are. His advice to Prospect – “do not publish journalism on this subject without having it checked by a genuine expert” – is a bit of a hostage to fortune, given that my article was read by Steve Jones and a colleague of Robert Plomin’s, who found none of these “errors”. If you do want the opinion of another real expert, rather than mine (and who could blame you?), you might want to look at what Steven Rose has to say on the topic, and then decide if there is any daylight between that and this.

3 comments:

Stephen Hsu said...

Hi Philip,

This is indeed a touchy subject. Given your interest, you might find this overview of use:
http://arxiv.org/abs/1408.3421

"Gene for" may be OK terminology for Mendelian traits that are controlled by something akin to an on/off switch. But for polygenic traits (also known as quantitative traits), such as height or intelligence, one has to talk about "causal variants" with specific effect sizes. As a former physicist, you might find it striking (as do I) that most of the variation in polygenic traits tends to be linear (additive). That is, to first approximation one can simply sum up individual effects to get a prediction for the phenotype value. Thus, we can talk about individual variants that (on average) increase height or intelligence by a certain fraction of a cm or even IQ point. There are deep evolutionary reasons for this (see section 3 in the paper linked above), as well as quite a bit of experimental evidence.

Finding causal variants is a matter of sample size, or statistical power. At present, almost 1000 such variants are known for height, accounting for about 20% of population variation. Available genomic datasets for which we have IQ scores are much smaller, hence the more limited results. I expect that as sample sizes continue to grow we will eventually capture most of the heritability for these traits; see section 4 and figure 13 in the paper. (Note heritability by itself can be estimated using different methods, from much smaller sample sizes.)

If you have any comments or questions please feel free to contact me.

Steve Hsu

PS I also sent you this message via email form on your author web page.

MichaelRosen said...

1. They've never found out what 'intelligence' is. All they have are the results of tests which are devised in order to produce the bell curve. This isn't science. It's social engineering.
2. If we want to have any decent sense of 'intelligence' it's an unmeasurable overall assessment of what a person's capabilities are.
3. The narrower they make intelligence - i.e. the ability to answer questions of a very particular sort in a very short time, on your own, without consulting anything or anybody (i.e. not a situation that prevails elsewhere in life) then the easier it is to fix distributions and claim genetic links. However, it's the premise that what is being tested is valid that is the problem
4. The key to figuring the bunkum is history. In the past, swathes of society were described as cretins, morons, uneducable, 'simian' (Burt) and so on. Society develops, more and more people get more and more education and qualifications. So the descriptions of the uneducable narrow down to a smaller percentage or proportion even as they build failure into norm-based high stakes exams, and create competition between schools in order to create failure! But historically, the nation has 'improved'. IQ results get better…layers of society once written off are now required to do less and less manual work and more and more kinds of work that were thought impossible for the proles to do in previous generations. How come? The answer is that the genetic link is rubbish. Environment (and interaction with it, we must always say, or we fall into another kind of determinism) has been the key.
5. It is vital for certain forces in society to produce tests which confirm hierarchies, which, when applied, reinforce the hierarchies in the schools they devise. Hierarchy…hmmmm…..now why would that have anything to do with capitalism….let me think….

The Question said...

Why Children's Literature should not be taught in universities:

1.They’ve never found out what “children’s literature” is. All they have are books written for children in order to produce books that children will read. This isn’t literature. It’s social engineering.
2. If we want to have any decent sense of “children’s literature’ it’s an unmeasurable overall assessment of how suitable for a child a work of literature is
3. The narrower they make children’s literature – i.e. books of a very particular sort for a very small age band, read on your own, without consulting anything or anybody (i.e. not a situation that prevails elsewhere in life, which mostly comprises shapes, colours and sounds) then the easier it is to fix it so that children only read children’s literature and claim that it is therefore children’s literature. However, it’s the premise that children’s literature is a valid genre that is the problem.
4. The key to figuring the bunkum is history. In the past, swathes of literature were described as adult, too old, not interesting, 'boring' (Wilde) and so on. Society develops, more and more people get more and more children and books. So the descriptions of the unreadable narrow down to a smaller percentage or proportion even as they build failure into child-based definitions of literature, and create competition between publishers in order to create failure! But historically, the nation has 'improved'. Children’s literature gets better…layers of literature once written off are now required reading and more and more kinds of literature exist that were thought impossible for the children to read in previous generations. How come? The answer is that the definition is rubbish. Literature (and interaction with it, we must always say, or we fall into another kind of determinism) has been the key.
5. It is vital for certain forces in society to produce educational qualifications which confirm hierarchies, which, when applied, reinforce the hierarchies in the schools they devise. Hierarchy…hmmmm…..now why would that have anything to do with capitalism….let me think….