Friday, 17 April 2015

Income, brain, race: Prof Kimberly Noble replies


Dear James,

Thank you again for sparking a thoughtful discussion on the paper. Below I will try to address some of the points you raise.

You are of course correct that the PING study is not – and was not designed to be – an epidemiologic or representative sample. Like much university-based research in the behavioral and neurosciences, we miss out on those who choose not to participate in research or who were who were never recruited in the first place. This is mitigated slightly by the relative socioeconomic, geographic and racial diversity in our sample, relative to most other similar studies. A true representative sample in brain research is challenging as it requires neuroscientific expertise to be available wherever participants are recruited. In practice, studies have to take place where the scanners are. But I agree in principal that an epidemiologic sample of brain development would be informative indeed, and well worth funding.

The original PING study ( was not designed with questions of SES in mind per se. As such, family income and parent education data were collected, as you point out, not as actual figures but in categories. While this is not how my lab chooses to assess socioeconomic factors in our studies employing primary data collection methods, this paper was a secondary data analysis, using the best dataset in existence to address the question of SES disparities in brain structure. It is, unfortunately, the sample we have, and comes with these inherent limitations.
I believe your larger concern, however, is the treatment of environmental and genetic mediators the observed effects. You write, “It seems to have escaped notice that the apparent SES/brain link might both be driven by a common factor of inherited intelligence.”

From the results of this study, we were unable to make any conclusions as to whether the observed effects were driven by genes, experience, or their interaction. We state this in our discussion: “Furthermore, in our correlational, non-experimental results, it is unclear what is driving the links between SES and brain structure.”

Paragraph 3 of the discussion then goes on to discuss possible environmental links (of which we measured none), and paragraph 4 discusses possible genetic links (of which we measured one).

From paragraph 4: “Notably, our results can only speak to the effects of GAF [genetic ancestry factor], a proxy for race. Thus, although the inclusion of genetic ancestry does not preclude the possibility that these findings may reflect, in part, an unmeasured heritable component, it reduces as far as possible the likelihood that apparent SES effects were mediated by genetic ancestry factors associated with SES in the population. Furthermore, associations between SES factors and brain morphometry were invariant across ancestry groups.”

You write: “There is a mention of ‘an unmeasured heritable component’ but it is then dismissed because the SES and brain measure relationships were invariant across racial groups. That is a different matter.”

Of course that is a different matter! The first sentence you refer to was a recapitulation of the finding that there was no main effect of genetic ancestry, and the second was meant to stress that neither were there any SES x genetic ancestry interactions. We simply cannot comment on other heritable factors because we did not measure them.

One thing in your commentary I find unclear. You state that we “could have given the parents the psychometric test battery for good comparability.” However, parent cognitive ability is, of course, itself influenced by both genetic and environmental factors, so it is not clear to me how this would have disambiguated this question of causality. Both cognitive ability and socioeconomic status have significant heritable and environmental components, and our study was not designed to address the question of their relative balance. Parent-child genes are correlated, but so are parent-child environments. Incidentally, scanning the parents would not solve the problem either, as parental brain morphometry would be both genetically and environmentally influenced, as well.

The bottom line is that, to truly establish the direction of causality, we need a random experiment. I am pleased to be part of a team of social scientists and neuroscientists in the US who are planning just that. We are currently piloting and fundraising for a large study in which a sample of low-income mothers will be randomized upon the birth of their child to receive a large or small monthly income supplement. We then plan to follow the families longitudinally to estimate the causal impact of an unconditional cash transfer on children’s cognitive, emotional and brain development. While this won’t answer all questions, it will provide definitive evidence on the extent to which young children’s cognitive and brain development is affected by poverty reduction.

Best wishes, Kim

Kimberly G. Noble, MD PhD
Associate Professor of Neuroscience and Education, Teachers College, Columbia University
Assistant Professor of Pediatrics, College of Physicians & Surgeons, Columbia University


  1. James, please ask her about the following.

    In order to obtain the result that got all the publicity the researchers had to correct for racial differences in brain size. The single largest effect (larger than the poverty result), and the most statistically significant, is the association of cortical surface area with African ancestry.

    .25 increase in African ancestry is roughly equivalent to reduction in income by $77k (from mean of ~$100k), in terms of effect on brain surface area. (See Table 1.)

    Strange that all the people who wrote or blogged about this article failed to notice this. It stands out like a flashing red light in the paper, if you read with comprehension.

  2. GAF is a social construct.

    1. I think it's pretty funny that GAF is called a proxy for "race". I would have thought it was the other way around. Data on race are usually based on appearance or self-report, relying on a presumption is that appearance and self-report reflect ancestry (albeit imperfectly, due to, e.g., arbitrary classification of people with mixed ancestry). Therefore, I would say that race is a convenient proxy for ancestry, but GAF is a direct measure.

    2. Yes, GAF gives you the full picture, in historical detail. Self-report usually matches that very closely, but cannot date how far back genes from other groups make their arrival.

  3. Sometimes it's no use saying that a data set is the best available. The question about data isn't about relative standards, but about absolute. Are they good enough to be useful in testing a hypothesis, or at least in helping you formulate a hypothesis? Are they capable of distinguishing cause and effect, avoiding problems from confounding, and so forth? If not they are best binned rather than published. Perhaps some Social Science would best be published in The Journal of Inadequate Data.

  4. I would advise Professors Noble and Sowell to carefully read the work of Professor William Kremen of UCSD. Kremen and colleagues have shown that brain surface area is probably what accounts for the link between brain size and IQ. Also they found that variation in brain surface area is extremely highly heritable (90% or more) and that shared environment accounts only for 0 to 5%.

  5. In the data, surface is more highly correlated to WM than is brain size. You could run the partial correlation and see if size has incremental validity, I guess.

  6. Very Interesting post. I have been wondering about this and your written in such a way that it is clearly understood. Thank you.
    I have some relevant information you can review below
    order BrainPlus IQ
    BrainPlus IQ ingredients