Thursday, October 22, 2015

Breastfeeding and IQ

Breastfeeding is VERY politically correct these days.  Mothers who do not breastfeed can be harassed by other mothers over it.  Why?  Because breastfeeding is thought to be  "more natural" and hence better for the baby.  But better in what way? One claim is that is helps the child's IQ.  But the studies have not been very supportive of that. So the latest very extensive study is of great interest.  Abstract below:

Breastfeeding and IQ Growth from Toddlerhood through Adolescence

By Sophie von Stumm &  Robert Plomin

The benefits of breastfeeding for cognitive development continue to be hotly debated but are yet to be supported by conclusive empirical evidence.

We used here a latent growth curve modeling approach to test the association of breastfeeding with IQ growth trajectories, which allows differentiating the variance in the IQ starting point in early life from variance in IQ gains that occur later in childhood through adolescence. Breastfeeding (yes/ no) was modeled as a direct predictor of three IQ latent growth factors (i.e. intercept, slope and quadratic term) and adjusted for the covariates socioeconomic status, mother's age at birth and gestational stage. Data came from the Twins Early Development Study (TEDS), a prospective cohort study of twins born between
1996 and 1994 in the United Kingdom, who were assessed 9 times on IQ between age 2 and 16 years (N = 11,582).

Having been breastfed was associated with a small yet significant advantage in IQ at age 2 in girls (β = .07, CI 95% from 0.64 to 3.01; N = 3,035) but not in boys (β = .04, CI 95% from -0.14 to 2.41). Having been breastfeeding was neither associated with the other IQ growth factors in girls (slope: β = .02, CI 95% from -0.25 to 0.43; quadratic: β = .01, CI 95% from -0.02 to 0.02) nor in boys (slope: β = .02, CI 95% from -0.30 to 0.47; quadratic: β = -.01, CI 95% from -0.01 to 0.01).

Breastfeeding has little benefit for early life intelligence and cognitive growth from toddlerhood through adolescence.

Von Stumm S, Plomin R (2015). Breastfeeding and IQ Growth from Toddlerhood through Adolescence. PLoS ONE 10(9): e0138676. doi:10.1371/journal.pone.0138676

The study is persuasive rather than conclusive.  I think IQ of the mother should have been controlled for.  I made the same criticism of a noted Brazilian study which did find some benefit from breastfeeding.

Another concern is that the measures of IQ used at different ages were not well correlated. They could obviously not be the same but correlations between them as low as .18 are a serious concern.

Overall, however, the general agreement of the studies on the matter leads me to agree that breastfeeding has no effect on IQ.  It may however have other benefits.

UPDATE:  Those .18 correlations in Table 1 are of course absolutely appalling so I have been thinking about that.  The simple thing to say is that the questions you ask a 4-year-old to assess his IQ and the questions you ask a 16-year old to assess his IQ are necessarily  very different -- so a high correlation is not to be expected.  There is however a conventional solution to that conundrum:  Use a spiral omnibus test -- where the questions start out very easy and gradually get harder.

The authors above, apparently, did not however have that luxury.  So their solution was a creative one which I rather admire.  They took the first eigenvector of the battery they did have and standardized that as IQ (mean 100; SD 15).

So what do we find from that?  It could be argued that they have for the first time made IQ tests that are valid for particular age groups.  And in that case what we see is that IQ is very variable  throughout the lifespan.  Being bright at 2 tells us little about  IQ at 16

And I think that is an important finding.  In particular it conforms to other findings that environment is important in early life but, as time goes by it is the genetic given that manifests itself.

Be that as it may, the measures of IQ used in the early years are clearly just not valid.  They do not correlate with well-accepted  measures from later life. Putting it more bluntly, trying to measure IQ at age 2 is just a no-go.  It fails.  It tells you  nothing.

In that case the slight effect seen at age 2 is a nonsense and not to be taken seriously.

And Table 1 in the article has another interesting implication.  It bears on the "Eleven Plus" exam used in England to filter access to Grammar (selective) schools. There was no IQ given for age 11 but there was for age 12.  And we see there that  the correlations for age 12 and up averaged around .6.  That is not ideal but, given changes in IQ throughout the early lifetime, is probably as good as can be expected. Those eigenvectors were not too bad as IQ measures!

Wednesday, October 21, 2015

Are firstborns smarter?

This is an old hypothesis that Robert Zajonc devoted considerable work to in the '70s.  He found an effect of up to 3 IQ points.  A recent international study with large samples has however recently re-examined the theory.  There is a popular account of the findings here. The journal abstract is as under:

Examining the effects of birth order on personality

Julia M. Rohrer et al.

This study examined the long-standing question of whether a person’s position among siblings has a lasting impact on that person’s life course. Empirical research on the relation between birth order and intelligence has convincingly documented that performances on psychometric intelligence tests decline slightly from firstborns to later-borns. By contrast, the search for birth-order effects on personality has not yet resulted in conclusive findings. We used data from three large national panels from the United States (n = 5,240), Great Britain (n = 4,489), and Germany (n = 10,457) to resolve this open research question. This database allowed us to identify even very small effects of birth order on personality with sufficiently high statistical power and to investigate whether effects emerge across different samples. We furthermore used two different analytical strategies by comparing siblings with different birth-order positions (i) within the same family (within-family design) and (ii) between different families (between-family design). In our analyses, we confirmed the expected birth-order effect on intelligence. We also observed a significant decline of a 10th of a SD in self-reported intellect with increasing birth-order position, and this effect persisted after controlling for objectively measured intelligence. Most important, however, we consistently found no birth-order effects on extraversion, emotional stability, agreeableness, conscientiousness, or imagination. On the basis of the high statistical power and the consistent results across samples and analytical designs, we must conclude that birth order does not have a lasting effect on broad personality traits outside of the intellectual domain.


So the effect was very small.  In fact, if we look at the supplemental material, we see that the difference in the British sample was just one IQ point -- totally unimportant for all intents and purposes.  So even the small effect found by Zajonc would seem to have been overstated.

Whether a difference of one IQ point requires explanation is unclear but several environmental explanations have been suggested in the links above.

What was NOT found is also interesting. That birth order had NO effect on any personality variable upsets a lot of theories -- but is consistent with genetics rather than the environment being the main influence on personality -- e.g. if you are a miserable whiner like most Leftists are, you were born that way.

One theory that would seem rather damaged by the findings even though it was not directly tested was the pet theory of Frank Sulloway.  Firstborns are conservatives and later-borns are  rebellious, says Sulloway.  Rebelliousness would seem to be a personality variable. See here for a dissection of the strange Prof. Sulloway and his theory.

Sunday, October 18, 2015

Larger brains do not lead to high IQs

One has to laugh.  I have often noted that people reporting research results tend to conclude what they want to conclude rather than what the data concerned actually shows.  I had a lot of fun reporting such cases in my own years as an active researcher.  See here. The heading above, taken from MedicalXpress, is another case in point.

It is politically correct.  Leftists don't like to think that there is ANYTHING inborn or hard-wired in human beings.  And, as a consequence, many of them reject ANY physical basis for IQ. And the heading above is just such a rejection. One problem: The findings they were allegedly summarizing in fact showed a stable and statistically significant correlation between brain size and IQ!

So what is going on?  Before I comment further, I reproduce the underlying journal abstract.

Meta-Analysis of Associations Between Human Brain Volume And Intelligence Differences: How Strong Are They and What Do They Mean?

Jakob Pietschnig et al.


Positive associations between human intelligence and brain size have been suspected for more than 150 years. Nowadays, modern non-invasive measures of in vivo brain volume (Magnetic Resonance Imaging) make it possible to reliably assess associations with IQ. By means of a systematic review of published studies and unpublished results obtained by personal communications with researchers, we identified 88 studies examining effect sizes of 148 healthy and clinical mixed-sex samples (> 8,000 individuals). Our results showed significant positive associations of brain volume and IQ (r = .24, R² = .06) that generalize over age (children vs. adults), IQ domain (full-scale, performance, and verbal IQ), sample type (clinical vs. healthy sample), and sex. Application of a number of methods for detection of publication bias indicates that strong and positive correlation coefficients have been reported frequently in the literature whilst small and non-significant associations appear to have been often omitted from reports. We show that although the positive association of brain volume and IQ seems to be robust, its strength has been overestimated in the literature. While it is tempting to interpret this association in the context of human cognitive evolution and species differences in brain size and cognitive ability, we show that it is not warranted to interpret brain size as a necessary cause or isomorphic proxy of human intelligence differences.


Got that?  They found that there was an association between the two variables but it was weak.  In other words, brain size was only one of the factors influencing IQ -- a conclusion which would surprise no-one familiar with the field.  Cortical complexity would obviously be a much more major influence. And cortical size considered independently of overall brain volume would also be a very promising variable for study.

The correlation (.24) IS weaker than we would expect from prior studies of head size.  Head size normally correlates over .3 with IQ.  One of the authors of this study was however the ultra-cautious Dutch psychologist Jelke Wicherts.  You can't put anything past him!  So he has ensured extensive allowance for confounding factors etc. So the figure reported above might reasonably be treated as a lower bound.  ALL correlations seem to shrink and all IQs seem to rise when the distinguished Prof. Wicherts gets hold of them! Wicherts is an Associate Professor in Psychological Methods specializing in errors with statistics and the measurement of intelligence (IQ).

The much-published Jelte Wicherts

And as further context, let me note that in psychological research generally, a correlation of .24 would be greeted with frabjous joy. My recollection is that a majority of correlations in psychological research are around that magnitude.  To have detected ANY effect of one variable on another is normally felt worthy of congratulation. And I am not entirely being mocking in saying that.  Any item of human behaviour is bound to be multi-causal so any one influence on the behaviour concerned MUST usually be small.  The behaviour will be the product of many influences, not one.

A final note:  The heading of the MedicalXpress article is not the only bit of political correctness in it.  They also state that average male and female IQs are the same. The wicked Richard Lynn's very extensive study of the data showed that women are in fact down by a couple of points. 

That's not the important point about female IQ, however. The smaller variability (SD) of female IQ has long been known and is well-accepted.  And that restricts female achievement at the top of the range.  I had better not spell that out further in case I get prosecuted for hate speech.  You can get away with saying things in academic language that would land you in trouble if you said  them in plain English -- JR