You are currently viewing Sexual maturation and intelligence, Rushton not vindicated

Sexual maturation and intelligence, Rushton not vindicated

A reader sent me Seb Jensen’s blogpost about nerds, which includes the following:

People in the Terman’s study of the gifted were more likely to be tall, broad-shouldered, have matured earlier sexually, have a stronger hand grip, and have a strong lung capacity relative to other children.

If we think back to my recent blogpost about how a new GWAS of age of onset of walking vindicates Rushton’s big theory, this finding does the opposite. In Rushton’s model, there is a trade-off between early development and later intelligence (and other phenotypes). A finding that highly intelligent children sexually mature earlier — within populations — does not fit this model. However, because the Terman study is very old, and it is known that Terman took certain liberties in his descriptions, it it sensible to check for replications. Here are some of them:

First study

Method Examination scores at age 16 were studied among 13,477 British twins participating in the population-based Twins Early Development Study. A pubertal development scale, a height-based proxy of growth spurt, and age at menarche were used as indicators of puberty. Associations between puberty, sex, and academic achievement were estimated in phenotypic mediation models and biometric twin models.
Results Earlier puberty was associated with higher academic achievement both in boys and girls. The exception was early age at menarche in girls, which associated with lower academic achievement. More than half of the sex differences in academic achievement could be linked to sex differences in pubertal development, but part of this association appeared to be rooted in prepubertal differences. The biometric twin modelling indicated that the association between puberty and academic achievement was due to shared genetic risk factors. Genetic influences on pubertal development accounted for 7%–8% of the phenotypic variation in academic achievement.
Conclusions Pubertal maturation relates to the examination scores of boys and of girls. This can give genes related to pubertal maturation an influence on outcomes in education and beyond. Sex differences in pubertal maturation can explain parts of the sex difference in academic achievement. Grading students when they are immature may not accurately measure their academic potential.

The correlations are quite small, but generally they are positive, not negative. The same-aged variables show the largest positive correlations (“PDS, 16” rows). Age of menarche (onset of menstruation in girls) shows only very tiny, but again positive correlations. In other words, children who scored higher on the achievement test (GCSE) — a good proxy for intelligence — were slightly earlier to develop sexually. This does not fit Rushton’s trade-off model.

Second study

  • Daniel Jr, W. A., Duke, P. M., Carlsmith, J. M., Jennings, D., Martin, J. A., Dornbusch, S. M., … & Siegel-Gorelick, B. (1982). Educational correlates of early and late sexual maturation in adolescence. The Journal of pediatrics, 100(4), 633-637.

From the National Health Examination Survey data, 5,735 Caucasian males and females, 12 to 17 years, were classified by age and stage of sexual maturation (Tanner). Early and late maturers were each compared to all other youth of comparable age and sex, in eight education-related categories: youth and parental aspirations and expectations concerning the level of education which would be achieved by the student, teacher reports of intellectual ability and academic achievement, and test scores (WISC and WRAT). Except at age 12, late maturing boys received significantly lower ratings than mid maturers in all these areas, and early maturing males received higher ratings. For females, no differences persisted across age groups. In advising male adolescents, physicians should be alert to the possibility that school functioning may be linked to maturational processes.

They used a robust measure of sexual development:

Tanner sexual maturation stages ]l were determined at the time of the physical examination. These stages, which correlate very well with other pubertal measures, provide an excellent noninvasive assessment of pubertal status. The correlation with bone age is 0.69 for males and 0.81 for females; the correlation with menarche is 0.74. During the physical examination, each youth was compared with standard photographs (breast and pubic hair for females, and testes, penis and pubic hair for males) in order to assign a sexual maturity rating. In analyzing the data, we averaged the Tanner ratings for each youth, thus obtaining one sexual maturity score per youth. For each girl, the mean of the two Tanner ratings for each breast was averaged with the pubic hair rating. Each boy’s score was derived by averaging his pubic hair and genitalia scores.

The stippled line are the late maturers (slow developers), their scores, both objective (WISC and WRAT) and subjective were below average, while the early maturers were above average. Clearly, the expectation from Rushton’s model failed here.

Third study

  • Noipayak, P., Rawdaree, P., Supawattanabodee, B., & Manusirivitthaya, S. (2016). Age at menarche and performance intelligence quotients of adolescents in Bangkok, Thailand: a cross-sectional study. BMC pediatrics, 16, 1-5.

Background The presence of an association between age at the onset of puberty and intelligence quotient (IQ) in young adolescents remains controversial. The aim of this study was to explore the association between age at menarche and performance IQ scores of young female adolescents in Bangkok, Thailand.
Methods A cross-sectional study was conducted among 537 students aged 11–15 years attending primary and secondary schools in the Dusit district of Bangkok, Thailand. The participants were selected based on two-step stratified sampling. Age at menarche and health and socioeconomic status were determined using a self-report questionnaire completed by participants. Performance IQ scores were determined using the Standard Progressive Matrices intelligence test (Thai version) administered by registered clinical psychologists.
Results Of the 537 participants, 0.4 had reached menarche at 8 years of age, 1.9 at 9 years, 10.1 at 10 years, 36.1 at 11 years, 37.6 at 12 years, 10.4 at 13 years, 3.4 at 14 years, and 0.2 % at 15 years. Age at menarche was inversely correlated with performance IQ (Pearson correlation −0.087, p = 0.047). The regression equation predicting performance IQ by age at menarche was performance IQ = 128.06 − 1.16*age at menarche (R 2 = 0.008). In univariate analysis, performance IQ was inversely correlated with age at menarche, body mass index (BMI), time spent watching television, and time sleeping, but was directly correlated with maternal age at birth (all p < 0.05). In multivariate analysis, age at menarche and BMI remained significantly inversely correlated with performance IQ (p < 0.05), while maternal age at birth was directly correlated with performance IQ. The model consisting of age at menarche, BMI, and maternal age at birth best predicted performance IQ.
Conclusion After adjusting for confounders, multiple regression analysis showed that age at menarche and BMI of young female adolescents living in the Dusit district of Bangkok, Thailand, were inversely correlated with performance IQ, whereas maternal age at birth was directly correlated with performance IQ.

It’s a p-hacked study, but that doesn’t matter here. They find a small negative correlation between IQ (Standard Progressive Matrices) and age of menarche, where Rushton’s model expects a positive correlation.

Discussion

Four unrelated studies all confirmed a slight positive correlation between maturation speed, whether measured by Tanner’s method, puberty, or teacher’s ratings. The studies were done far apart in time and space, the Terman study in the 1920s or 1930s USA, The British study in 2010 or so, the second US study in 1960, and the Thai study in 2016. Clearly, this pattern is robust to variation in study settings. Given the consistency of the results, it doesn’t seem likely that finding additional studies would change the pattern.

So what do we make of this? From Rushton’s theory’s perspective, the reason that the races differ in age of menarche — and this is true enough — is that Africans develop earlier due to selection on life history speed. This fits and makes sense, but the within group findings all seem to suggest this trade-off doesn’t exist at the individual level. This makes it hard to understand how this race-level pattern could arise from the individual forces of selection.

To save Rushton’s theory in this case, one would have to posit some confounders that cause the correlation to be unexpectedly reversed within population, while the direct genetic (pleiotropic, shared causal genes) correlation is actually positive. It’s very rare to find that genetic and phenotypic correlations differ in direction (Cheverud’s Conjecture). The best I can think of is that this correlation exists in the wild (ancestral human populations), but that modern societies with plenty of nutrition have caused its reversal. We know that age of menarche starts earlier in modern populations, here’s some data from India:

And here’s western data:

I can’t think of a plausible mechanism that would have reversed the relationship between developmental speed and intelligence. Maybe we can posit there are two different genetic mechanisms for this pattern. First, there are some genes that control the trade-off in life history speed, and these have a negative genetic correlation between the two phenotypes in questions. Second, there are some more genes that control general body functioning, where worse genes just cause everything to work less well in general, causing a positive correlation. In this ad hoc theory, there would be more of the second kind of genes, or their effects would be stronger. This kind of model can be tested using local genetic correlations, as has been done for e.g. autism and intelligence. This pattern showed that while most genes (genetic variants to be precise) have shared positive correlations with intelligence and autism, some are negative. This results in potentially two kinds or clusters of autism, the more typical high IQ Asperger’s types (typical programmers), and the totally dysfunctional, institutionalized, low intelligence types. In this ad hoc model, then, Rushton’s model predicts that it is the first subset of genes that have given rise to the between race patterns we see, not the latter.

It is possible to sort of save Rushton’s theory in another way, but hypothesizing that selection differed between races such that Africans and Europeans (say) had different selection pressures for maturing earlier and also independently for intelligence, and that this so happens to result in a race-level pattern consistent with life history theory. But it doesn’t appear that singular selection for life history speed gave rise to this pattern.

In general, it has been many years since Rushton did his literature review, and his bold, encompassing theory has not always held up when analyzed closer. It would be good if a team of researchers would re-investigate this model and check all the claims in light of modern research. I don’t find it plausible enough to spend the time doing all of this work, but I’m happy enough to look into a few specific cases.