On eating, thinking, learning and being a woman

In the August issue "Letters", Richard Lynn challenges Jim Flynn's comment that the "Flynn effect" – of substantial increases in scores on intelligence tests observed in many countries during the 20th century – is plausibly explained by "better education and smaller families". Lynn's preferred explanation is improvements in nutrition.

Associations between nutrition and ability are incontrovertible. Malnutrition severely impairs ability, better nutrition can obviously improve it. Thus in societies where the average level of nutrition is low, improvements have marked effects. In very prosperous societies, effects are much smaller and causal relationships are hard to determine because even so apparently "measurable" a factor as nutrition is a strong proxy for many other factors that also strongly determine intelligence. For example, there is excellent evidence for strong associations between maternal nutrition, infant birth weight and infant and indeed lifespan ability but levels of maternal nutrition are also directly linked to infant nutrition and both are also linked to lower lifetime risk of pathology, greater longevity and also, unsurprisingly, to greater affluence  socio-economic advantage and so with longer education. Similarly, better nutrition throughout the lifespan is a marker for socio-economic advantage, and so also for level of education, better health care and so general health and smaller family size, which are all associated with higher intelligence. In datasets that do not provide "pure" indices of each of these strongly confounded factors even powerful statistical modeling cannot identify unique determinants or even satisfactorily rank-order their effects.
 
Jim Flynn has spent much of his working lifetime showing that no single one of the (intensely interrelated) demographic indices in the data available to him accounts for all of the observed gains in population mean intelligence test scores. Unless Richard Lynn really does adopt the extreme position that only improvements in nutrition have any effect, his comments simply reflect an opinion that nutrition has had a greater effect on improvements in intelligence test scores than Flynn's preferred candidates of education and reduction in family size. This assertion is uninteresting without statistical comparisons supported by reliable data which, as far as I know, are not yet available.

Lynn does not base his disagreement with Flynn on empirical data but solely on the argument that changes in education and family size cannot account for substantial improvements in IQ test scores of two-year-olds during the 20th century. He feels that the same argument discounts my comment in The Psychologist (Rabbitt, 2006, p.674) that "growing exposure to, and awareness of the kinds of problems found in intelligence tests is enough to account for the small increases observed".

The context for my comment was a report of striking retention of problem solving by elderly people who, having completed a 10 minute intelligence test, performed significantly better when they were given it again 8 years later, although they had no further experience of it in the meanwhile (Rabbitt et al., 2004). Lynn wonders whether this can be "a plausible explanation for the increase in the IQs of two-year olds and whether an increase of approximately 27 IQ points over the period from 1917 to the present can be considered small".

Lynn's argument only works on the premise that "improvements in education" can mean only "improvements in what children have been taught in schools" rather than what they can gain from better educated parents. There is excellent evidence for strong associations between infant ability and parental educational achievement. Further, the "intelligence tests" on which two-year-olds are graded are, essentially, markers of pace of cognitive development based on successful completion of "milestone" tasks or activities. It is not sensible to argue that the likelihood of such encounters is independent of the richness of domestic environment and parental vocabulary and is also unaffected by increasing parental understanding of, and anxiety about the chronologies of markers by which infant progress is formally assessed. Parents are not only "educated" but pressured by popular books on child rearing and by the media on the value for infant cognitive development of shared activities such as reading or telling stories and of engagement in explanations of the world. On the issue of the "size" of the "IQ score" (rather than raw test score) increases, I need better evidence than I can currently find to validate Richard Lynn's estimate of "over 27 IQ points".

Even such an apparently unambiguously determinable factor as "Sex" illustrates the problem that demographic indices are usually strong proxies of each other, that that there are multiple causes for individual differences in ability and that the demographic indices available for analysis are complexly inter-related, strongly interactive, and that their rank order of effects differs even between apparently very similar social groups. During the University of Manchester Longitudinal Study of people aged between 41 and 92, men scored significantly higher than women on three different tests of fluid general intelligence. (Cattell and Cattell 1960 Culture Fair total correct, Men, N = 1294, min = 1, max = 44 mean = 26.8, sd = 7.63, Women N = 2973, min = 1, max = 44, mean = 32.66; sd = 19.78. Heim (1970) AH4-1 total correct , Men, N = 1823, min 0, max, 62; sd= 11.74; Women, N = 4240; min =  0, max = 62; mean = 30.7; sd = 11.37; Heim AH4-2 Total correct: Men, N = 1821; min = 0, max = 64; mean = 31.48; sd = 11.1; Women , N = 4239; min = 0, max = 64; mean = 27.68; sd = 10.58.) Nevertheless, "Sex" turns out to be a proxy for many other potent variables, and this is not valid evidence that women are less intelligent than men because:

(1) Male volunteers were much keener to demonstrate high levels of performance, joined the study only if they were confident they would do well and tended to leave once their curiosity was satisfied. Women tended to join for social reasons rather than for brief, competitive self-assessment and so persisted much longer with the study.

(2) Women  aged 41 to 92 in 1983 when recruitment began had shorter educations.

(3) For this and for other reasons they had also experienced less intellectually demanding professions.

(4) Even when categorised in terms of their spouses' occupations the women volunteers were less socio-economically advantaged. There were very strong links between socio-economic advantage and intelligence test scores.

(5) Significant interactive effects of Sex x City of Residence (Manchester or Newcastle) on intelligence test scores clearly show that Sex differences in mental abilities can be significantly modified by subtle, and from these data inscrutable, environmental and socio-economic factors.

(6) On some tests there are significant Age x Sex interactions, with younger women scoring lower than younger men and older women scoring higher than older men.

(7) Average scores on depression inventories are significantly higher for women than for men. Even within ranges of depression scores well below the sample median higher depression scale scores are significantly associated with lower IQ test scores.

(8) Finally, and perhaps most interestingly, although men scored slightly but significantly higher on the three tests on tests of general fluid intelligence women scored higher on all tests of verbal memory and learning and on two different tests of information processing speed.

Vivid claims that particular factors are uniquely strongly associated with intelligence test scores and so must have special practical importance or theoretical significance do provide simple, dramatic statements that attract media attention. It is much less newsworthy to draw attention to the weaknesses of our indices of categorisation of social and demographic factors and to the uncertainties and complexities of interpretation that this entails. Nevertheless I believe that this meeker approach will better help us to understand the complex aetiology of individual differences in mental abilities.