Tuesday 30 September 2008

Pioneering studies of IQ

Pioneering studies of IQ by G.H. Thomson and J.F. Duff – An example of established knowledge subsequently ‘hidden in plain sight’

Bruce G. Charlton. Medical Hypotheses. 2008; 71: 625-628

Summary

Perhaps the earliest authoritative measurement of a social class gradient in IQ, with a stratification of occupations among the parents of children with different IQs, is seen in two fascinating papers published in 1923 and 1929 in the British Journal of Psychology. The authors were GH Thomson and JF Duff (both of whom were later knighted) and the papers’ main findings were confirmed by later researchers. Results of an intelligence test administered to 13419 children aged 11–12 were analyzed according to parent’s occupation. The average children’s IQ at extremes of social class among their parents included clergymen-121, teachers-116 and bankers and managers-112 at the upper end; while at the lower end there were ‘cripples and invalids’-94, cattlemen-93, hawkers and chimneysweeps-91, and the ‘insane, criminal’-88. More than 100 specific categories of parental occupations were then combined into 13 social classes, with their children’s average IQ as follows: Professional-112; Managers-110; Higher Commercial-109; Army, Navy, Police, Postmen-106; Shopkeeping-105; Engineers [ie. apprenticed craftsmen, such as mining engineers]-103; Foremen-103; Building trades-102; Metal workers, shipbuilders-101; Miscellaneous industrial workers-101; Miners and quarrymen-98; Agriculture-98; Labourers-96. A follow-up study compared an ‘intelligent’ group (IQ 136 plus) with a matched IQ 95–105 ‘control’ group. IQ testing at age 11–12 was predictive of teacher’s reports of higher levels of intelligence and health at age 16; and better performance in official examinations. The occupations of fathers, grandfathers and uncles were consistent with occupation being indicative of ‘an inherited quality’ (i.e. IQ) and there was regression from parents to grandparents and uncles among the ’intelligent’ but not among controls. Other findings included a wider variance in intelligence among boys than girls, and descriptions of the predictive value of IQ in estimating future education, examinations and health. Although the distribution, heredity and predictive value of childhood IQ measurements was once quite widely understood, for the last few decades IQ research has been regarded as morally-suspect and IQ scientists subjected to vilification, persecution and sanctions. Ignorance and misunderstanding of IQ is the norm among intellectual elites in schools, universities, the media, politics and public administration. Consequently IQ research is actively-shunned, and has near-zero influence on public policies. Since this area of science has been so comprehensively ‘disappeared’ from public consciousness as a result of socio-political pressure; it seems probable that other similarly solid and vital domains of scientific knowledge may also be ‘hidden in plain sight’.

***

Perhaps the earliest authoritative measurement of a social class gradient among the parents of children with different IQs is seen in two fascinating papers published in 1923 [1] and 1929 [2] in the British Journal of Psychology.

The authors were Godfrey H Thomson and James Fitzjames Duff – both of whom were later knighted. The research was done from Armstrong College of Durham University in the city of Newcastle upon Tyne, England – Armstrong College later became a part of King’s College which then became the independent University of Newcastle [3]. Thomson was Professor of Education in Newcastle, then later moved to the Chair of Education in Edinburgh from where he provided ‘Eleven plus’ examinations for much of Britain [4] and [5]. Duff moved to Manchester University and ultimately returned to Durham to become Warden (equivalent to Vice Chancellor – the senior administrative position in UK universities) of the collegiate Durham division of the university [3] and [6]. Duff is credited with initiating Durham’s ascent from a tiny theological and teaching college to become one of the premier UK universities [3].

As well as the intrinsic fascination of these trail-blazing researches, the two papers provoke reflection on the effect on science of changes in the national socio-political ethos. The fate of Sir Godfrey and Sir James’s papers provides an example of how once widely-accepted knowledge, generated by very senior and prestigious establishment figures, can later become generally disregarded or even denied, despite abundant scientific confirmation and elaboration by later researchers.

It seems that even in modern times, and in a liberal democratic society such as the UK where information is freely and easily accessible, scientific knowledge can apparently be ‘disappeared’ when it comes into conflict with the dominant socio-political agenda: can become, as it were, ‘hidden in plain sight’.


The social distribution of intelligence in Northumberland

The 1923 Duff and Thomson study began when an intelligence test was administered on February 24 1922, to all children aged 11 and 12 at state elementary schools in Northumberland excluding Newcastle and Tynemouth; yielding an enormous sample of 13419 children (6930 boys; 6695 girls). Further information on parental occupation was provided by teachers. The children’s IQ was then tabulated according to their parent’s (implicitly father’s) occupation.

Average IQ was 99.6, 877 children had an IQ of 120 plus and 1337 had an IQ less than 80. Boys exhibited a slightly larger apparent standard deviation than girls (no specific numbers were given by the authors), with a greater proportion of the most intelligent children being boys (IQ 130-9 - 80 boys, 49 girls; IQ above 140 – 12 boys, 4 girls) and also a greater proportion of the least intelligent being boys (IQ below 80 – 715 boys, 622 girls).

Although private schools were not sampled, and consequently there were no children with parents of the very highest social classes, nonetheless the parents social classes ranged widely from clergymen, lawyers, teachers, chemists, bankers and managers at the top; to farm labourers, brewery and mineral-water workers, ‘cripples and invalids’, cattlemen, ‘hawkers and chimneysweeps’ and the ‘insane, criminal’ at the bottom.

The average IQ (rounded to the nearest integer) of the children of some well-represented extremes of social class among the parents was clergymen-121, teachers-116 and bankers and managers-112 at the upper end; while at the lower end there were farm labourers-94; brewery workers-94; ‘cripples and invalids’-94, cattlemen-93, hawkers and chimneysweeps-91, and the ‘insane, criminal’-88. In between, by far the largest number of parents was the 5659 coal miners (average IQ of children-98).

One surprising statistic is that the children of n = 16 ‘Doctors, dentists, vets’ [i.e. veterinarians] had a reported average IQ of only 102 – the same as builders and below plumbers! My guess is that (in this particular time and place) most rural or semi-rural resident ’doctors, dentists, vets’ who sent their children to state schools were not college-educated, but had instead been trained by apprenticeship: more like craftsmen than professionals.

More than 100 specific categories of parental occupations were then combined into 13 social classes, with their children’s average IQ as follows: Professional-112; Managers-110; Higher Commercial-109; Army, Navy, Police, Postmen-106; Shopkeeping-105; Engineers [ie. apprenticed craftsmen, such as mining engineers]-103; Foremen-103; Building trades-102; Metal workers, shipbuilders-101; Miscellaneous industrial workers-101; Miners and quarrymen-98; Agriculture-98; Low grade occupations, labourers-96.

Finally the parents occupations were divided into two simple divisions of ‘brain work’ having an average IQ of 107 versus ‘hand work’ having an average of IQ 99.

Duff and Thomson also comment that although there are striking stepwise average differences in IQ by parental social class, the parental occupation according to the 13 social classes only predicted child’s IQ with a Pearson correlation of 0.28. In other words, each social class contained a range of IQs, with considerable overlapping between classes.


Following-up the children of highest intelligence

The 1929 Duff paper was a follow-up of the highest-IQ children (IQ 136 plus) which were termed the ‘intelligent’ group with an IQ 95-105 ‘control’ group matched from the same schools. Parents and teachers were asked for information, but the replies were incomplete; and data was obtained on only 64 ‘intelligent’ and 28 ‘control’ subjects.

It was found that IQ testing at age 11–12 was predictive of teacher’s reports of higher levels of intelligence and health at age 16; higher career aspirations; and also better performance in the Durham School Certificate examinations, especially the highest levels of examination results.

Occupations of fathers, grandfathers and uncles were surveyed in terms of their social class. The most striking analysis was in terms of the percentage of fathers that were at the level of skilled labourer or higher: there were 64% of fathers in the intelligent group at this level and 28% of fathers in the control group. By comparison among the intelligent group 49% of grandfathers and 52% of uncles were at this level; while in the control group 33% of grandfathers and 40% of uncles.

Duff commented that this pattern was consistent with occupation being indicative of ‘an inherited quality’ with a regression from parents to grandparents and uncles among the intelligent – but no consistent regression among the average control group when the data as a whole is analyzed. He concludes: “Intelligence is not the sole factor that determines occupation; but that it is an important factor cannot be doubted.”


Science then and now

Reading the articles after eighty years there are striking differences when compared with modern practice. Most surprisingly there is no Reference section and only five footnotes (in the earliest paper). My impression is that this paucity is partly due to embryonic nature of the field – with very little prior relevant published research; and partly due to the fact that the authors were writing for a small, familiar audience of scientific peers, who did not need to have spelled-out precisely how this piece of research fitted into the development of the subject. In those days background assumptions were often simply taken for granted, rather than referenced. The methodology was, by modern standards, skimpy – supplemented with an offer to supply extra detail to ‘anyone interested’. The general tone of these papers is therefore somewhat like a letter addressed to other members of an exclusive club.

At this early stage in the science, researchers were almost simultaneously devising methods and applying them to gather data. The social class categories used were generated specifically for this paper, and apparently on ‘commonsense’ grounds – since no detail is given about the principles underlying the classification.

Yet, for all the apparent arbitrariness and subjectivism of style (as is seems to us nowadays), and the incompleteness of the follow-up study, these two papers seem to have been both prescient and essentially correct (as judged by subsequent knowledge) and their main findings have been substantially replicated or expanded:

1. It has been confirmed that men have a wider variance in intelligence than women – with a greater proportion of both high-scorers and low-scorers [7].

2. Although Duff and Thomson’s studies did not directly measure parental IQ, the authors’ assumption was that occupations reflected IQ. Many later studies have confirmed that there is a significant social class/occupational gradient in average IQ – the size of this gradient depending upon the degree of specificity with which social class is defined e.g. [8], [9], [10] and [11].

3. Thomson and Duff’s 1923 analysis demonstrated what later epidemiologists of the 1990s re-discovered for health and social class [12] – that socio-economic differences are not absolute or fixed in size; rather the gradient is much greater when socio-economic position is analyzed precisely than when measured imprecisely. Here there was a gradient of 33 IQ points from 121 down to 88 when 100-plus specific occupations are used; a gradient of 16 IQ points from 112 down to 96 when specific occupations are collapsed into 13 groups; and a gradient of only 8 IQ points from 107 down to 99 for the dual categories of head-work versus hand-work.

4. Childhood IQ has been confirmed to be predictive of future educational (and also occupational) attainments e.g. [13], [14], [15] and [16].

5. It has been confirmed that childhood measurements of IQ are predictive of subsequent health e.g. [9], [17] and [18].

6. IQ is confirmed to be substantially heritable, and exhibits regression to the mean consistent with the degree of heritability e.g. [13], [14] and [15].

The contemporary invisibility of IQ research
Duff and Thomson were both knighted, ending their careers as highly respected and influential figures in the UK educational establishment. The main findings of these papers from the 1920s have been amply replicated in the modern consensus on IQ [e.g. [20] and see above]. And the basic understanding of the distribution, heredity and predictive value of childhood IQ measurements which they pioneered was widely appreciated.

However, for the last few decades IQ research has generally been regarded as a morally-suspect activity and the candid discussion of IQ is taboo among the intellectual elites in schools, universities, the media, politics and public administration. IQ scientists have been – and still are – subjected to vilification, persecution and sanctions [15], [19], [21], [22] and [23]. This 80 year old knowledge is typically regarded by mainstream public discourse as surprising, shocking and controversial – or the facts may even be denied outright.

Consequently, despite its remarkable prescience and importance, this pioneering work on IQ, plus three generations of supporting scientific literature, is ignored or actively-shunned – and has near-zero influence on modern public policies.

Since this area of science has so been comprehensively ‘disappeared’ from public consciousness in the face of socio-political pressure, it seems probable that other similarly solid and vital domains of scientific knowledge may also be hidden in plain sight.


References

[1] J.F. Duff and G.H. Thomson, The social and geographical distribution of intelligence in Northumberland, Brit J Psychol 14 (1923), pp. 193–198.


[2] J.F. Duff, Children of high intelligence: a following-up enquiry, Brit J Psychol 29 (1929), pp. 413–438.


[3] E.M. Bettenson, The University of Newcastle upon Tyne: a historical introduction, 1834–1971, University of Newcastle upon Tyne, Newcastle upon Tyne, UK (1971).


[4] G.H. Thomson, Education of an Englishman, Moray House, University of Edinburgh, Edinburgh (1968).


[5] Wikipedia. Godfrey Thomson. http://en.wikipedia.org/wiki/Godfrey_Thomson; 2008 [accessed 10.07.2008].


[6] Sir James Fitzjames Duff. thePeerage.com: A genealogical survey of the peerage of Britain as well as the royal families of Europe. http://thepeerage.com/p18060.htm; 2008 [accessed 10.07.2008].


[7] L.V. Hedges and A. Nowell, Sex differences in mental test scores, variability, and numbers of high-scoring individuals, Science 269 (1995), pp. 41–45.

[8] M. Argyle, The psychology of social class, Routledge, London (1994).


[9] C.L. Hart, I.J. Deary, M.D. Taylor, P.L. MacKinnon, G. Davey Smith and L.J. Whalley et al., Scottish mental health survey 1932 linked to the Midspan Studies: a prospective investigation of childhood intelligence and future health, Public Health 117 (2003), pp. 187–195.


[10] D. Nettle, Intelligence and class mobility in the British population, Brit J Psychol 94 (2003), pp. 551–561.

[11] R. Lynn and T. Vanhanen, IQ and global inequality, Washington Summit, Augusta, Georgia, USA (2006).


[12] G.D. Smith, M.J. Shipley and G. Rose, Magnitude and causes of socioeconomic differentials in mortality: further evidence from the Whitehall Study, J Epidemiol Commun Health 44 (1990), pp. 265–270.

[13] L.M. Terman and L.H. Oden, The gifted child grows up: Volume 4 (Twenty five years follow up of a superior group), Stanford University Press, Stanford, CA, USA (1959).


[14] A.R. Jensen, The g factor Praeger: the science of mental ability, Westport, CT, USA (1998) p. 15.


[15] R.J. Herrnstein and C. Murray, The bell curve, Forbes, New York (1994).


[16] G. Park, D. Lubinski and C.P. Benbow, Contrasting intellectual patterns predict creativity in the arts and sciences, Psychol Sci 18 (2007), pp. 948–952.

[17] G.D. Batty, I.J. Deary and G.S. Gottfredson, Premorbid (early life) IQ and later mortality risk: systematic review, Annals Epidemiol 17 (2007), pp. 278–288.

[18] L.S. Gottfredson, Intelligence: is it the epidemiologists elusive ‘fundamental cause’ of social class inequalities in health?, J Personality Social Psychol 86 (2004), pp. 174–199.


[19] H.J. Eysenck, Rebel with a cause: autobiography of Hans Eysenck, W.H. Allen, London (1990).


[20] U. Neisser et al., Intelligence: knowns and unknowns, Amer Psychol 51 (1996), pp. 77–101.

[21] I.J. Deary, Intelligence: a very short introduction, Oxford University Press, Oxford (2001).


[22] L.S. Gottfredson, Applying double-standards to ‘divisive’ ideas, Perspect Psychol Sci 2 (2007), pp. 216–220.


[23] J. Malloy, James Watson tells the inconvenient truth: faces the consequences, Med Hypotheses 70 (2008), pp. 1081–1091.