The Mismeasure of Man

by Stephen Jay Gould

  Spearman on the inheritance of g

  Two of Spearman’s primary claims appear in most hereditarian theories of mental testing: the identification of intelligence as a unitary “thing,” and the inference of a physical substrate for it. But these claims do not complete the argument: a single, physical substance may achieve its variable strength through effects of environment and education, not from inborn differences. A more direct argument for the heritability of g must be made, and Spearman supplied it.

  The identification of g and s with energy and engines again provided Spearman with his framework. He argued that the s-factors record training in education, but that the strength of a person’s g reflects heredity alone. How can g be influenced by education, Spearman argued (1927, p. 392), if g-ceases to increase by about age sixteen but education may continue indefinitely thereafter? How can g be altered by schooling if it measures what Spearman called eduction (or the ability to synthesize and draw connections) and not retention (the ability to learn facts and remember them)—when schools are in the business of imparting information? The engines can be stuffed full of information and shaped by training, but the brain’s general energy is a consequence of its inborn structure:

  The effect of training is confined to the specific factor and does not touch the general one; physiologically speaking, certain neurons become habituated to particular kinds of action, but the free energy of the brain remains unaffected.… Though unquestionably the development of specific abilities is in large measure dependent upon environmental influences, that of general ability is almost wholly governed by heredity (1914, pp. 233–234).

  IQ, as a measure of g, records an innate general intelligence; the marriage of the two great traditions in mental measurement (IQ testing and factor analysis) was consummated with the issue of heredity.

  On the vexatious issue of group differences, Spearman’s views accorded with the usual beliefs of leading western European male scientists at the time (see Fig. 6.9). Of blacks, he wrote (1927, p. 379), invoking g to interpret the army mental tests:

  On the average of all the tests, the colored were about two years behind the white; their inferiority extended through all ten tests, but it was most marked in just those which are known to be most saturated with g.

  In other words, blacks performed most poorly on tests having strongest correlations with g, or innate general intelligence.

  Of whites from southern and eastern Europe, Spearman wrote (1927, p. 379), praising the American Immigration Restriction Act of 1924:

  The general conclusion emphasized by nearly every investigator is that, as regards “intelligence,” the Germanic stock has on the average a marked advantage over the South European. And this result would seem to have had vitally important practical consequences in shaping the recent very stringent American laws as to admission of immigrants.

  6.9 Racist stereotype of a Jewish financier, reproduced from the first page of Spearman’s 1914 article (see Bibliography). Spearman used this figure to criticize beliefs in group factors for such particular items of intellect, but its publication illustrates the acceptable attitudes of another age.

  Yet it would be incorrect to brand Spearman as an architect of the hereditarian theory for differences in intelligence among human groups. He supplied some important components, particularly the argument that intelligence is an innate, single, scorable “thing.” He also held conventional views on the source of average differences in intelligence between races and national groups. But he did not stress the ineluctability of differences. In fact, he attributed sexual differences to training and social convention (1927, p. 229) and had rather little to say about social classes. Moreover, when discussing racial differences, he always coupled his hereditarian claim about average scores with an argument that the range of variation within any racial or national group greatly exceeds the small average difference between groups—so that many members of an “inferior” race will surpass the average intelligence of a “superior” group (1927, p. 380, for example).*

  Spearman also recognized the political force of hereditarian claims, though he did not abjure either the claim or the politics: “All great efforts to improve human beings by way of training are thwarted through the apathy of those who hold the sole feasible road to be that of stricter breeding” (1927, p. 376).

  But, most importantly, Spearman simply didn’t seem to take much interest in the subject of hereditary differences among peoples. While the issue swirled about him and buried his profession in printer’s ink, and while he himself had supplied a basic argument for the hereditarian school, the inventor of g stood aside in apparent apathy. He had studied factor analysis because he wanted to understand the structure of the human brain, not as a guide to measuring differences between groups, or even among individuals. Spearman may have been a reluctant courtier, but the politically potent union of IQ and factor analysis into a hereditarian theory of intelligence was engineered by Spearman’s successor in the chair of psychology at University College—Cyril Burt. Spearman may have cared little, but the innate character of intelligence was the idée fixe of Sir Cyril’s life.

  Cyril Burt and the hereditarian synthesis

  The source of Burt’s uncompromising hereditarianism

  Cyril Burt published his first paper in 1909. In it, he argued that intelligence is innate and that differences between social classes are largely products of heredity; he also cited Spearman’s g as primary support. Burt’s last paper in a major journal appeared posthumously in 1972. It sang the very same tune: intelligence is innate and the existence of Spearman’s g proves it. For all his more dubious qualities, Cyril Burt certainly had staying power. The 1972 paper proclaims:

  The two main conclusions we have reached seem clear and beyond all question. The hypothesis of a general factor entering into every type of cognitive process, tentatively suggested by speculations derived from neurology and biology, is fully borne out by the statistical evidence; and the contention that differences in this general factor depend largely on the individual’s genetic constitution appears incontestable. The concept of an innate, general, cognitive ability, which follows from these two assumptions, though admittedly a sheer abstraction, is thus wholly consistent with the empirical facts (1972, p. 188).

  Only the intensity of Sir Cyril’s adjectives had changed. In 1912 he had termed this argument “conclusive”; by 1972 it had become “incontestable.”

  Factor analysis lay at the core of Burt’s definition of intelligence as i.g.c. (innate, general, cognitive) ability. In his major work on factor analysis (1940, p. 216), Burt developed his characteristic use of Spearman’s thesis. Factor analysis shows that “a general factor enters into all cognitive processes,” and “this general factor appears to be largely, if not wholly, inherited or innate”—again, i.g.c. ability. Three years earlier (1937, pp. l0–11) he had tied g to an ineluctable heredity even more graphically:

  This general intellectual factor, central and all-pervading, shows a further characteristic, also disclosed by testing and statistics. It appears to be inherited, or at least inborn. Neither knowledge nor practice, neither interest nor industry, will avail to increase it.

  Others, including Spearman himself, had drawn the link between g and heredity. Yet no one but Sir Cyril ever pursued it with such stubborn, almost obsessive gusto: and no one else wielded it as such an effective political tool. The combination of hereditarian bias with a reification of intelligence as a single, measurable entity defined Burt’s unyielding position.

  I have discussed the roots of the second component: intelligence as a reified factor. But where did the first component—rigid hereditarianism—arise in Burt’s view of life? It did not flow logically from factor analysis itself, for it cannot (see pp. 280–282). I will not attempt to answer this question by referring either to Burt’s psyche or his times (though Hearnshaw, 1979, has made some suggestions). But I will demonstrate that Burt’s hereditarian argument had no foundation in his empirical wor
k (either honest or fraudulent), and that it represented an a priori bias imposed upon the studies that supposedly proved it. It also acted, through Burt’s zealous pursuit of his idée fixe, as a distorter of judgment and finally as an incitement to fraud.*

  BURT’S INITIAL “PROOF” OF INNATENESS

  Throughout his long career, Burt continually cited his first paper of 1909 as a proof that intelligence is innate. Yet the study falters both on a flaw of logic (circular reasoning) and on the remarkably scant and superficial character of the data themselves. This publication proves only one thing about intelligence—that Burt began his study with an a priori conviction of its innateness, and reasoned back in a vicious circle to his initial belief. The “evidence”—what there was of it—served only as selective window dressing.

  At the outset of his 1909 paper, Burt set three goals for himself. The first two reflect the influence of Spearman’s pioneering work in factor analysis (“can general intelligence be detected and measured”; “can its nature be isolated and its meaning analyzed”). The third represents Burt’s peculiar concern: “Is its development predominantly determined by environmental influence and individual acquisition, or is it rather dependent upon the inheritance of a racial character or family trait” (1909, p. 96).

  Not only does Burt proclaim this third question “in many ways the most important of all,” but he also gives away his answer in stating why we should be so concerned. Its importance rests upon:

  … the growing belief that innate characters of the family are more potent in evolution than the acquired characters of the individual, the gradual apprehension that unsupplemented humanitarianism and philanthropy may be suspending the natural elimination of the unfit stocks—these features of contemporary sociology make the question whether ability is inherited one of fundamental moment (1909, p. 169).

  Burt selected forty-three boys from two Oxford schools, thirty sons of small tradesmen from an elementary school and thirteen upper-class boys from preparatory school. In this “experimental demonstration that intelligence is hereditary” (1909, p. 179), with its ludicrously small sample, Burt administered twelve tests of “mental functions of varying degrees of complexity” to each boy. (Most of these tests were not directly cognitive in the usual sense, but more like the older Galtonian tests of physiology—attention, memory, sensory discrimination, and reaction time). Burt then obtained “careful empirical estimates of intelligence” for each boy. This he did not by rigorous Binet testing, but by asking “expert” observers to rank the boys in order of their intelligence independent of mere school learning. He obtained these rankings from the headmasters of the schools, from teachers, and from “two competent and impartial boys” included in the study. Writing in the triumphant days of British colonialism and derring-do, Burt instructed his two boys on the meaning of intelligence:

  Supposing you had to choose a leader for an expedition into an unknown country, which of these 30 boys would you select as the most intelligent? Failing him, which next? (1909, p. 106)

  Burt then searched for correlations between performance on the twelve tests and the rankings produced by his expert witnesses. He found that five tests had correlation coefficients with intelligence above 0.5, and that poorest correlations involved tests of “lower senses—touch and weight,” while the best correlations included tests of clearer cognitive import. Convinced that the twelve tests measured intelligence, Burt then considered the scores themselves. He found that the upper-class boys performed better than the lower-middle-class boys in all tests save those involving weight and touch. The upper-class boys must therefore be smarter.

  But is the superior smartness of upper-class boys innate or acquired as a function of advantages in home and schooling? Burt gave four arguments for discounting environment:

  1. The environment of lower-middle-class boys cannot be poor enough to make a difference since their parents can afford the ninepence a week required to attend school: “Now in the case of the lowest social classes, general inferiority at mental tests might be attributable to unfortunate environmental and post-natal influences.… But such conditions could not be suspected with the boys who, at a fee of gd a week, attended the Central Elementary School” (1909, p. 173). In other words, environment can’t make a difference until it reduces a child to near starvation.

  2. The “educative influences of home and social life” seem small. In making this admittedly subjective assessment, Burt appealed to a fine intuition honed by years of gut-level experience. “Here, however, one must confess, such speculative arguments can convey little conviction to those who have not witnessed the actual manner of the respective boys.”

  3. The character of the tests themselves precludes much environmental influence. As tests of sensation and motor performance, they do not involve “an appreciable degree of acquired skill or knowledge.… There is reason, therefore, to believe that the differences revealed are mainly innate” (1909, p. 180).

  4. A retesting of the boys eighteen months later, after several had entered professions or new schools, produced no important readjustment of ranks. (Did it ever occur to Burt that environment might have its primary influence in early life, and not only in immediate situations?)

  The problem with all these points, and with the design of the entire study, is a patent circularity in argument. Burt’s claim rested upon correlations between test performances and a ranking of intelligence compiled by “impartial” observers. (Arguments about the “character” of the tests themselves are secondary, for they would count for nothing in Burt’s design if the tests did not correlate with independent assessments of intelligence.) We must know what the subjective rankings mean in order to interpret the correlations and make any use of the tests themselves. For if the rankings of teachers, headmasters, and colleagues, however sincerely attempted, record the advantages of upbringing more than the differential blessings of genetics, then the ranks are primarily a record of environment, and the test scores may provide just another (and more imperfect) measure of the same thing. Burt used the correlation between two criteria as evidence for heredity without ever establishing that either criterion measured his favored property.

  In any case, all these arguments for heredity are indirect. Burt also claimed, as his final proof, a direct test of inheritance: the boys’ measured intelligence correlated with that of their parents:

  Wherever a process is correlated with intelligence, these children of superior parentage resemble their parents in being themselves superior.… Proficiency at such tests does not depend upon opportunity or training, but upon some quality innate. The resemblance in degree of intelligence between the boys and their parents must, therefore, be due to inheritance. We thus have an experimental demonstration that intelligence is hereditary (1909, p. 181).

  But how did Burt measure parental intelligence? The answer, remarkable even from Burt’s point of view, is that he didn’t: he merely assumed it from profession and social standing. Intellectual, upper-class parents must be innately smarter than tradesmen. But the study was designed to assess whether or not performance on tests reflects inborn qualities or the advantages of social standing. One cannot, therefore, turn around and infer intelligence directly from social standing.

  We know that Burt’s later studies of inheritance were fraudulent. Yet his early and honest work is riddled with flaws so fundamental that they stand in scarcely better light. As in the 1909 study, Burt continually argued for innateness by citing correlations in intelligence between parents and offspring. And he continually assessed parental intelligence by social standing, not by actual tests.

  For example, after completing the Oxford study, Burt began a more extensive program of testing in Liverpool. He cited high correlations between parents and offspring as a major argument for innate intelligence, but never provided parental scores. Fifty years later, L. S. Penrose read Burt’s old work, noted the absent data, and asked Burt how he had measured parental intelligence. The old man replied (in Hear
nshaw, 1979, p. 29):

  The intelligence of the parents was assessed primarily on the basis of their actual jobs, checked by personal interviews; about a fifth were also tested to standardize the impressionistic assessments.

  Hearnshaw comments (1979, p. 30): “Inadequate reporting and incautious conclusions mark this first incursion of Burt into the genetic field. We have here, right at the beginning of his career, the seeds of later troubles.”

  Even when Burt did test subjects, he rarely reported the actual scores as measured, but “adjusted” them according to his own assessment of their failure to measure true intelligence as he and other experts subjectively judged it. He admitted in a major work (1921, p. 280):

  I did not take my test results just as they stood. They were carefully discussed with teachers, and freely corrected whenever it seemed likely that the teacher’s view of the relative merits of his own pupils gave a better estimate than the crude test marks.

  Such a procedure is not without its commendable intent. It does admit the inability of a mere number, calculated during a short series of tests, to capture such a subtle notion as intelligence. It does grant to teachers and others with extensive personal knowledge the opportunity to record their good judgment. But it surely makes a mockery of any claim that a specific hypothesis is under objective and rigorous test. For if one believes beforehand that well-bred children are innately intelligent, then in what direction will the scores be adjusted?*

  Despite his minuscule sample, his illogical arguments, and his dubious procedures, Burt closed his 1909 paper with a statement of personal triumph (p. 176):

  Parental intelligence, therefore, may be inherited, individual intelligence measured, and general intelligence analyzed; and they can be analyzed, measured and inherited to a degree which few psychologists have hitherto legitimately ventured to maintain.