A review of: N.J.MACKINTOSH
(ed.), 1995, Cyril Burt: Fraud or Framed? Oxford : Oxford University Press. Pp. vii + 156. ISBN 0 19 852336 X. stlg19-99.
By Chris Brand
Review published in Nature 377, 6548, 394-395, October 5, 1995.
Sir Cyril Burt (1883-1971) thought that the broad heritability of IQ was high_ yielding his notorious correlation of .77 between monozygotic twins reared apart (MZA's). Yet he estimated the narrow heritability as only .52: parents could pass on genetically to their children only a half of their IQ advantage (or disadvantage). Partly by this route, parents could transmit a third of their advantage (or disadvantage) in socio-economic status (SES). Countering regression to the mean, the class system would be refreshed by social mobility. The majority of the brighter children would come from `working class' homes and children's own IQ's would account for some 50 per cent of eventual SES variance in their own generation. General intelligence (g), which Burt had once found to relate to speed of intake of simple perceptual information, was frankly more important in life than were `personality traits'; but there were other sources of variation in mental abilities ("group factors"), and education should accord with children's ability profiles. There were also innate differences between races in g; but "they are small" in comparison with the big differences between individuals. Progressive enough for Burt to be knighted under a Labour government, Burt's position infuriated rising social scientists and geneticists who abjured anything that could be linked to `negative' eugenics. "Wouldn't it be great if it could be shown that Burt was really just an old fraud!" muttered one London educationist to Arthur Jensen in 1957. After Burt's death, closer scrutiny of his key work led at last to its being denounced as involving casualness or fraud, and bolder accusations of `fascism' soon followed. Here, the eminent learning theorist, Nicholas Mackintosh, leads a hand-picked team of scholars in a re-examination of Burt's character and figurework. Under-reporting of the details of what Burt himself often admitted to be "precarious" studies had always limited the scientific usefulness of Burt's research reports; and Burt's .77 figure is replaced today by the .78 estimate from the 43 MZA pairs studied in Minnesota. On the other hand, the charge of fraud against Burt has been rejected by most experts as `not proven', chiefly because Burt's figures have crazily rough edges that no true fraud would have failed to smooth down_notably the .77 figure and also correlations for height and weight that remained the same across two big changes in sample size. Accepting all this, Mackintosh's contributors try to focus mainly on what only probably happened: "we are not trying a case in a court of law," they declare. As a verbal token, the term `fabrication' is often used, rather than `fraud'. Thus liberated, the authors all find something quite interesting to say. Hans Eysenck colourfully exposes a level of deviousness in committee work that would put Burt well within the top 50% of vice-chancellors and right up among the equally fabrication-prone Kepler, Newton and Freud. Stephen Blinkhorn rehearses quite brilliantly, if a little self-indulgently, how Burt had every right to feel aggrieved at being written out of the history of factor analysis by Spearman and Thurstone. Nicholas Mascie-Taylor agrees to exonerate Burt from the best-known charge that his figures on social mobility were too perfect; but he details convincingly Burt's `deliberate deceptiveness' about his inadequate data on social mobility; and Jensen finds Burt a "brilliant eccentric" who was inexplicably and inexcusably "furtive." The book as a whole is fairly and indeed beautifully written. Such points, however, advance no very general claim about Burt. Eysenck, the giant of trait psychology, implicitly acknowledges the problem when he admits two quite different `sides' to Burt's character. Fortunately, Mackintosh, who writes three of the chapters, has more of an agenda and realizes the need to come off the fence _at least in his epilogue. Taking Burt's 1966 MZA study first, Mackintosh establishes a magisterial authority by a sustained defence of the possibility that Burt had twin data from the 1920's. All that is missing from Mackintosh's account is any consideration of how, in 1966, Burt might have increased the number of MZA pairs on which relevant and supportive data were available. He could have embraced some of the better-separated pairs reported in J. Shields' Monozygotic Twins (1961). Hi-jacking of Shields' twins would explain Burt's otherwise peculiar statement that his 1966 paper would "bring together the evidence now available both from our own studies and from more recent investigators"; and why he quite openly told Eysenck in 1971 that "our own studies" were mostly complete by 1939. It would also explain Burt's wish, in 1969, while "calculating data on twins for Jencks", to have access to University College Library: simply, he needed access to the copious raw data in Shields' book from which he (or Miss Conway) had once learned the crucial correlation of .77 that corresponded with Burt's own. Plainly, at least something as `devious' as this happened; but no-one knows what. Having cleared Burt of fabrication of the twin data, however, Mackintosh maintains suspense for the volume with an engaging new line of attack. Perhaps the biggest difficulty for Burt-baiters is that Burt was spectacularly correct on so many points. How could he possibly have managed without data? So it is reasonable that an apparent error on Burt's part should arouse suspicion. As background to his 1969 figures (unsatisfactory,as usual) on declining educational standards in Britain from 1914, Burt explains that children's g levels had not changed. To demonstrate this required him to have tested children in the 1960's with his 1914 tests; and Burt claims to have done that. However, when this sort of exercise has been undertaken by others, as first in the 1940's, g-scoring has actually been rising at some two IQ points per decade. With this as the strongest case that Mackintosh finds he can make for fabrication, many would-be accusers will hold their peace. After all: the secular IQ rise has occurred not on Burt's tests but chiefly on others where important scoring opportunities can be gained by guessing and/or skipping harder items instead of persevering; Burt plainly resisted the temptation to recognize an IQ-type rise that would helpfully have thrown `declining educational standards' into sharper relief; and even James Flynn, the chief exponent of the rise, does not himself believe it to have been in true intelligence_only in whatever conventional IQ tests measure. Burt thus actually notches up yet another success_at least while agreeing with Flynn is Mackintosh's touchstone. Eysenck and Jensen themselves have admittedly interpreted Flynn's work as indicating a substantial g rise, so Mackintosh might squeeze more testimony to Burt's psychopathy from them. Properly considered, however, Mackintosh's academic whodunit marks a further step towards Burt's rehabilitation.
