|Galton Institute Home Page||September 1996 Newsletter Contents||Newsletter Index|
Review: Cyril Burt: Fraud or Framed? Edited by N J. Mackintosh, pp 156. (Oxford University Press 1995) £19.99
This is the latest of several books, and numerous articles, dealing with Sir Cyril Burtís work in educational psychology and politics. The contributors (A.R. Jensen, S.F. Blinkhorn, H.J. Eysenck and C.G.N. Mascie-Taylor, with three chapters by Prof. Mackintosh himself) are all pretty high-powered, and they present a well balanced account of the allegations of fraud and fabrication first made public 5 years after Burtís death in 1971, at the age of 88. But it is not likely to be the last word in a controversy which is as much political as scientific.
First of all, there can now be no doubt that Burtís reputation as the first and most influential of educational psychologists is well deserved, despite the rather discreditable attempts to rubbish it after his death. The lengthy mathematical chapter by Blinkhorn on the early history of factor analysis and kinship studies makes it clear that what Burt did here was both original and important, and that where he differed from Spearman and others as he sometimes did (and they seem to have been a quarrelsome lot!), as often as not it was Burt who was right.
But Burtís main contribution was not so much mathematical as in his work as an educational psychologist employed by the London County Council between 1913 and 1932, when he accumulated copious statistics on the performance of children in schools and wrote several highly regarded books. And he became firmly convinced, as indeed was almost everybody else at that time, that human intelligence was to some substantial extent inherited genetically and determined at conception, and so not much affected by environmental factors after birth.
But, and this was the important point, since there were far more lower down on the social scale than at the top, these might well include as many with superior intelligence as higher up, even if the percentage was lower. And that they did not do as well as children from the professional classes was because they were unable to get the education suited to their high ability for social and economic reasons. Burt therefore favoured the provision of free Grammar School education for the 20% of children, regardless of social class, who would be able to benefit from it. That implied selection, intended to give everyone a fair chance of getting the education best suited to their abilities, whether they could afford to pay for it or not; and in this it was largely successful, although mistakes were of course sometimes made and were difficult to rectify later.
Burt was of course avowedly ťlitist, but after the war most teachers and educational psychologists, and many parents, held increasingly egalitarian views where education was concerned, believing that all schools should be "comprehensive", offering the same opportunities to everyone regardless of ability. Taking it for granted that there were few if any differences in the inborn ability of normal children, if some did better than others that must be due to differences in upbringing and early education, rather than to genetical inheritance. And that could and should be corrected by remedial teaching, designed to bring those with "Special Needs" up to the same level as all the rest.
The crucial question here is whether or not everyone is born with the same potential to do well, or whether there really are genetically inherited differences in ability determined at conception, so that some are destined to do better than others through no merit or fault of their own. At first sight it seems obvious that the big differences which can easily be observed cannot be due wholly to environmental effects operating after birth. And how could our larger brains and high intelligence have been evolved otherwise than through natural selection, which certainly is not egalitarian in its effects? But it isnít easy to produce a positive proof of this. The observed correlation between parents and children in IQ, and the regression of this towards the mean both up and down, is certainly consistent with genetical inheritance, but it doesnít prove that this must have been responsible.
Burt thought that he had done so with his studies of twins separated soon after birth, and brought up by different foster families. Here identical twins not only looked alike but were closely similar in IQ and educational attainments, as well as in quite trivial ways later on in life. Separated dizygotic twins, on the other hand, were no more alike than they were to their other brothers and sisters. This is a valid approach to the problem, but it does have serious practical difficulties. First of all, twins are only very rarely separated and, when they are, the adoptive parents are usually selected to be as similar to the biological mother as possible, so they will seldom have provided significantly different environments. But, more importantly, Burtís records of these, going back to the 1920s, are deficient and it is far from clear how he assessed the IQ of the twins. Furthermore, although in the end he published details of no fewer than 53 pairs of separated twins, from which he estimated that 77% of the IQ was genetically inherited, this was not done until 1966, when he was long retired and could not himself have done all the relevant tests. After Burtís death, it was suggested that at least the later of these results must have been faked. But, although Burt may have implied that they were new results, he never actually said so, and it is now clear that most if not all of them were based upon work done long before but mislaid during the war. Most unfortunately the records were all incinerated shortly after Burtís death, on the advice of an eminent educational psychologist. In his chapter, Professor Jensen says that Burt had told him that he still had several crates of raw data on these tests, and after his death Jensen wrote to ask that these should be kept until he could examine them when he was next in England: but by then it was too late. Probably this early work would be judged inadequate by modern standards, but the records might have disposed of the allegations of fraudulent fabrication. Since Burt was in no way responsible for the loss of the evidence, however, he ought here to be given the benefit of the doubt That said, there is no question that much of what Burt produced in old age was muddled and badly presented, and would never have been published had he not been the editor of the journal in which it appeared. But this is a long way from saying that it must have been deliberately fraudulent: indeed one feels that he could have made a much better job of it had he really intended to fabricate his data.
Professor Mackintosh, in a detailed analysis of Burtís twins and other kinship studies, comes to the damning conclusion that "We know that he [Burt] was sometimes prepared to adjust his data, and at other times to make false claims about them, in order to make them appear more convincing. On balance, I believe that the evidence makes it more probable than not that some of the data he reported existed only in his imagination, in other words that he fabricated them".
With the loss of the original records, this is about as far as anyone can now go. But a belief that "on balance ... it is more probable than not" that Burt may have fudged some of his data is far from accepting that this has been established beyond all reasonable doubt. On the other side of the balance must be placed the fact that the supposed fabrications would have been done in his old age, when Burt was suffering from MťniŤreís disease with all its psychological effects. And he knew that the new generation of educational psychologists were hostile to almost all of his earlier work. Many of the posthumous accusations have anyway now been shown to be groundless, as Mackintosh himself acknowledges. For example, the famous "missing ladies", supposedly fictitious assistants in the collection of data before the war, turn out not to have been missing at all: the last of them is known to have been married in Dublin, where she died some years ago but is survived by her grown-up children.
In his final chapter, Mackintosh asks "Does it matter?". It does, and for two reasons. Firstly, if Burt has been wrongly condemned for deliberate scientific fraud, rather than mere muddle in old age, we owe it to his memory to say so. And secondly, if he was right that human intelligence and educability is to a significant extent genetically heritable, that has important implications for educational policy, which are still relevant today.
The results of Burtís pre-war twins studies which would probably be judged unsatisfactory by modern standards, quite apart from the original records being no longer available, cannot now be safely accepted. But his methodology was sound enough, and the work has anyway now been repeated by T.J. Bouchard et al., 1990, (Science 250: 223-8) who, with a sample of more than 100 sets of twins and triplets collected well after Burtís death, arrived at a figure of about 70% for the genetical heritability of IQ scores. Although those who donít like the political implications of Burtís own work are happy to dismiss it as fraudulent, nobody has suggested that the data used in this later study were fabricated, which has produced a result near enough to Burtís earlier estimate of 77%.
Twenty years ago, with the enthusiastic agreement of the politically correct establishment, Sir Cyril Burt was being described as a "senile liar" and a "wicked old fraud" who was responsible for the "biggest scientific scandal since the Piltdown hoax", denying to generations the benefits of a fully comprehensive system of State education. But a better comparison would be with Gregor Mendel, whose results were shown by modern statistical methods to have been too good to be true, long after his death. It seems clear that Mendel had adjusted his figures by excluding results which he thought were due to experimental error or were otherwise anomalous. In the 1860ís that would have been considered a normal and proper thing to do, and there is anyway now no question that the conclusions from these experiments were both valid and important. Similarly, Burtís conclusion that human intelligence and educability was to an important extent genetically heritable, and so not much affected by environmental factors operating after birth, seems now to be well established, however shaky its foundations may have been. This has important consequences for educational policy and practice, which now require serious consideration. Any system of education which takes it for granted that there are no inborn differences in childrenís educability, when in fact there are, is not going to give the best results.
C. B. Goodhart