Swann (1985)

Notes on the text
Preliminary pages Membership, Contents, Introduction

Part I: Setting the scene
Chapter 1 The nature of society
Chapter 2 Racism: theory and practice
Chapter 3 Achievement and underachievement
Chapter 3 continued

Part II: Education for all
Chapter 4 Ethnic minorities and education: historical perspective
Chapter 5 Multicultural education: further studies
Chapter 5 continued
Chapter 6 'Education for all': a new approach

Part III: Major areas of concern
Chapter 7 Language and language education
Chapter 8 Religion and the role of the school
Chapter 9 Teacher education; employment of ethnic minority teachers
Chapter 9 continued

Part IV: 'Other' ethnic minority groups
Introduction
Chapter 10 Chinese children
Chapter 11 Cypriot children
Chapter 12 Italian children
Chapter 13 Ukranian children
Chapter 14 Vietnamese children
Chapter 15 'Liverpool Blacks'
Chapter 16 Travellers' children
Reflections and conclusions

Part V:
Main conclusions and recommendations

Appendices

Report of the Committee of Enquiry into the Education of Children from Ethnic Minority Groups

Chairman: Lord Swann

Cmnd. 9453

London: Her Majesty's Stationery Office 1985
© Crown copyright material is reproduced with the permission of the Controller of HMSO and the Queen's Printer for Scotland.

Chapter 3 Achievement and underachievement (continued)

Annex D: The IQ Question
A Paper by Professor NJ Mackintosh and Dr CGN Mascie-Taylor
[pages 126 - 163]

Introduction: The nature of IQ tests

If a child does badly at school, lagging behind other children in learning to read, being assigned to lower streams or classes, failing exams and finally leaving school at 16 with few if any educational qualifications, it may seem only natural to say that the child was not good at school work, perhaps that he was not academically minded or was just not very bright. We may thus think that we have explained the child's performance at school by reference to his ability or capacity (or lack thereof). And if it could further be shown that the child also obtained low scores on standard intelligence or IQ tests, it might seem that this explanation had been confirmed: science would have documented the fact that the child lacked the intellectual ability needed for success at school.

Even when we are dealing with an individual child, there are reasons for questioning several of the steps in this argument. When we are dealing with large groups of children, the doubts multiply. It has long been known that children from working-class families are, on average, academically less successful than those from middle-class families. (1) Is this really because they are naturally less intelligent? And it has also been apparent for some time that children in this country from certain ethnic minorities tend to perform less well at school than do indigenous children. (2) Is this because they too, as a group, are on average less intelligent?

Even to pose the question will seem repugnant or insulting to many people. But it would be idle to pretend that no one has ever taken such possibilities seriously. In a celebrated article published in 1969, the American educational psychologist, AR Jensen noted that blacks in the USA were not only less successful at school than whites, but also scored significantly lower on standard IQ tests, and concluded that it is 'a not unreasonable hypothesis that genetic factors are strongly implicated in the average Negro-White intelligence difference.' (3) The British psychologist HJ Eysenck is characteristically blunter: 'All the evidence to date' he has written, 'suggests the strong and indeed overwhelming importance of genetic factors in producing the great variety of intellectual differences which we observe in our culture, and much of the difference between certain racial groups.' (4)

There is a whole network of assumptions underlying these arguments, some reasonable enough, others distinctly more questionable. Before considering some of the evidence, it will be as well to make clear what some of these assumptions are. The first, and one of the more questionable, is that there is something called 'intelligence' which can be accurately measured by an IQ test; more especially, that a child's IQ reflects a single capacity or even a set of abilities with which he is endowed by nature and which determines, or sets limits to, what he will achieve at school. The distinction is sometimes drawn between ability or potential, measured by an IQ test, and achievement or performance, measured by school exams.

This is no place for a detailed discussion of IQ tests, (5) and to be brief it will be necessary to be dogmatic. It is important to insist at the outset that an IQ test simply measures a sample of a person's behaviour at a given point in time, what he knows or has learned, how well he can solve certain kinds of problem. But school exams also purport to measure a child's knowledge and what he has learned, and will often test the ability to solve certain kinds of problem (consider a mathematics exam). The distinction between an IQ test and a school exam is not trivial, for the former is more likely than the latter to test knowledge that has not been formally taught at school and to require the child to solve problems rather different from any that he may have come across before. But it is a difference of degree, not of kind. An IQ score does not, indeed could not, provide any magically direct insight into a child's intellectual capacity divorced of all he has been taught or learned for himself. It measures his potential only in much the same way that (even if more accurately than) school exams also measure potential. Teachers and examiners assume that an exam result tells them not only what a child has learned, but also how well he is likely to do in the future. Scores on IQ tests will also predict a child's likely performance in other situations - for example how well he will do at school. In that sense, but in no other, they may be said to measure the child's potential.

A common assumption is that an IQ score reflects a child's potential in the sense that it measures a fixed, innate ability. It is true that IQ scores are relatively stable: a child's IQ at the age of 6 or 7 will predict quite well his IQ at 16 or 17. (6) But large changes are quite common. It is also probable that inheritance plays some part in determining IQ. That is to say, it seems probable that some of the differences in IQ observed in the population of the US or the UK are caused by genetic differences between members of these populations. How important genetic factors may be cannot be realistically determined from any available data. The claim of Jensen and Eysenck that no less than 80 per cent of the variation in IQ is genetically caused is not justified on the basis of published evidence, and recent estimates, based on more recent and better data, have given estimates of the order of 50 per cent. (7) But the most reasonable conclusion is that there are far too many problems inherent in all the data to justify this sort of precise, quantitative statement; the safest claim is that probably somewhere between one quarter and three quarters of the variation seen in IQ in most Western societies is due to genetic difference between members of those societies.

Perhaps the most contentious assumption underlying the whole argument, however, is that IQ tests could ever provide a fair measure of the intelligence of children from working-class families, let alone those from ethnic or racial minorities. Devised by white, middle-class psychologists, standardised on white children, validated by their ability to predict performance in white schools, IQ tests, it is argued, will inevitably reflect white, middle-class values, must be biased against other groups, and could not possibly provide a realistic assessment of their abilities. The argument may seem reasonable and persuasive. But it needs examination to disentangle what is possibly true from what is probably misleading. As has been justly remarked, criticising IQ tests 'for reflecting class differences is rather like blaming the weighing machine when it shows an undernourished child to be below weight.' (8) If a child has been deprived of intellectual stimulation or educational opportunity, it is small wonder that his intellectual performance will reflect this fact. An IQ test is no more able to gauge a child's true innate potential regardless of the circumstances of his upbringing than is a pair of scales to measure his true potential weight regardless of what he has been fed. To repeat: IQ tests measure a sample of a child's actual behaviour, what he knows and has learned. Some children may have lacked the opportunity to acquire the knowledge crucial for answering certain questions, just as a starved child may have been fed a diet lacking critical nutrients. To claim that IQ tests are biased is often only a misleading way of making the point that IQ tests measure skills and knowledge which not all children may have been able to acquire; in other words, that differences in IQ scores are partly due to differences in the environmental experiences of different children. But we already knew that.

What is commonly meant by the claim that an IQ test is biased against a particular group of children is that it does not reveal their true intelligence. But if this assertion is to carry any weight, it must mean that the IQ test provides a lower estimate of their intellectual performance than does some alternative measure. Bias is now a relative term. For example, one IQ test might be more biased against, say, children from working-class families than was another IQ test. Psychologists have indeed tried valiantly to construct 'culture-fair' tests which require less specialised knowledge, but no one would seriously argue that there could ever exist a test that required no such knowledge at all. Nevertheless there is reasonably good evidence that tests depending heavily on vocabulary produce larger differences between the scores of working-class and middle-class children and between those from large and small families than do so-called non-verbal IQ tests. (9) In other words, verbal tests are, by this criterion, more biased against working-class children than are non-verbal tests. But, contrary to widespread opinion, the differences between the IQ scores of blacks and whites in the US are, at least usually, no greater on verbal tests than on non-verbal tests. (10) Whether the same is true of any differences between West Indian and white children in this country is a question addressed below.

It is, of course, possible that all IQ tests are biased against working-class or black children - by comparison with some other measure. The problem, then, is to find the other measure. An obvious candidate might be performance on school tests or exams. But, again contrary to widespread opinion, there is no evidence that IQ tests are more biased against working-class children either in this country or in the US, or against black children in the US, than are conventional exams or measures of school performance. (11) The question whether they are biased against ethnic minorities in this country is considered below. But for these other groups, where any differences have been found they have almost invariably been in the opposite direction. That is to say, if a working-class and middle-class child have similar IQs, the chances are that the middle-class child will do better on school exams: the IQ test gives the working-class child a higher score than does the exam. Of course, one could argue, this shows only that exams are even more biased against working-class children. In this limited sense, it certainly suggests that by comparison with IQ tests, school exams are biased against working-class children. But that does not tell us whether IQ tests themselves are biased in their favour or against them.

The critic who claims that IQ tests do not give a true measure of the intelligence of, say, working-class children, must find some other, better measure of their intelligence. We have already argued that IQ tests cannot provide any magical insight into a child's true innate potential. There is no reason to believe that any other measure could achieve this miracle. Indeed, those who claim that IQ tests do not really measure intelligence must face up to the problem that is has proved remarkably difficult to find any other test that looks as if it might be measuring intellectual abilities but does not also give results that agree rather closely with existing IQ tests. To this extent, IQ tests, although no doubt very far from perfect, are as good a measure of intelligence or cognitive ability as we have.

Black-White differences in IQ

With no further preamble, then, let us turn to the question of immediate concern. Children from ethnic minorities do not necessarily do so well at school in this country as do indigenous children. Might this be because they are naturally less intelligent? More specifically, for example, is there any evidence that the average IQ of West Indian children in this country is lower than that of whites? (12)

In the US there has been substantial evidence available for 50 years or more that blacks on average obtain significantly lower scores than whites on standard IQ tests. (13) The difference is usually said to average about 15 points, although it varies considerably from study to study. (14) But this tells us little or nothing about the standing of West Indian children in this country: there have been, not surprisingly, far fewer studies comparing the IQ scores of West Indians and whites in the UK and they go back for not much more than 10 to 15 years. The results of four studies are shown in Table 1. The largest, by Yule, Berger, Rutter & Yule (15) which was published in 1975 (but for which the data was collected about 5 years earlier), obtained scores on non-verbal group IQ tests from approximately 14,000 white and 350 West Indian children in their last year at LEA primary schools. In addition, 105 white children and 100 West Indians were given individual IQ tests. These latter results are shown with the West Indian children divided into those born in the West Indies and those born in this country. It is apparent that the West Indian children have lower scores than the whites, although equally apparent that the differences diminish when the West Indian children are born, and therefore receive all their education, in this country. A second, smaller published study, by McFie and Thompson, (16) reported individual IQ scores for a group of West Indian children, aged 5-15 years and a group of white children matched by age and sex to the West Indian sample. The results are very similar, and again show that the longer the West Indian children have been in this country, the better they perform.

The final two sets of data shown in the Table have not previously been published; they come from two large national surveys. The National Child Development Study (17) has provided an exhaustive survey of all children born in a particular week in 1958. There are data for over 10,000 white children and for smaller numbers of various ethnic minorities. At the age of 11, these children were given various educational tests, including verbal and non-verbal IQ tests. The results show that the scores of the West Indian children for whom data are available are significantly lower than those of the white majority, being particularly low for those children resident in this country for less than 4 years. The Child Health and Education Study has provided a similar survey of all children born in a particular week in 1970; the children were given a variety of tests at the age of 10, including four sub-tests from the recently developed British Ability Scales. (18) All but one of the West Indian children identified in the survey had been born in this country, so the results are not broken down by length of residence in the UK. The results show that West Indian children obtained significantly lower scores than white children, but it is notable that the difference is rather smaller than in earlier studies, being less than 10 points on the two main verbal tests (definitions and similarities), just over 6 points on the non-verbal matrices test and only 3 points on the digit span.

There are differences between these various studies, both in the extent to which West Indian children lag behind whites and in whether the lag is greater on verbal or non-verbal tests. Such differences are hardly surprising, for the studies have differed widely not only in the samples of children studied, and the date when they were studied, but also in the type of tests employed. For example, the West Indian children in CHES are doing rather better that those in earlier studies. Their total IQ score of 92.4 is only 8 points below the white mean. This cannot be simply because they were all born in this country (compare their scores with those of the UK-born children in the Yule, Berger, Rutter and Yule study). It may imply that the position of West Indian children has indeed improved in the 10 years since the other studies were undertaken. But it is equally possible that it is the change in tests that is responsible for this apparent improvement. As it happens, the British Ability Scales, unlike the other tests employed in these studies, were standardised on a sample of children that would have included some children of West Indian and Asian origin. Although it does not follow that such children will necessarily do better on this test than on others, it is obvious enough that the only way to find out whether there really have been changes since 1970 in the standing of West Indian children is to administer the same test to random samples of children born in different years over the period between 1960 and 1974. Unfortunately no study has done this.

The differences between the studies reported in Table 1, however, should not be exaggerated. There is also a fair degree of consistency between them. There can be little doubt that on average West Indian children in this country obtain significantly lower scores on a variety of IQ tests than do white children. There is no suggestion that this poor performance is confined to verbal tests, although in three of the four studies it was certainly larger on the verbal than on the non-verbal tests. The overall difference is usually not as large as the 15-point difference said to hold between American blacks and whites and it is considerably smaller than this when the West Indian child has been resident in this country for any length of time. This is not entirely surprising, for there is evidence that blacks from the West Indies are more successful than other blacks in the US (19) It is important, therefore, not to generalise too readily from the American case. Nevertheless, when all is said and done, there is usually still a difference of about 5 to 12 points on a variety of tests between West Indian and white children in this country even when all children have been born here.

Genetic interpretations

Given that there does appear to be a real difference between the IQ scores of West Indian and indigenous children in this country, is there any reason to believe that it might be due to genetic differences between the two groups? How could one set about finding the answer to this question? It seems obvious that the only evidence that would be sufficient to prove that black-white differences in IQ, either here or for that matter in the US, were genetic in origin would be the demonstration of a difference in IQ scores between randomly selected groups of black and white children brought up in strictly comparable conditions. Given the nature of American and British societies, the experiment seems hardly feasible. It is difficult to believe that black and white children could ever experience truly comparable environments, and it would always be possible to appeal to some uncontrolled differences in their experience to explain away any remaining differences in their IQ scores. What is striking, however, is that in the three studies that have attempted to approximate these ideal conditions (albeit not very successfully), such differences have been extremely small. Unfortunately, only one of these studies is British. It looked at 4 to 5 year old children living in residential nurseries. (20) At this age, it is important to acknowledge, 'IQ' scores do not predict later IQ particularly well; it is not clear, therefore, whether the several measures taken of the children will have had much bearing on their IQ when they were older. It is also clear that neither the black nor the white children were a representative sample of the black and white populations of this country, since the occupations of their natural parents were more likely to be non-manual than is true of the population as a whole. In other words, the study is very far from ideal. Nevertheless, its outcome is impressive: the test scores of the 30 children, one or both of whose parents were black, were somewhat higher than those of the 24 children who had two white parents.

The two other studies are both concerned with American blacks. One looked at illegitimate children born to, and brought up by, German women, but fathered by American (and a small proportion of French) servicemen between 1945 and 1952. (21) Some of these servicemen were black, others white. This had no effect, however, on the IQ score of the children. The more recent study looked at children brought up by white adoptive families in Minnesota: (22) some of the children had two white natural parents, some had two black parents, some one black and one white. The difference between the IQ scores of the white and mixed-race children was small and statistically insignificant. The black children did have significantly lower scores (approximately 12 points lower) than either of the other groups, and this might suggest a genetic effect. But they had also been adopted at a later age than the mixed-race children and had been with their adoptive families for a significantly shorter period of time before IQ scores were obtained. Moreover, their adoptive parents were less well educated than those adopting the mixed-race children. In combination, as the authors noted, these environmental factors may be sufficient to account for the observed differences in IQ.

All three studies have numerous flaws: the samples are small and sometimes seriously unrepresentative; in the one British study, the IQ scores are questionable; in each case, some or all of the children are of mixed parentage, and in the Minnesota adoption study, the results from the children with two black parents are distinctly equivocal. Nevertheless, these are the only studies that even approximate to the ideal, and on balance there is little doubt which way they point: measured differences in IQ scores between black and white children in the US and between West Indian and white children in this country are probably largely environmental in origin. The data do not compel this conclusion, for they are very imperfect. But there is no other direct evidence to contradict it. So why should anyone have concluded otherwise? Why, for example, should Jensen have argued that the difference in average IQ between blacks and whites in the US is partly caused by genetic differences between the two groups? (23) Although there is no guarantee that his conclusions would apply to the difference between West Indians and whites in this country, they should not, for that reason alone, simply be dismissed out of hand. It is important to see what his arguments are and whether there is any reason to accept them. They are, in fact, largely indirect and amount to the assertion that there are no plausible environmental factors which could account for the observed differences in IQ. In particular, Jensen has stressed three points:

1. Differences in IQ between American blacks and whites are not just due to differences in 'socio-economic status' (SES), since even when one compares children from families of similar SES, there is still a large difference in IQ. In California, indeed, Jensen has claimed that black children from high SES families have marginally lower IQs than white children from low SES families.

2. The difference between American blacks and whites is very much larger than that observed between whites and other ethnic minorities in the US (e.g. American Indians and Mexicans) who are on average even poorer and hold worse jobs than blacks; while some ethnic minorities (notably people of Japanese or Chinese origin) obtain higher scores than the white majority on non-verbal IQ tests.

3. The difference is not just due to bias in the tests. In the first place, it holds up, and indeed is often greater, on supposedly unbiased non-verbal IQ tests than on possibly biased verbal tests of intelligence. Secondly, Jensen argues, there is no evidence that IQ tests in general are particularly biased against blacks in the US, for they predict academic achievement for blacks just as well as they do for whites.

The environment

Little enough is known about the environmental factors that affect IQ in any population - let alone those that might be responsible for differences between ethnic groups or classes within that population. It is known, for example, that differences in IQ are correlated with differences in social class or SES, but many psychologists (Jensen and Eysenck to name but two) have argued that these may partly reflect genetic rather than environmental differences between the classes. It is known that certain demographic variables, such as family size, and other related factors such as birth order, are correlated with IQ, but although these effects are almost certainly environmental in origin, it is not certain just how they are produced. (24)

For our purposes, one critical question is whether differences between ethnic groups may be accounted for by appeal to the same sorts of factors as those supposedly responsible for differences within, say, the majority group, or whether it is necessary to appeal to some unique factors affecting the cognitive or scholastic performance of an ethnic minority. It is perfectly clear, for example, that West Indians in this country and blacks in the US are, on average, poorer and occupy jobs of lower status that the white majority; (25) since these factors affect IQ in the white population, it seems entirely probable that they may also help to explain the lower test scores of West Indian children. There are other, equally well established, differences in their backgrounds. A higher proportion of West Indian that of white families contain only a single parent; a higher proportion of West Indian than of white mothers go out to work; and a higher proportion of West Indian than of white children are looked after by child-minders. (26) Is it not possible that some of these differences are related to, perhaps even cause, the observed differences in the children's IQ scores?

Jensen's first argument was that differences in IQ between blacks and whites in the US cannot be explained this way. Paradoxically enough, his fiercest critics accept this initial premise, although clearly not the conclusions he draws from it. A number of writers, both in this country and in the US, have rejected all attempts to explain differences between blacks and whites by appeal to the family background of the blacks. (27) Such explanations are said to imply that there is something 'pathological' about black family life or culture, and are seen as no less insulting than Jensen's own suggestion that differences in IQ are genetic in origin. These writers have seen a racist society rather than a pathological family background as the major cause of these differences. Black children are said to lack self-esteem: constant exposure to racial slurs and low expectations (for example from teachers) eventually ensures that black children will live down to these expectations and perform poorly both in IQ tests and in school in general. The argument, in other words, is that there is something unique about the experience of children from some ethnic minorities that is directly responsible for their relatively poor performance on a variety of tests.

That certain ethnic minorities in this country (and in the US) suffer from racial discrimination will probably not be disputed. The question is whether such discrimination has a direct effect on a child's performance in school. An alternative possibility is that racial discrimination is a major cause of the impoverished social circumstances of their parents and affects the children's IQ scores only indirectly, by ensuring that they grow up in the sorts of circumstances that contribute to a low IQ score in the indigenous majority also. In what follows, therefore, we shall attempt to see first, how far differences between blacks and whites can be accounted for in terms of the sorts of environmental factors that are thought to affect IQ in the white population; secondly, whether any of Jensen's other indirect arguments have any force when applied to the British case; and finally, whether there is any evidence to support the argument that racial discrimination provides a unique factor directly affecting the IQ scores of black children in this country.

Class and family background

Jensen's first argument was that black-white differences in IQ are not simply a consequence of differences in SES, for high-SES blacks actually have rather lower scores than low-SES whites. Equating for SES, he has claimed, reduces what is otherwise a 15 point or greater difference in IQ by no more than 2 or 3 points. (28) This is consistent with data surveyed by Shuey in 1966: (29) from 13 studies undertaken between 1921 and 1964, she concluded that the average difference in IQ between blacks and whites of similar SES was 12.80 points. But many of these studies are too flawed to be taken seriously: the attempts to equate socio-economic status were frequently crude and clearly unsuccessful. (30) Moreover, Shuey also analysed a further 14 studies carried out between 1922 and 1958 which had looked at blacks and whites living in similar neighbourhoods, and found that the average difference was now reduced to 8 points. And there are other more recent American studies which have obtained results wholly at variance with Jensen's claim and Shuey's summaries. In one large and careful survey carried out in Boston, Philadelphia and Baltimore, (31) individual IQ tests were given to a total of over 15,000 black and white children from either high or low SES families: differences between blacks and whites of similar SES averaged between 4 and 6 IQ points - considerably less than the 9 or 10 points separating children from the high and low SES backgrounds.

Much the same appears to be true in this country. The data from the National Child Development Study, shown in Table 2, reveal a 15 point difference between West Indian and white children in total IQ. But, as can be seen, this difference is immediately reduced to 11 points by adjusting for the fact that a very much smaller proportion of West Indian than of white children have fathers with non-manual jobs (the actual proportions are shown in Table 3). Table 2 also shows the results of a second British study by Grace who gave IQ tests to 11 year old indigenous and West Indian children attending the same, inner-city schools, and therefore approximately equated for the neighbourhood in which they lived. (32) It is clear that the difference in IQ scores is very much less than most of those shown in Table 1.

Father's occupation and neighbourhood are hardly the only factors known to be related to a child's IQ. If equating black and white children on variables as simple as these has such an effect on the difference in their IQs, it seems possible that a more comprehensive attempt to match them for potentially important environmental variables might well reduce the difference to quite trivial proportions. And so it has seemed. (33) Two studies published over 10 years ago found essentially no differences between the IQ scores of West Indian and white children when some of the more obvious disparities in social circumstances between them were reduced or controlled. (34) Their results are also shown in Table 2. Houghton studied 5 year old West Indian and white children living in a decaying city centre, attending the same schools and with similar nursery school experience. All but two of the West Indian children had either been born or lived at least two years in this country. Although the average IQ scores were low, the difference between the two groups of children was trivial and statistically insignificant. Bagley deliberately selected West Indian children from relatively advantageous backgrounds, by excluding those whose father was absent or unemployed, those with three or more brothers and sisters or living in crowded conditions, and by requiring that the child should have attended British schools since the age of 5 (the children were 7-10 years old at the time of the study). He matched the 50 West Indian children so obtained with 50 English children in terms of age, sex and father's occupation. As can be seen, both groups of children had slightly above average IQs, with the West Indians doing very slightly better.

Neither of these studies is very large and their conclusions should not necessarily be accepted without question. In Houghton's study, for example, the children were only 5 years old at the time of testing, and one can wonder about the social composition of a white population left behind in a decaying city centre. Bagley's West Indian children were an admittedly highly selected and unrepresentative sample. In order to obtain more evidence on this crucial question, we have undertaken analyses of the National Child Development and Child Health and Education studies whose summary data were shown in Table 1. Although neither study contains particularly large groups of West Indian children, they have one virtue almost unique in this area. The children were selected, on the sole basis of the date of their birth, from the entire country. They constitute, therefore, a reasonably representative and, for our purposes, randomly chosen sample of the population. Both studies, moreover, provide information on the child's family and social circumstances and thus enable one to form some idea of the relationship between social circumstances and test scores. There are numerous factors besides parental occupation that correlate with IQ scores in the white samples and in terms of many of these factors West Indian children can be said to suffer disadvantage. Our analyses suggest indeed that, much, although by no means all, of the original differences in overall IQ between West Indian and white children can be accounted for in terms of a relatively small number of social variables.

A measure of the social disadvantage suffered by the West Indian children is shown in Table 3. (35) In the NCDS sample, for example, a significantly higher proportion of West Indian than of white households contain four or more children, occupy crowded accommodation that involves sharing basic amenities and suffer from serious financial problems; West Indian fathers are more likely than white fathers to have manual rather than non-manual jobs or to have been unemployed for part of the preceding year, and more families have no male head. In the CHES sample, not surprisingly, data are not always available for exactly the same set of factors, but as can be seen from the table, West Indian fathers are still more likely to have manual rather than non-manual occupations; West Indian families are larger and poorer than white families, more likely to have no male head, and to live in inner cities or council estates.

In each study, all of the factors listed in Table 3 are significantly related to children's IQ scores in the population as a whole and are likely therefore to be related to the average difference in IQ scores between West Indian and white children. Many of them, of course, tend to go together: other things equal, households with large numbers of children are more likely to be overcrowded; unemployment tends to produce financial difficulties. It is obviously inappropriate therefore to add the contribution for each factor independently. But the statistical technique of multiple regression allows one to take account of the effect of one variable having made allowance for that part of its effect which it shares with other variables.

The results of a series of multiple regression analyses of the NCDS and CHES data are shown in Table 4. The data are broken down by sex and by type of IQ score (verbal and non-verbal in NCDS and the four sub-tests in CHES), and for the NCDS data there is a separate analysis for those children who had been in the UK for four years or more. Several points are immediately apparent. When the factors listed in Table 3 are taken into account, the differences between West Indian and indigenous IQ scores are considerably reduced, although they are not completely eliminated. Moreover, this general conclusion covers a great deal of variation within and between each of the two studies.

In the NCDS analyses, differences of over 10 points in both verbal and non-verbal IQ between West Indian and indigenous males are reduced to differences of less than 5 points; in the case of females, however, the residual difference, especially in non-verbal IQ scores, is considerably larger. If the analysis is confined to those children who had been in the country for four or more years before being tested, all these differences are reduced without much change in the general pattern. In the CHES data, the original unadjusted differences between West Indian and indigenous children are rather smaller than in the NCDS data, at least for the matrices and digit span tests. And the regression analysis, although accounting for a somewhat smaller proportion of the overall differences, thus leaves rather smaller residual differences, at least for these tests. The differences between the sexes that was so apparent in the NCDS data have essentially disappeared in the CHES data - indeed girls do rather better than boys on the matrices test.

Another way of looking at the effect of these social variables, and one that takes advantage of the very much larger number of indigenous than of West Indian children in these studies, is to select individual white children who match individual West Indian children in terms of the factors listed in Table 3. Thus if a particular West Indian boy lives in a household containing more than four children, has a father with a manual job who has been unemployed for part of the past year, has attended more than two schools etc, one searches for a white boy in the sample with the same set of characteristics. We have undertaken such an analysis for both the NCDS and CHES samples. In neither case is it in fact possible to find matches for all West Indian children for whom complete data were available (an indication, perhaps, of the discrepancy between the social circumstances of West Indian and white children). In NCDS it was possible to match only 50 of the 72 West Indian children; in CHES, 77 out of 92 children were matched. In NCDS, the average total IQ score of the matched indigenous children is only 5.6 points greater than that of the West Indian children, a difference that reduces to 5.2 points if the analysis is confined to the 40 West Indian children who entered the country before the age of 7. In CHES, the difference in total IQ between the West Indian and matched indigenous children is only 2.6 points.

These various analyses show that it is possible to account for a significant proportion of the difference in IQ scores between West Indian and white children. But in no case was the difference completely abolished. This may cast some doubt on the generality of the conclusions suggested by the Houghton and Bagley studies shown in Table 2. It may even suggest the possibility that the differences unaccounted for by the analyses must be of genetic origin. One immediate problem with that conclusion is to decide which, if any, of these differences is to be accepted. The suggestion that there is a residual core of genetically determined difference in the intelligence of West Indian and white children hardly prepares one for the way this difference varies in such a capricious manner from less than 3 points in some cases to nearly 10 in others. If one were to argue, for example, that non-verbal tests provide a potentially less biased measure of intelligence than verbal tests, then in the CHES data at least, the residual differences in the digit span and matrices tests were small and statistically not significant.

There are, however, rather more general reasons why the conclusions that can be drawn from a multiple regression analysis of this sort are strictly limited. It is obvious enough that large-scale surveys will probably neither identify all the social and environmental factors that can affect IQ (why should they be only social, economic or demographic?), nor measure the ones that have been identified very precisely. It is entirely possible that the residual differences in Table 4 could be reduced yet further by including more variables in the regression analysis. For example, in the NCDS sample, one additional factor that has a highly significant effect on IQ scores is the teacher's rating of the child's 'social adjustment'. West Indian children obtain significantly lower ratings than white children, and when this factor is included in the analysis it reduces the residual differences in Table 4 by between 2 and 3 IQ points. Similarly, in the CHES survey, children were asked to complete questionnaires designed to measure the extent to which they attributed success or failure to their own efforts (or lack of them) or to external factors beyond their control. Here too, there were significant differences between the scores obtained by West Indian and white children, and since these scores were related to IQ scores in the sample as a whole, their inclusion in the regression analysis again reduces the residual differences in Table 4 by anything up to 2 points.

If this suggests that Table 4 may be overestimating the importance of genetic factors, there are equally good reasons for supposing that it might underestimate them. It is one thing to show that differences in IQ scores between two groups are related to differences in social circumstances between them. But is is quite another matter to show that the differences in social circumstances cause the differences in IQ. The discovery of a correlation between the two variables does not prove that one causes the other. Consider, for example, the effect of the teacher's assessment of a child's social adjustment in the NCDS data. Low ratings of social adjustment go with low IQ scores, and West Indian children obtain lower ratings than do white children. Have we therefore identified one cause of the difference between West Indian and white IQ scores? It is not difficult to think of other explanations of the observed pattern of correlations. Perhaps, for example, some teachers are inclined to label children as maladjusted (a euphemism for behaving badly in school) simply because they are West Indian. The label does not cause a low IQ independently of skin colour. It is simply a reflection of skin colour and in itself irrelevant to IQ.

There is, moreover, reason to believe that genetic factors contribute to differences in IQ within the white population. Perhaps, then, unfortunate social circumstances do not cause a low IQ; they are themselves a consequence of a genetically determined low IQ. The possibility that such genetic effects contribute to the observed correlation between social circumstances and IQ cannot be dismissed. But the importance of this contribution depends, among other things, on one's guess as to the heritability of IQ in the population as a whole. There is good reason to believe that this may be very much less than the figure of 80 per cent espoused by Jensen and Eysenck, and this would mean that the genetic contribution to the correlation between social circumstances and IQ was unlikely to be of overriding significance. Its importance, moreover, is likely to be greater for some circumstances than for others. Parental education, for example, is correlated with a child's IQ score, but it will also be related to the parent's own IQ score, and parent's IQ is surely the one factor most likely to be genetically related to a child's IQ. But parental education was not one of the variables entered into these regression analyses. Father's occupation was, of course, entered and both Jensen and Eysenck (in common with many other psychologists) have argued that differences between occupations or social classes are a consequence of genetic differences in IQ. They may be right, although it is worth insisting that there is virtually no direct evidence to support this assumption. And it does not take much imagination to see that the reason why many West Indian adults hold poor jobs is at least as likely to be discrimination on the part of the employers as genetic inferiority on the part of West Indians. (36)

The importance of the analyses shown in Tables 2 and 4, therefore, does not lie so much in the absolute size of any residual, unexplained difference between West Indian and white scores. In some cases, that difference remains quite large; in others it has vanished completely. But it is impossible to be certain whether the role of social factors has been underestimated, because the analysis failed to measure all relevant variables, or whether their importance has been overestimated, because the analysis included factors which themselves reflect genetic differences. The real importance of these analyses is that they suggest that there are indeed social circumstances which predict a significant proportion of the difference between average West Indian and white IQ scores and that they are exactly the same as those which predict differences within the white population. This simultaneously undermines both Jensen's position and that of his sternest critics. It cannot disprove a genetic hypothesis, but when combined with the direct evidence considered earlier it makes it even less likely that genetic factors are an overriding cause of black-white differences in IQ: whatever may be the case in the US, it is simply not true to claim that no known environmental factors can account for more than a small fraction of the difference between West Indian and white IQ scores in this country. And since the factors which have been identified are those that contribute to differences within the white population, there is that much less reason to believe that there is something unique about the experience of West Indian children in this country which causes their low IQ scores. If racial discrimination is responsible for their poor performance it seems more likely that this is due to its effects on their parents' social circumstances.

Comparison with other ethnic minorities

Jensen's second argument was that the poor performance of American blacks on IQ tests cannot be attributed to their social circumstances, since other ethnic groups such as Mexican immigrants or American Indians, who are even more impoverished, obtain higher IQ scores. Without examining this claim in more detail, it is worth asking whether there are relevant comparisons to be made between West Indian children in this country and children from other ethnic minorities.

The second major group of recent immigrants to the UK has been that from the Indian sub-continent. If only because of the obvious problems raised by their not all speaking English as a first language, there has been for a long time a concern for the educational attainment of their children. But evidence becoming available in the mid-1970s suggested that, although they might find considerable initial difficulties, these children rapidly adapted to British schools and performed relatively well on standard tests of reading and mathematics, as well as on public examinations such as CSE and O Levels. (37) There is rather less information available on their performance on IQ tests, but Table 5 shows the results of two earlier studies, along with analyses of the NCDS and CHES surveys. (38) Yet again, there is considerable variability from one study to another, but some common trends can be discerned.

The most notable conclusion is that length of stay in this country has a very marked effect on performance. In the three studies where data are available, there is a striking and consistent improvement in the scores obtained when children have been resident in this country for a reasonable length of time and, presumably, have received most of their education here. This is as true for a supposedly 'culture-fair' test like Raven's Matrices as it is for verbal tests. The second impression is that children who have been long resident in the UK usually obtain scores well within the normal British range. In several cases, indeed, there is essentially no difference between the scores obtained by long-resident Asian children and those of indigenous children. The most serious exception to this generalisation is to be found in the CHES data where children from Pakistan obtain consistently low scores, and both Indian and Pakistani children do particularly badly on the two verbal tests (definitions and similarities). (39)

The CHES data suggest that earlier complacency about the performance of Asian children may have been somewhat misplaced. As a comparison with Table 1 will reveal, these CHES data show that Pakistani children do not obtain average scores any higher than those of West Indian children, and even Indian children are not consistently better. In this, they contrast rather strikingly with data from the NCDS survey, where long-resident Asian children scored some 8 to 10 points above the West Indian children. It is not at all easy to say why this might be. The number of long-resident Asian children in NCDS is very small and they may have, by chance, been an unrepresentative sample. This does not seem entirely likely, for the NCDS data agree tolerably well with those of the other studies shown in Table 5. It should also be remembered that CHES and NCDS employed quite different sets of tests, and that the two studies tested the children at slightly different ages (10 in CHES and 11 in NCDS). The possible importance of the former factor should not need emphasis: this is not the first time that we have seen differences between various groups come and go depending on the type of test employed. Although there was only a one year difference in the age at which children were tested, it is conceivable that this might have been important. There is evidence that both Indian and Pakistani children may improve by comparison with West Indians between the ages of 8 and 12. (40)

A final possibility is that the social circumstances of Asian immigrants have changed over the 12 years separating the two studies. There is some evidence to support this. The data for the NCDS sample have been analysed by others. (41) In terms of the sorts of factors considered in Table 3, the families of Asian children born in this country were substantially more advantaged than those of West Indian children and did not suffer by comparison with indigenous families. The CHES data tell a different story. Although there are some differences between Indian and Pakistani families, the general tendency is for both groups to resemble West Indian rather than indigenous families in terms of the proportion of fathers in manual occupation, number of children in the family, family income, free school meals and neighbourhood. The actual proportions are shown in Table 6. (42)

The importance of these factors is suggested by the results of a series of regression analyses, similar to those undertaken on the West Indian data, whose results are also shown in Table 6. The effects of these analyses are in general more marked for Pakistani than for Indian children, since their social circumstances are in general poorer. From having started with scores lower on average than those of the Indian children, the Pakistanis often end up, after the regression analysis, with slightly higher scores. But the more important comparison is with the West Indian children whose results were shown in Table 4. By and large, the two Asian groups produce a pattern of results very similar to that observed in the West Indian case. The regression analysis significantly reduces the gap separating all three groups from the indigenous majority, and although the residual differences for the two verbal tests remain considerable, those for the digit span and matrices tests are small and statistically insignificant. The only notable departure from the West Indian results is that the sex difference observed in the matrices scores in West Indian children is reversed: in the present case it is the males, rather than females, who do better. One could speculate about the cultural factors that might be responsible for this difference, but in the absence of further data it would remain speculation. It does not, however, seem likely that the difference reflects any genetic, sex-related differences between the different ethnic groups.

Two conclusions can perhaps be drawn from this discussion. First, there is some suggestion that children of Indian and Pakistani immigrants may no longer be obtaining higher test scores than those of West Indian immigrants, and if true this might be because their social circumstances are no longer as markedly superior as they used to be. This must be a tentative conclusion, to be regarded with considerable suspicion unless it can be confirmed by other, larger surveys. Secondly, however, one can assert with some confidence that the comparison of West Indian and Asian children in this country provides little support for the sort of argument proposed by Jensen on the basis of his comparisons of blacks with other ethnic minorities in the US. The present comparison suggests that where there is a difference in the IQ scores of West Indian and Asian children, there is also a demonstrable difference in their social circumstances. Where their circumstances are similar, there is no significant difference in measured IQ. There is no evidence for the combination of superior IQ and poorer circumstances which Jensen claims to find in his American comparisons. (43)

Test bias

Jensen's final argument was that the poor performance of American blacks on IQ tests cannot be due to bias in the tests themselves. The argument rests on two claims. The first is that blacks, unlike other minority groups, obtain low scores not only on possibly biased verbal IQ tests, but also on less biased non-verbal tests such as Raven's Matrices, sometimes indeed performing even worse on the latter than on the former. The second is that there is no reason to accept that IQ tests are in general biased against blacks for they predict school achievement as well for blacks as they do for whites. Indeed, he has argued, if there is any difference it is that IQ tests overestimate the scholastic attainment of blacks. The relationship between IQ and school performance among ethnic minorities in this country raises a number of wider issues which it will be more convenient to discuss at a later point in this paper. We can anticipate one of the conclusions of that discussion here by noting that the picture is on the whole fairly similar to that reported by Jensen for the US. By and large, IQ scores do not seriously underestimate the performance of West Indian or Asian children at school. In some cases, the reverse is true.

This is worth noting, but it may not be sufficient to establish the conclusion which Jensen wishes to draw. The fact that blacks do at least as badly at school as their IQ scores would lead one to expect may prove that IQ tests are not more biased than are other measures of school achievement. But it is surely possible that both sets of measures might be biased against blacks. It is this possibility that Jensen seeks to dispute by arguing that in the US blacks perform badly not only on verbal IQ tests which clearly require knowledge of the majority culture, but also on a variety of non-verbal or 'culture-fair' IQ tests which, it is to be supposed, do not require such knowledge.

The argument will not convince everyone, for it rests on the unproven assumption that because a particular IQ test employs non-verbal, diagrammatic and abstract material, it no longer requires culturally specific knowledge or skills. Jensen's favourite example of such a test is Raven's Matrices. But two of the studies reported in Table 5 show differences of between 10 and 20 points on this test between Asian children who have and those who have not resided for any length of time in this country. The ability to solve abstract non-verbal problems may depend as much on education and experience of a particular type as does the ability to answer questions about the meanings of English words. It is only a rather simple-minded view of intelligence and of differences in cultural tradition that could blind one to this possibility.

But it is not necessary to rely on these general arguments. The question at issue is whether there is any evidence that West Indian children in this country, unlike other ethnic minorities, obtain low scores not only on verbal but also on non-verbal IQ tests. As usual the data are fragmentary and somewhat less clear than one might wish. But the answer they suggest is surely no. In the two previously published studies shown in Table 1 (Yule, Berger, Rutter & Yule; McFie & Thompson), West Indian children obtained rather higher scores on non-verbal than on verbal tests, regardless of their length of stay in this country. In the one previously published study shown in Table 5 (Ashby, Morrison & Butcher) Asian children resident less than 4 years in the UK obtained higher non-verbal than verbal scores, but those here more than 4 years obtained slightly higher verbal scores. The only two studies that permit a direct comparison between West Indian and Asian children are the National Child Development and Child Health and Education Studies. The former found no difference between verbal and non-verbal IQ scores in either group. In the latter, both groups did better on the non-verbal matrices test than on the two main verbal tests (and best of all on the recall of digits).

Relative performance on verbal and non-verbal tests clearly depends on the precise tests employed. The most common result appears to be for ethnic minorities to obtain lower scores on verbal tests, a finding entirely consistent with the suggestion that they are more biased than non-verbal tests against minority groups. But the difference is neither large nor consistent. And there is certainly no suggestion that it is confined to Asian children. If recently arrived in this country, such children will often not speak English as a first language (if at all). It would hardly be surprising if they were at a particular disadvantage on verbal tests. The fact that there is no good evidence of any difference between the relative performance of West Indian and Asian children on verbal and non-verbal tests, therefore, provides striking disconfirmation of Jensen's supposition that blacks necessarily differ from other ethnic minorities by scoring poorly on both kinds of test. (44)

Conclusions

We have considered the three types of indirect argument put forward by Jensen to support his claim that the difference between black and white IQ scores in the US is probably largely genetic in origin, without finding much reason to accept that any of them applies to the case of West Indian children in the UK. First, it turns out to be quite easy to find social and environmental factors that account for a significant part of the average difference between West Indian and white IQ scores in this country. Secondly, there is no evidence that West Indian children obtain lower IQ scores than other ethnic minorities suffering from comparable levels of social and economic deprivation. Finally, there is no reason to suppose that West Indian children differ from Asian children in the pattern of their scores on verbal and non-verbal tests, for both tend, if anything, to obtain rather lower scores on verbal tests. If non-verbal scores are taken as the less biased measure of intelligence, the differences between both minority groups and the indigenous majority are often quite small. The implication of this indirect evidence, then, is much the same as that of the direct evidence considered earlier. If there are genetic differences for IQ between various ethnic groups in this country, they are likely to be extremely small.

Racial discrimination

It is time to turn to Jensen's critics, to those who have argued that the poor performance of black children on IQ tests cannot be attributed to their family background, but would rather lay the blame directly on their experience of a racist society. As we have already seen, to the extent that social variables related to differences in IQ within the white population are also related to differences in IQ between West Indian and white children, it is unnecessary to appeal to factors unique to the experience of West Indian children to explain their low IQ scores. If racial discrimination is affecting their test scores, it must be doing so indirectly through its effects on the social status and circumstances of their parents.

In practice, however, the regression analyses reported above did not completely abolish the differences in IQ scores between West Indian and white children, but left some sizeable differences unaccounted for. It is possible that some of these further differences are due to the direct effects of discrimination on the children themselves. It is certainly not unreasonable in principle to suggest that the attitudes of a hostile and contemptuous white majority have so affected the self-confidence and ambitions of West Indian children that they are unable or unwilling to succeed in school or obtain high scores on IQ tests. The question is whether there is any evidence for this.

The first requirement of any comprehensive account of ethnic differences in scholastic performance is that it should be able to explain why children of Asian origin by and large do better than those of West Indian. (45) On the face of it, the suggestion that the Asian community in this country suffers less racial discrimination than the West Indian might seem hard to defend. But it is obviously necessary to distinguish between different forms of discrimination: our present concern is with attitudes towards the intelligence of an ethnic group, not whether they are attacked in the streets by hooligans. And there can be little doubt that teachers have markedly different views of the abilities of Asian and West Indian children - as the following unguarded remarks show:

'West Indians are boisterous and less keen on education than Asians. This is well known obviously.'

'They are slower than Asian children - not as bright.' (46)

Teachers' expectations

But the suggestion that the attitudes of society at large and of teachers in particular may have deleterious effects on the performance of West Indian children is quite plausible. Intuitively, one can readily see how constant denigration, whether overt or more subtle, might sap a child's confidence in his own abilities and cause him to fail. And the suggestion has the backing of received, or at least of popular, psychological opinion. The notion of the self-fulfilling prophecy was given wide currency by Rosenthal and Jacobson's study, 'Pygmalion in the Classroom'. (47) Rosenthal and Jacobson claimed to have shown that children's IQ scores could increase dramatically if their teachers were led to believe that a special test administered to all children in class had identified these particular children as 'late bloomers' due to show a marked spurt. Although the study has achieved widespread fame, most of those who have examined it closely have remained unconvinced. (48) The increase in IQ scores occurred in only two of the six grades tested (that is, in only 19 children). The IQ tests were administered by the teachers themselves rather than by a trained tester - leading one to wonder who was fulfilling the prophecy. And some of the scores obtained, which ranged from zero to 300, must mean either that it was not being administered properly or that the test itself was an absurd one. Moreover, although several ingenious and better-controlled studies have established quite clearly that a teacher's expectations can have a significant effect on a child's general performance, no subsequent study has replicated Rosenthal and Jacobson's finding that IQ scores can be so affected. (49) This is an important conclusion, for it suggests that although one might be able to blame teachers' attitudes for the generally poor school performance of West Indian children, it is less likely that such attitudes are responsible for their low IQ scores. And to the extent that IQ is related to school performance, of course, this in turn suggests that prejudice is in fact unlikely to be the sole cause of their poor school performance.

This conclusion is based on the failure of several, no doubt somewhat contrived, experimental studies to demonstrate a direct effect of teachers' attitudes and expectations on their pupils' IQ scores. It is possible, of course, that the attitudes towards West Indian children actually held by teachers in British schools and the expectations they develop are both subtler and more powerful determinants of their pupils' performance. All one can say is that there is relatively little direct evidence that convincingly establishes such an effect. Measured IQ scores may be rather less malleable than is sometimes supposed. Certainly, in the US, there is now a considerable body of evidence to contradict the common assertion that blacks obtain low scores on IQ tests because they are tested by white testers. Although a few studies have found that children obtain higher IQ scores when tested by an examiner of their own race, others have reported exactly the opposite result, and the majority have found no difference either way. (50)

Motivation and self-esteem

But there are still other possibilities. A person's performance on an IQ test can be significantly affected by his attitudes and motivation. There is evidence from the US that blacks and whites alike obtain lower scores if they are too anxious (they may also perform less than optimally if they are too relaxed and do not take the task seriously). It would hardly be surprising if some black children perceived some situations as more threatening than white children do. They might therefore feel too anxious in a situation where white children are performing at their best. But although these effects are sometimes statistically significant, they are rarely large, and invariably insufficient to eliminate differences between black and white children. (50) One can only conclude that if this sort of factor really is important in the real world, then it has not been satisfactorily captured in the artificial setting of an experimental study. That is not, perhaps, too surprising, but it follows that it is only an act of faith that allows anyone to assert that 'test anxiety' is the real cause of blacks' poor performance.

Anxiety is not the only motivational factor which has been thought to contribute to poor performance. It has been suggested by some writers in this country that West Indian children lack confidence in their own abilities, have low expectations or, in psychological jargon, poor 'self-esteem'. More particularly, it is possible that West Indian children who do badly at school have especially low self-esteem. One study identified a group of West Indian children whose scores on standardised reading tests were as good as those of white, middle-class children, and showed that their family background and attitudes were markedly more independent, self-reliant and indeed hostile towards white society than were those of other West Indian families. (51) Unfortunately, the two groups of families also differed in social class, the former being distinctly more middle-class than the latter, and it is at least possible that other attributes of middle-class family life were responsible for the superior performance of their children. Another study reported that differences in self-esteem between West Indians and whites were confined to boys: West Indian girls were as self-confident as their white counterparts. (52) It is sometimes claimed that West Indian girls do better at school than boys - and even that they achieve higher scores on IQ tests, but there are numerous studies which have found no such effect, and some which have not confirmed the original differences in self-esteem. (53)

The evidence is far from persuasive, and the picture looks even less convincing when other studies are considered. It has been shown that West Indian children born in the West Indies may have higher self-esteem than those born in this country. (54) But we have seen that it is the latter group who score higher on IQ tests and do better at school. There is some evidence that increases in the proportion of immigrant pupils in a school may improve the self-esteem of immigrant pupils, but no evidence that it will improve their scholastic record. (55)

It is not even clear that West Indian children in general have lower self-esteem than whites. Some studies have shown such a difference, others the reverse. (56) The numerous conflicts in evidence in this area of research may reflect genuine differences in attitude between different groups: attitudes may well have changed, for example, over the last 20 years; some West Indian children, in some circumstances, may well be more self-confident than others placed in different circumstances. But one suspects that part of the problem may arise from the ill-defined nature of the concept of self-esteem.

Even if the relationship between measures of self-esteem and IQ or scholastic performance were very much clearer than they are, we could still not assert that low self-esteem had been shown to be the cause of poor performance. Once again it might simply be a reflection of that performance. The child who, for whatever reason, is doing badly at school may well be inclined to agree with such statements as 'I'm not doing as well in school as I'd like to' or 'I often get discouraged at school'. These are two of the items to be found in the most widely used self-esteem inventory.

Very much the same could be said about attitudes to school itself. It seems probable that children who are successful at school might also like school more than those who are not. This would not be surprising but would hardly prove that their success was caused by their favourable attitude. In fact, there is little reason to believe that West Indian children do have particularly unfavourable attitudes to school. Certainly, their parents commonly show significantly more interest in, and concern for, their children's education than do white parents of comparable socio-economic status. (57) And there have been several reports showing that West Indian children themselves express attitudes to school at least as favourable as those held by indigenous children. (58) As we shall show below, they obtain rather more total CSE and O Level passes than indigenous children with comparable IQ scores or in comparable social circumstances, and it is possible that this partly reflects the greater value they place on school achievement.

Changes in performance over time

Although it seems entirely plausible to suppose that a child's IQ score and performance at school should be affected both by his own attitude to school and by his teachers' attitudes towards him, it is difficult, particularly in the case of IQ scores, to find evidence to prove that such an effect is operating to depress the achievement of West Indian children in this country. There remains one further line of evidence that may bear on this issue. It has been suggested that West Indian children, unlike those from other ethnic minorities, fall progressively further behind their white contemporaries as they go through school. What starts as a relatively small difference between the various groups at the age of 7 or 8 it has been claimed, increases, in the case of West Indian children, to a much larger difference at the age of 12 to 16. Such a failure to keep pace with their contemporaries would suggest that West Indian children suffer from some environmental disadvantage: an obvious possibility is that this disadvantage is the prejudiced attitude of school and society. If there were widespread evidence of such a relative decline in the performance of West Indian children, therefore, this might strengthen the case for believing that they suffer from discrimination, for example at the hands of their teachers, and that this discrimination is one factor directly contributing to their poor performance. In fact, the evidence, particularly in the case of IQ scores, is far from convincing.

One recent study, widely reported in the press, is said to have shown that 'West Indian children are failing in school to such an extent that their intelligence scores go into a sharp decline between the ages of 8 and 12'. (59) This is a remarkable interpretation to put on the findings actually reported. At the age of 8, the IQ scores of West Indian and white children from the same schools were 98.3 and 106.7 respectively, while at the age of 12 they were 98.7 and 107.8. There is no suggestion that West Indian IQ scores have declined and the gap between them and white children has increased by a trivial 0.7 points. The claim that they have declined is based on the results of a somewhat misleading regression analysis which also included data from Indian and Pakistani children in the same schools. These children, although initially doing no better than the West Indians, did indeed show striking gains in IQ scores, of 10 and 8 points respectively, from the ages of 8 to 12. Their inclusion means that, on average, children with low IQ scores at 8 had higher scores at 12 and it is, in effect, by comparison with this average that West Indian children are being said to show a decline. But this decline appears to be only by comparison with Asian children, who are making striking gains (the importance of which has already been alluded to) not by comparison with whites. (60)

The results of an analysis of scores obtained on tests of reading by children in the Inner London Education Authority are also said to show serious evidence of a progressive decline by West Indian children. (61) Indigenous children obtained scores of 98.1 and 97.8 at age 8 and 15 respectively, while the comparable scores for West Indian children were 88.1 and 85.9, a decline of 2.2 points and an increase in the gap between indigenous and West Indian children of 1.9 points (Indian and Pakistani children showed gains). Here the decline is real, if small, but once again it is increased by a statistical analysis which is said to show that the increase in the gap at age 15 is really as large as 3.7 points. (62)

Comparison with the ILEA data can be obtained by analysing the scores on reading tests given at age 11 and 16 in the National Child Development Study. Here the gap between West Indian and indigenous children is 14.3 points at age 11 and 14.1 points at age 16. The size of the gap is somewhat larger than in the ILEA data, presumably because there the indigenous children scored about 2 points below the national average. But there is no suggestion that West Indian children are falling further behind their white contemporaries. This may, of course be partly because some of the decline occurs before the age of 11 (this is certainly true in the ILEA survey). But it is also possible that the statistical analysis of the ILEA data has exaggerated the extent of the decline. If that is true, then the overall picture, from the relatively small number of studies available, is one of, at most, relatively trivial declines in West Indian performance over time.

There is little justification, particularly in the case of IQ scores, for the claim that West Indian children fall seriously further behind their white contemporaries as they progress through school.

It was difficult, if not impossible, to find convincing evidence that the school achievements, let alone IQ scores, of West Indian children in this country are adversely affected by such psychological factors as their attitudes to school, society and themselves, or directly by society's or their teachers' attitudes to them. The fact that West Indian children do not appear to fall seriously further behind white children in school as they grow older, although surely welcome in itself, tends to undermine one final possible reason for accepting this explanation of their relatively poor performance. It does not, of course, prove that such an explanation is wrong. The idea that a child's achievements at school, and perhaps even his IQ, could be affected by psychological factors is by no means unreasonable. It is surely more unreasonable to suppose that the only environmental factors to affect IQ should be economic and demographic - just because there are the variables most readily measured in large-scale social surveys. But the fact remains that the evidence for other, more psychological factors is almost totally lacking.

IQ, social circumstances and school performance

The question which we originally set out to answer was whether differences in school performance between indigenous children and those from any ethnic minority should be attributed in whole or in part to differences in their natural intelligence. Although children from some ethnic minorities certainly obtain, on average, lower scores on standard IQ tests than do indigenous children, we have found rather little reason to believe that this reflects genetic differences for intelligence. On the contrary, such differences in IQ scores as we have found are clearly related to the same sort of differences in social circumstances that are associated with differences in IQ among the indigenous majority. If we could show that children's IQ scores were related to school performance in much the same way whether they are white, West Indian or Asian, the implication would be that any differences between groups in school performance were also related to differences in their social circumstances.

IQ and school performance

How well do IQ scores predict the school performance of children from ethnic minorities? The question is, in part, one concerned with potential bias in IQ tests. If it turned out that IQ scores seriously underestimated the performance of which, say, West Indian children proved capable in school, in the sense that a West Indian child with an IQ of 90 did very much better in school than a white child with the same IQ score, this would imply that IQ tests were biased against West Indians. In the US, Jensen has claimed, there is no evidence that IQ tests are biased against blacks in this way. What is the position in this country?

There are several reports on the performance of West Indian and Asian children on standard tests of reading and mathematical ability. Their general findings are broadly comparable to those that have looked at IQ scores: both West Indian and Asian children, but particularly the former, tend to obtain lower scores than do indigenous children. (63) This suggests that average differences between groups in IQ scores are related to the average differences found when children are directly tested for what they have learnt in school. But it is also important to know whether, within each group, children who obtain higher IQ scores also do better on these other tests. If this were not true, we should be inclined to doubt that IQ had the same meaning for each group. The National Child Development and Child Health and Education Studies gave both IQ and reading and mathematics tests to children, and we have calculated, for each ethnic group, the correlation between IQ scores and scores on these other tests. (64) The results are shown in Part A of Table 7. As can be seen, the correlations are all high, and there is no serious suggestion that their size differs systematically between groups (statistically, none of the differences is significant). By this measure, then, IQ scores do mean much the same in each group.

But there is a further question which we can ask, for these correlations tell us only that, for example, a West Indian child with an IQ of 105 is likely to obtain higher scores on reading and mathematics tests than one with an IQ of 85. They do not say whether he will obtain the same scores as a white child with an IQ of 105. It is possible that West Indian and white children with similar IQ scores obtain systematically different scores on reading and mathematics tests. In order to answer this question, it is necessary to perform another regression analysis which, in effect, compares the reading and mathematics scores actually obtained by West Indian children with the scores they would have obtained, given their IQ, if the relationship between reading or mathematics and IQ scores were exactly the same in West Indian children as it is in white children. The results of these regression analyses are shown in Part B of Table 7. It is clear that the actual scores obtained by both West Indian and Asian children, in both studies, are not only lower than the average white scores (approximately 100 on both tests in both studies), their scores are also, in virtually all cases, lower than the scores one would have predicted on the basis of their IQ. In other words, IQ scores tend, if anything, to overestimate the performance of both West Indian and Asian children on these standard tests of school achievement.

Many of these discrepancies are extremely small and statistically not significant. But some are too large to be ignored and require some explanation. The discrepancies between obtained and predicted scores for the Asian children in NCDS, although quite sizeable in the case of reading, are largely due to the inclusion of children who had been in the country for less than 4 years. The scores obtained by Asian children who entered the UK before the age of 7 are as good as those of indigenous children, on both reading and mathematics tests, and do not differ from the scores one would have expected on the basis of their IQ. In CHES, there is essentially no discrepancy between predicted and obtained scores for Indian children, while Pakistani children, although doing worse on reading tests, do slightly better on mathematics tests, than their IQ scores would have led one to expect. Possibly more serious discrepancies are to be found in the West Indian children in CHES, for they perform at a significantly lower level on both reading and mathematics tests, particularly the latter, than their IQ scores would predict. Why this should be so is not clear. Perhaps their IQ scores are misleadingly high. But an alternative possibility is suggested by the arguments we considered in the previous section. If West Indian children are not doing as well at reading and mathematics as one might have expected, this might be because of the way they are taught or the preconceptions teachers have about their abilities. There is some evidence, as we have seen, that teachers' expectations can have a greater impact on measures of school achievement than on IQ scores. There is, of course, no evidence to show that this is what has actually been happening to the children in question. Moreover, other studies nave not reported such discrepancies between IQ scores and measures of school achievement in West Indian children. In the NCDS data, shown in Table 7, the discrepancies are quite small. And in another study, which reported IQ and reading scores for indigenous, West Indian and Asian children, at age 12 the West Indian children were reading as well as the Asian children in spite of obtaining significantly lower IQ scores. (65) In this case, therefore, it was the Asian children who read less well than their IQ would have led one to expect.

Perhaps of even more interest than the scores obtained on reading and mathematics tests are the results of public examinations such as CSE and O Levels taken at the age of 16+. How well do West Indian and Asian children do on such exams and how well is their performance predicted by the scores they obtain on IQ tests? Published evidence suggests that West Indian children obtain significantly fewer 'good' passes (O Levels A-C and CSE 1) than indigenous children, but that Asian children are probably doing just as well as the indigenous majority. (66) There are, however, no published data on the relationship between IQ scores and examination performance and we have therefore analysed data from NCDS to provide some preliminary information. The results of our analyses are shown in Table 8. The picture is distinctly complex.

Since a sizeable minority of children, whether white, West Indian or Asian, obtained no graded passes in CSE or O Levels at all (either because they did not enter or because they failed), the data are analysed first to see whether there is any difference in IQ scores between children with at least one graded pass and those with none. As can be seen from Part A of the table, indigenous children with at least one pass have, unsurprisingly, a higher IQ score than those with none. But although the same is true for Asian children, in West Indian children there is virtually no difference between the two groups. In West Indians alone, for reasons which are not immediately apparent, IQ scores do not seem to predict which children will have some success in these exams and which will have none.

Part B of the table gives the correlation, in those children with at least one passing grade, between IQ scores and two measures of success: the total number of passes achieved at any age, and the number of good passes achieved by the age of 16. Both of these measures of examination performance show moderately high correlations with IQ in all ethnic groups. Although the correlations for West Indian children are slightly lower than those for the other children, the differences are not large. In all groups, children with higher IQ scores tend to do better than those with lower scores.

But, as with the reading and mathematics tests, we also need to know whether a West Indian or Asian child with a high IQ obtained as many passes as a white child with a similar IQ score. Once again, therefore, we need to perform a regression analysis which compares the number of passes obtained and the number predicted on the basis of IQ. This information is given in Part C of the table. It is clear that both West Indian and Asian children actually obtain rather more total CSE and O Level passes than would have been expected, given their IQ scores, if they had been white. In the case of Indian children this underestimate of examination performance is statistically significant. It should be noted, however, that both groups of children have rather fewer 'good' CSE or O Level passes by the age of 16 than their IQ scores would have predicted (although these differences are not significant).

The number of children involved in these analyses is relatively small and, yet again therefore, their results should be treated with some caution. For example, the data for the Asian children do not accord with more recent, published data which suggest that Asian children obtain as many good passes as white children, and thus considerably more than in this study. But in the case of West Indian children there are now some other fragments of evidence to suggest that they may be obtaining rather better examination results than is sometimes supposed. (67) There is no room for complacency, for there seems to be no question but that they obtain significantly fewer good examination passes than white or Asian children, but the picture is not one of unrelieved gloom.

What then can we say about the relationship between IQ scores and these measures of school achievement? There is no simple answer. While IQ scores tended to overestimate the performance of Asian, and perhaps more strikingly of West Indian, children on tests of reading and mathematics, there is equally good evidence that they can underestimate the performance of both groups, again more particularly of West Indians, on public examinations. Yet again, we can only speculate why this should be so. It is at least possible that both West Indian and Asian children and their parents place a higher value on examination credentials than their white contemporaries and are more willing to stay on at school or to attend a college of further education in order to obtain such credentials.

But we should not exaggerate the discrepancies between IQ and school achievement in ethnic minorities. The fact is that many of the relationships between the two are very similar in West Indian, Asian and indigenous children. The correlations between IQ and reading, mathematics and examination passes are reasonably high and of much the same value in all three groups. Although some studies have reported discrepancies between IQ scores and some measures of school achievement, they have not always been confirmed, and the discrepancy between IQ and a different index of school achievement may be in exactly the opposite direction. It would surely be unreasonable to argue, on the basis of this evidence, that IQ tests systematically underestimate the academic potential of West Indian or Asian children in this country. If the criterion of bias is that one test gives a lower estimate of performance than another, there is no more reason to suppose that IQ tests are invariably biased against West Indians in this country than against blacks in the US. More important, if high and low IQ scores have much the same implication for school performance in other ethnic groups as they do in whites, then one might expect to find that factors that affected West Indian or Asian IQ scores would also affect their school achievements.

Social circumstances and school performance

If IQ scores are related to school achievement in much the same way in West Indian, Asian and white children, and if the differences between the IQ scores of the three groups of children are related to differences in their social circumstances, it seems at least possible that differences in their school achievement will also be related to these differences in social circumstances.

A final set of regression analyses, similar to those whose results are shown in Tables 4 and 6, were conducted to examine this possibility. The measures of school achievement we used were the results of the reading and mathematics tests from NCDS (at both 11 and 16) and CHES (at age 10), and the examinations results from NCDS. Performance on all of these was correlated, in the indigenous population, with the same set of social circumstances that were related to variations in IQ scores. The results of the regression analyses are given in Table 9; not surprisingly, they show that much, although by no means all, of the initial difference between either West Indian or Asian and white children is accounted for by the differences in their social circumstances.

Several comments on these results are in order. The data from NCDS are for all available children regardless of how recently they had entered the country. This certainly explains the particularly poor performance of the Asian children at age 11. As can be seen, by age 16 they have made significant gains, particularly in the mathematics test; and if the analysis is confined to those children who entered the country before the age of 7, the residual differences at age 11 are trivially small, and by age 16 Asian children obtain higher scores than white children even before the regression analysis is carried out. As with IQ tests, the West Indian children do not show such a marked effect of length of residence in this country, and show relatively little improvement from 11 to 16. The data from CHES indicate that West Indian children obtain lower scores on reading and mathematics tests than do either Indian or Pakistani children and that the residual differences between their scores and those of the white children, i.e. the differences unaccounted for by differences in their social circumstances are considerably larger than in these other groups. The results shown in Table 7 should have prepared one for this finding, since there we saw that West Indian children, unlike Asians, obtained lower scores on these tests than their IQ scores would have led one to expect. Once again, we can offer no more than speculation why this should be so, and must stress that it is an isolated result to be treated with the same caution.

But the overall pattern of results is as we might have expected. Differences in their social circumstance can account for a significant part of the differences in school performance observed between ethnic minorities and white children. In the case of the examination data from NCDS, it can be seen that both West Indian and Asian children obtain more total passes than do white children in comparable social circumstances, just as they obtained more passes than white children with comparable IQ scores. And the difference in the number of good grades obtained is very small once social circumstances have been taken into account. Although it is important to remember that regression analyses such as these can never prove that differences in test scores are actually caused by differences in social circumstance, it is at least worth considering the possibility that elimination of some of the more glaring instances of social inequality might significantly affect differences in school achievement.

Summary and conclusions

At the risk of seeming unduly pedantic, it is necessary to reiterate that the evidence available to answer any of the important questions about the attainments and abilities of ethnic minorities is very imperfect. There are relatively few studies providing information about their performance on standard IQ tests; in most of these the data were collected 10 or more years ago; their results do not always agree with one another; and in virtually no case is there any information that would make it possible to disentangle possible causes of differences between different ethnic groups. It has been necessary from time to time to call upon American data to answer certain questions, but there is no guarantee that such data are relevant to the British case. The only safe conclusion is that few secure conclusions can be drawn - and that many of those most confidently asserted reflect preconception and prejudice rather than sober evaluation of the evidence. With that proviso, it is worth seeing in what direction the evidence seems to point.

1. Several studies undertaken in this country over the past 10 to 15 years have shown that children from ethnic minorities on average attain somewhat lower scores on standard IQ tests than indigenous children.

2. Much of this difference has been due to the particularly poor performance of children who have only recently immigrated into this country. This is particularly true in the case of the Asian community, where children of Indian or Pakistani origin born in this country have in some, although by no means all, studies both achieved IQ scores well within the range of the indigenous population and also performed equally well on school exams. Whether this is still true, or true of all groups of Asian children, may be open to question.

3. In the case of children of West Indian origin, this does not seem to be the whole story: while those born in this country have higher scores than recent immigrants, they still score, on average, about 5 to 12 points below the population mean. Despite claims to the contrary, however, there is little reason to believe that the size of this gap increases as children progress through school. Indeed, there is some evidence that West Indian children may do rather better at CSE and O Level examinations than the gap in IQ scores might lead one to expect.

4. Much of this difference in IQ scores between West Indian and indigenous children appears to be related to differences between them in such factors as parental occupation, income, size of family, degree of overcrowding, and neighbourhood. All of these factors are related to IQ among whites, and when they are taken into account, the difference between West Indian and indigenous children is sharply reduced, to somewhere between 1 and 7 points.

5. These findings tend to argue against those who would seek to provide a predominantly genetic explanation of ethnic differences in IQ, but they equally imply that such differences are not due to a special set of factors unique to the West Indian experience. Although discrimination against West Indian families in this country may have an important indirect effect on their children's IQ scores by ensuring that they live in impoverished circumstances, there is less reason to believe that such discrimination, whether by society as a whole or by teachers and IQ testers in particular, has any direct effect on the West Indian child's performance. There is, moreover, relatively little evidence that specifically supports either this or the genetic position. Such imperfect attempts as have been made to study the intellectual development of black and white children brought up in comparable surroundings have found few if any differences in their IQ scores. Conversely, there is not much reason to believe that teachers' expectations have any large effects on their pupils' IQ scores (although they may affect other aspects of their performance at school), and although motivational and attitudinal factors have sometimes been found to have significant effects on IQ scores, the effects are neither consistent nor large. At best such factors may make a modest contribution to observed ethnic differences in IQ scores; they are unlikely to be the most important cause.

6. The evidence is not compelling, then, but on balance it does seem to point one way rather than others: ethnic differences in IQ scores are probably largely caused by the same factors as are responsible for differences in IQ within the white population as a whole. And much the same conclusion probably applies to ethnic differences in more specific measures of school performance such as tests of reading or mathematics or public examinations. Here too, such differences as there are between different ethnic groups seem to be largely related to the same social factors that are related to differences within the indigenous populations. If, therefore, we wish to affect the IQ scores of children from ethnic minorities in our society, or indeed their school performance, we might make a start by improving the social and economic circumstances of their families.

Finally, it seems worth stressing that the possibility that such a programme might have beneficial effects on IQ scores is surely not the best reason for wishing to improve the conditions in which a substantial minority of the community is forced to live. By the same token, even if it turns out that racial prejudice is not a direct cause of the low IQ scores obtained by many West Indian children, this is no reason to countenance such prejudice. There are many things in life substantially more important than IQ. Even in a narrow educational context, no one should be particularly interested in IQ scores. Educationalists should be concerned rather with how well children do at school, how adequately they master certain basic skills and, if need be, with their examination results - since these results will affect their future chances in life in a multitude of ways. They should ask, for example, why West Indian children do not read as well as white children, and rather than wondering whether this is due to a difference in IQ, they would be better advised to tackle the problem directly - by relying on those factors (such as parental involvement (68)) actually shown to be capable of affecting a child's ability to read.

The only reason for being concerned with IQ scores is that they tend to predict educational attainment (and even that not particularly well). Attempts to invest a child's IQ score with greater importance usually rest on one of two assumptions: that it provides a direct measure of his true worth, or that it predicts his later success in life significantly more accurately than other measures of educational attainment. Neither assumption is justified. There is a third fact that, in some minds at least, has inflated the importance of IQ scores: the possibility that there might be significant and ineradicable differences in the average IQ of different social, ethnic or racial groups has been thought by some to justify prejudice or discrimination against all members of the groups with lower scores. But the justification is spurious. As has often been pointed out, an average difference between two groups, even one as large as the 15 point difference claimed to hold between blacks and whites in the US, conceals even larger differences within each group and therefore a very large degree of overlap between the two. Indeed, even with this 15 point difference, it has been argued, the chances are that the difference in IQ between a randomly selected black child in the US and a similarly chosen white child from the same social class will be only 1 point larger than the difference one would expect to find between a brother and sister brought up in the same family. (69) And the fact of the matter is that we do not discriminate against all groups with lower than average IQ scores. No one has ever suggested that we should discriminate against twins; but there is excellent evidence that the average IQ of twins is about 5 points below that of the rest of the population. (70) We do not think that this matters, and we should rightly question the good sense or good will of anyone who claimed that it did.

Table 1 Summary of four studies comparing IQ scores of indigenous children and children of West Indian origin

Table 2 Summary of four studies showing IQ scores of West Indian and indigenous children approximately matched for father's occupation or neighbourhood

Table 3 Percentage of indigenous and of West Indian families living in various social circumstances

Table 4 Results of multiple regression analyses of West Indian and indigenous children's IQ scores using variables listed in Table 3

Table 5 Summary of four studies comparing IQ scores of indigenous children and children of Indian or Pakistani origin

Table 6 Social circumstances of Indian and Pakistani families and results of regression analyses of children's IQ scores taking these factors into account

Table 7 Prediction of performance on tests of reading and mathematics by IQ scores

Table 8 Prediction of examination performance at age 16+ by IQ score at age 11 for indigenous, West Indian and Asian children

Table 9 Results of multiple regression analyses of West Indian, Asian and indigenous scores on reading and mathematics tests and public examinations using social variables listed in Table 3

Footnotes

1. JWB Douglas, The Home and the School, MacGibbon & Kee, 1964; C Jencks, Inequality, Basic Books, 1972; M Rutter & N Madge, Cycles of Disadvantage, Heinemann, 1976.

2. See, for example: Interim Report of Committee of Inquiry into the Education of Children from Ethnic Minority Groups, HMSO, 1981, pp. 6-8; MJ Taylor Caught Between, NFER-Nelson, 1981, pp. 111-122; S Scarr, BK Caparulo, B M Ferdman & R B Tower, Brit. J Develop. Psychol., 1983, 1 ,31-48.

3. AR Jensen, Harvard Educ. Rev., 1969, 39, 1-129.

4. HJ Eysenck, Race, Intelligence and Education, Temple Smith, 1971, p. 130.

5. There is not even complete agreement on what is to count as an IQ test, although the traditional distinction between them and tests of achievement (e.g., of reading or mathematics) is followed here. IQ tests are usually divided into individual or group tests, depending on whether they are administered to one person at a time, as are the Stanford-Binet or Wechsler tests, or to a large group at the same time. A more important distinction is between 'verbal' and 'non-verbal' tests, the former concentrating (not necessarily exclusively) on tests of vocabulary, meanings of phrases, verbal analogies and the like; while the latter may use numerical, pictorial or diagrammatic material, as in Raven's Matrices test which can be administered to a group with only minimal instructions. The Wechsler test is divided into two parts, verbal and performance, the former including tests of general knowledge and mathematics as well as vocabulary, and the latter comprising typical non-verbal pictorial and diagrammatic material.

6. See, for example: AR Jensen, Bias in Mental Testing, Methuen, 1980, pp. 277-281.

7. The proportion of the total variation in any characteristic found in a given population that can be attributed to genetic differences between members of that population is termed the 'heritability' of the characteristic.

Heritability can, in principle, be calculated for any characteristic that can be measured, whether it be height, weight, colour of hair or of skin, musical ability or IQ. Two points should be stressed. First, heritability is a statistic applicable only to a particular population at a particular time. Secondly, high heritability does not mean that the characteristic in question is genetically fixed and therefore unamenable to environmental influence. The heritability of height, for example, is generally taken to be very high, of the order of .90, in most modern populations. But changes in the diet of a particular population can still produce large increases in average height: that of the Japanese, for example, has increased by over 5cm since 1945.

Both Jensen and Eysenck have repeatedly argued that the heritability of IQ is about .80, i.e. that 80 per cent of the observed variation in IQ in modern Western populations is due to genetic variation within these populations. Relatively few other authorities would now agree with them. ND Henderson (Ann. Rev. Psychol., 1982,33, p. 413) states that 'an estimate of 50 per cent seems more in vogue' or 'between .4 and .7.' S Scarr and L Carter-Saltzman in Handbook of Human Intelligence, (Ed. R J Sternberg), Cambridge University Press, 1982, p. 792) write: 'Most investigators in behavior genetics conclude from the evidence that about half (+/-.1) of the current differences among individuals in US and European white populations in measured intelligence result from genetic differences among them.'

8. PE Vernon, Intelligence and Cultural Environment, Methuen, 1969, p.70.

9. RS MacArthur & WB Elley, Brit. J. Educ. Psychol., 1963,33,107-119; JD Nisbet & NJ Entwhistle, Brit. J. Educ. Psychol., 1967, 37 188-193.

10. JC Loehlin, G Lindzey & JN Spuhler, Race Differences in Intelligence, Freeman, 1975, pp.I77-195.

11. LA Messé, WD Crane, SR Messé & W Rice, J. Educe. Psychol., 1979, 71, 233-241; AR Jensen, Bias in Mental Testing, pp. 465-515; S Scarr & D Yee, Educ. Psychologist, 1980, 15,1-22.

12. It is clearly somewhat inappropriate to refer to UK citizens as West Indian children. In earlier studies, many children so characterised may have been born there, but this is not true of more recent studies. We use the term, somewhat hesitantly, and later the term Asian, Indian or Pakistani to refer to children whose families have recently migrated from the West Indies or Indian sub-continents. Where we report the results of original analyses of The National Child Development Study, the criterion for counting a child as West Indian or Asian was that he should be described as of the appropriate ethnic appearance, and that both his parents should have been born in the West Indies or in India, Pakistan or Bangladesh. This is, obviously, a rather conservative criterion. For the more recent Child Health and Education Study, where data on parental place of birth were not available, the criterion was simply that the child be described in the parental interview as of West Indian, Indian or Pakistani (including Bangladeshi) ethnic groups.

13. A M Shuey, The Testing of Negro Intelligence, Social Sciences Press, 1966; J C Loehlin, G Lindzey & J N Spuhler, Race Differences in Intelligence.

14. IQ tests are normally 'standardised' to give a mean or average score for the population as a whole of 100, with a 'standard deviation' of 15 points. That is to say, approximately two thirds of the population will obtain IQ scores between 85 and 115, or one standard deviation on either side of the mean. If, therefore, there is a 15 point difference between the mean IQ scores of blacks and whites in the US, this would imply that whereas about 50 per cent of whites obtained IQ scores of 100 or more, only about 15 to 20 per cent of blacks will do so.

15. W Yule, M Berger, M Rutter & B Yule, J. Child Psychol. Psychiat., 1975.16, 1-17.

16. J McFie & J A Thompson, Brit. J. Educ. Psychol., 1970,40,348-351.

17. The National Child Development Study is organised by the National Children's Bureau, 8 Wakley Street, London EC1. We are very grateful to the Bureau for making the data available and in particular to K Fogelman for all his help. The raw scores on the two IQ tests have been converted to IQ scores with a mean of 100 and standard deviation of 15.

18. The Child Health and Education Study is directed by Professor NR Butler, University of Bristol. We are very grateful to him and to his colleagues Dr Mary Haslam and Mr B Howlett for making the data available and for all their help. Our analyses of the data from this and the National Child Development Study were financially supported by a grant from the Economic and Social Research Council.

The British Ability Scales (CD Elliott, NFER-Nelson, 1979) consist of a set of 23 separate sub-tests or scales, only four of which were used in this study: word definitions, similarities, matrices and recall of digits. The first is a test of vocabulary, and the second a test of conceptual grouping or categorisation; by any criterion, these would be regarded as verbal IQ tests, The matrices sub-test is similar to Raven's Matrices, requiring the child to complete diagrammatic sequences, and is equally clearly a non-verbal IQ test. In the recall of digits sub-test or digit span, the child listens to a sequence of numbers and simply has to repeat them back to the tester; it is a test used in many other IQ tests (in the Wechsler tests and some versions of the Stanford-Binet test), where it is usually classified as a verbal test (as it is in the British Ability Scales). The basis for such classification is that factor analysis reveals that the digits test tends to agree with other verbal tests more than with non-verbal tests, although it is acknowledged that it usually correlates less strongly with other verbal tests than they do with one another. We have calculated the inter-correlations between these four sub-tests in the unusually large sample provided by this study. The correlation between definitions and similarities was twice as large as that between either of these sub-tests and recall of digits, and these were, in turn, no larger than the correlation between recall of digits and matrices. We also performed a factor analysis (varimax rotation) on these scores: the first factor was one on which definitions and similarities loaded more strongly than digits or matrices; the second was one on which the latter two tests loaded more strongly than the first two. By this criterion, the similarities and definitions sub-tests go together as a measure of verbal IQ, while the recall of digits and matrices tests go together as a measure of non-verbal IQ. Rather than insist on this grouping, however, we have chosen to present the results of the four sub-tests separately, allowing readers to make up their own minds.

19. N Foner, Internat. Migration Rev., 1979 13, 284-297.

20, B Tizard, O Cooperman, A Joseph & J Tizard, Child Development, 1972 43, 337-358.

21. K Eyferth, Archiv. Gesamte Psychol., 1961, 113, 222-241.

22. S Scarr & R A Weinberg, Amer. Psychologist, 1976, 31, 726-739.

23. See, for example: A R Jensen, Straight Talk about Mental Tests, Methuen, 1981, pp. 207-232. An excellent discussion of Jensen's data and arguments is provided by J R Flynn, Race, IQ and Jensen, Routledge, 1980. It will be obvious to the reader of Flynn's book that our own discussion owes a great deal to his painstaking and scrupulously honest account.

24. For a good discussion, see: M Rutter & N Madge, Cycles of Disadvantage, pp. 110-123.

25. MJ Taylor, Caught Between, pp. 27-35; M Ghodsian, J Essen & K Richardson, New Community, 1980, 8, 195-205; S Scarr et al. Brit. J. Develop. Psychol., 1983,1,31-48; AR Jensen, Educability and Group Differences, Methuen, 1973, pp. 168-169.

26. MJ Taylor, Caught Between, pp. 35-45; M Rutter, B Yule, J Morton & C Bagley, J. Child Psychol. Psychiat., 1975, 16, 105-124; S Scarr et al. Brit. J. Develop. Psychol., 1983, I, 32-48.

27. For example: SS Baratz & JC Baratz, Harvard Educ. Rev., 1970,40,34-36; Black People's Progressive Association and Redbridge Community Relations Council, Cause for Concern: West Indian Pupils in Redbridge, 1978.

28. AR Jensen, Straight Talk about Mental Tests, p.192.

29. AM Shuey, The Testing of Negro Intelligence.

30. For a critical analysis of the studies relied on by Shuey, see: P M Green, Dissent, 1976, 284-297.

31. PL Nichols & VE Anderson, Soc. Biol., 1973, 20, 367-374.

32. AM Grace, Unpublished MEd Thesis, University of Nottingham, 1972.

33. One relevant study in California (J R Mercer & W C Brown, In The Fallacy of IQ (Ed. C. Senna), The Third Press, 1973) found that a relatively small number of factors, father's occupation and education, type of neighbourhood, home ownership, mother's participation in various local organisations, and family structure, were able to account for most of the difference between the IQ scores of black and white children. It is important to note that to 'account for' such a difference in terms of a particular set of social factors does not mean that the social factors have been shown to cause the difference, but only that they were correlated with it and when allowance is made for them the IQ difference disappears. Whether this implies a direct causal relationship is another matter, discussed on p. 22.

34. VP Houghton, Race, 1968, 147-156; C Bagley, Soc. Econ. Studies, 1971 10.

35. It must be stressed that the information provided in Table 3 comes from two independent studies which asked rather different questions and collected and tabulated data in rather different ways. Even information seemingly as straightforward as the number of children, for example, is not comparable, for in NCDS this item refers to the number of children in the household (which might contain more than one family) while in CHES, it refers to the number of children in the family. In no sense, therefore, should the table be used to infer whether or not there has been any change in the social circumstances of West Indian (or indigenous) children over the 12 years separating the two studies.

A similar analysis of NCDS data has been published by C Bagley in Self-Concept, Achievement and Multicultural Education (ed. G Verma & C Bagley), Macmillan, 1982. One problem with this sort of analysis is that data on all the variables may not be availablefor all the children in the original sample. In the NCDS sample, for example, complete data for the variables listed in Table 3 are available for only 72 of the original 113 West Indian children of whom only 54 had been in the UK for more than 4 years. It is difficult to know whether they are a random sample from the initial set of 113, but the mean IQ score of 88.4 for these 72 does not differ greatly from the mean of 86.3 for all 113 children, while the mean of 89.7 for the 54 who had been in the UK for more than 4 years hardly differs at all from the mean of 89.0 for the 74 children in the entire sample who had been in this country for this length of time. In the CHES sample, complete data are available for 92 of the original West Indian children, with a mean IQ score of 91.5 compared to that of 92.4 for the entire sample. Although this difference in total IQ is trivially small, by an odd chance it is largely made up of a difference in scores on the recall of digits. The 92 children in the regression analysis had a mean score of 97.3 on this sub-test compared with a score of 99.2 for the entire sample.

36. MJ Taylor, Caught Between, pp. 27-35. Runnymede Trust, Britain's Black Population, Heinemann, 1980, pp. 55-72.

37. Interim Report of the Committee of Inquiry into the Education of Children from Ethnic Minority Groups, pp. 6-8; J Essen & M Ghodsian, New Community, 1979, 7, 422-429.

38. B Ashby, A Morrison & HJ Butcher, Res. Educ., 1970, 4, 73-80; R Sharma, unpublished PhD Thesis, University of London, 1971; other data on the performance of Asian children on IQ tests and in school are discussed in the report prepared for this committee by MJ Taylor.

39. As was true for the West Indian children in CHES, the large majority of Indian and Pakistani children in this study had lived in the UK for at least 5 years by the time of these tests - 159 of 170 Indian children and 79 of 91 Pakistani children. Exclusion of the small number of recent immigrants makes very little difference to the results shown in Table 5: they did obtain rather lower scores on the two verbal sub-tests, but actually obtained slightly higher scores on the matrices and recall of digits sub-tests.

40. S Scarr et al. Brit. J. Develop. Psychol., 1983,1,31-48.

41. M Ghodsian et aI., New Community, 1980,8, 195-205.

42. The data shown in Table 6 are from all Indian and Pakistani children regardless of their date of entry into the UK Exclusion of the small numbers who were not born here makes virtually no difference to the results of the analyses. Statistical analyses of the data on the social circumstances of West Indian, Indian and Pakistani children given in Tables 3 and 6 confirms, what is evident from the figures, that both Indian and Pakistani families are significantly less likely than West Indian families to have no male head, and that Indian children are significantly less likely than West Indians to receive free school meals or to live in inner cities or council estates. The only other significant differences are those where Pakistani families are actually worse off than West Indians.

43. There are, of course, numerous other minority groups now resident in the UK They are mostly too small to provide significant numbers of children in national surveys like NCDS and CHES. But some thorough local studies have occasionally turned up enough children to provide meaningful data. One such is W Yule et aI., J. Child Psychol. Psychiat., 1975, 16, 1-17. This found, for example, that Turkish Cypriot children obtained scores on IQ tests lower than those of West Indian children.

44. We have found one British study where West Indian children, selected to have similar scores to white children on verbal tasks, obtained reliably lower scores on Raven's Matrices (JF Payne, unpublished MA Thesis, University of Keele, 1967). But there were no data on any other ethnic group and it is impossible to know whether the results have any generality. American studies, on which of course Jensen's arguments were originally based, give little reason to believe so. Although some have reported particularly poor performance by American blacks on this test (C Higgins & C Sivers, J. Consult. Psychol., 1958, 22, 465-468; AR Jensen, Educability and Group Differences, pp. 309-312), others have found no such deficit (I J Semler & I Iscoe, J. Educ. Psychol., 1966, 57, 326-336; JM Mandler & NL Stein, J. Exp. Psychol., 1974, 102, 657-669). There is some evidence that blacks in the US find particular difficulty with the 'Blocks Design' sub-test of the Wechsler tests: this requires one to manipulate diagrammatic, spatial relationships possibly similar to those called for in the Matrices test. The evidence is discussed by JC Loehlin, G Lindzey and JN Spuhler, Race Differences in Intelligence, pp. 186-187; see also G R Reynolds & A R Jensen, J. Educ. Psychol., 1983, 75, 207-214. JM Mandler & NL Stein, Psychol, Bull., 1977, 84, 173-192, have published a thorough, critical review of this whole issue, entitled 'The myth of perceptual defect'; their conclusion is that there is no good reason to believe that blacks in the US perform especially poorly on non-verbal, perceptual or spatial tests.

45. We are assuming here that there probably is a difference between the IQ scores of West Indian and of Asian children, and that their relatively similar performance at age 10 in CHES should be taken as the exception rather than the rule. The evidence from various measures of school achievement, tests of reading and mathematics and public examinations, tends to support this. Provided thay have been living in this country fora reasonable length of time, Asian children do better at school than West Indian children (see below, pp. 41-48). Even in CHES, West Indian children obtain lower scores on tests of reading and mathematics than Indian children, and lower mathematics scores than Pakistani children.

46. Cited by: S Tomlinson, Educational Subnormality: A Study in Decision Making, Routledge & Kegan Paul, 1981, pp. 146-147. .

47. R Rosenthal & L Jacobson, Pygmalion in the Classrooom, Holt, Rinehart & Winston, 1968.

48. See: JD Elashoff & RE Snow, Pygmalion Reconsidered, Jones Publishing Company, 1971.

49. Elashoff & Snow, op. cit.; see also WB Seaver J. Pers. Soc. Psychol., 1973, 28, 333-342; WD Crane & PM Mellon, J. Educ. Psychol., 1978, 70, 39-49.

50. See: AR Jensen, Bias in Mental Testing, pp. 596-618; W Samuel, DJ Soto, M Parks, P Ngissah & B Jones, J. Educ. Psychol., 1976, 68, 273-285; W Samuel, J. Educ. Psychol., 1977, 69, 593-604.

51. C Bagley, M Bart & J Wong, In Race, Education and Identity (ed. GK Verma and C Bagley), Macmillan, 1979.

52. C Bagley, K Mallick & G K Verma. In Race, Education and Identity.

53. W Yule et aI., J. Child Psychol. Psychiat., 1975, 16, 1-17, found that West Indian girls performed better on IQ tests than West Indian boys, but neither the NCDS nor the CHES data show such differences: in NCDS the total IQ for boys is 86.7, for girls 85.5; in CHES, the comparable scores are 92.6 and 92.1. AM Grace (see footnote 32) also found no evidence that West Indian girls obtained higher scores than boys. One study which has reported higher self-esteem in West Indian boys than girls is: DM Louden, New Community, 1978, 6, 218-234.

54. P Lomax, Educ. Rev., 1977, 29,107-119.

55. C Bagley et al. In Race, Education and Identity.

56. D Milner, Children and Race, Penguin, 1975; A G Davey & P N