Cyril Burt
Abridged and revised by the author from 'The evidence for the concept of intelligence', Brit. J. Educ. Psychol., vol. 25 (1955), pp. 158-77.
I The Non-statistical Evidence
Current criticisms
The concept of'intelligence', and the attempt to measure intelligence by standardized tests, have of late become the subject of attack from several different quarters. The objections urged are partly practical and partly theoretical. Yet few of the critics show a clear or correct understanding of what the term really designates or of the reasons that led to its introduction. Two misconceptions are widely current. Writers who are chiefly interested in the more practical issues declare that intelligence 'is a popular and relatively unambiguous word', and denotes a quality that 'all can recognize, though few can define': as such it is 'too elusive to measure'. Hence, instead of pinning Z.Q.s on to the coat of each child, we ought, so they tell us, to leave all such decisions to the intuitive insight of the teacher. Writers who discuss the more technical aspects of the subject seem for the most part to suppose that the concept, if not the name, was invented by a small band of statistical enthusiasts - one critic mentions Spearman, Pearson, and myself - who deduced their theories by primitive factorial procedures that have since been 'publicly discredited': the more accurate methods of Thurstone and his American followers, they say, have since clearly shown that the intellectual achievements of different individuals are the product, not of a single general factor, but of a number of more specialized 'primary abilities'. And this at once accounts for the difficulty that besets all at- tempts to produce an agreed definition. As Captain Kettle ob- served, when asked why the pictures of the Saghalien sea-serpent showed such incredible differences: ' " Spects it's because there's no such crittur"; so each just draws his fancy.
The definition of intelligence
Most of the critics who protest about 'the spate of incongruous definitions' rest their complaint on the results of the famous Symposium organized some twenty years ago.' The editor of an American journal submitted two searching questions about the nature of intelligence to a dozen differing psychologists, and re-ceived a dozen different replies. But the varying descriptions sug-gested were not (as is commonly supposed by those who still quote them) intended to be'definitions' in the strict logical sense; they were, in the language of J. S. Mill, merely 'attempts to ex-plain the thing', not 'attempts to interpret the word'. As the editorial letter shows, the purpose of the discussion was primarily a practical one - to determine how intelligence appears to operate, with a view to ascertaining 'what material may most profitably be used in constructing tests'. But that is a separate question, and, except incidentally, will not concern us here. The questions I now want to settle are prior to all these: namely, (i) how precisely is the word to be defined, and (ii) what evidence is there for believing that something really exists corresponding to the definition pro-posed ? Instead of taking the term for granted and hunting round for a plausible formula, as is most frequently done, a sound scientific procedure requires us to start with the relevant facts: so let us take the second of our two questions first.
History of the concept
Many of the current criticisms spring largely from a manifest ignorance as to how the concept originated. A glance at the earlier literature is, therefore, necessary first of all.a As a brief historical review will show, long before the advent of statistical analysis several converging lines of evidence had already drawn attention to an important property of the mind, for which some special name seemed desirable.
------
1.'Symposium on Intelligence and its Measurement', J. Educ. Psychol,,
12, 1921, pp. 123-47 and 195-216.
2. A more detailed account will be found in my 'Historical Sketch',
which forms the first chapter of the Board of Education Report on Psycholo-
gical Tests of Educable Capacity (pp. 161).
------
(1) Observational. The earliest attempts to analyse and classify mental activities were based partly on the observation of various types of person in everyday life and partly on introspection. Plate, to whom we owe the basic distinctions, drew a clear contrast be-tween 'nature' and 'nurture', and then distinguished three'parts' or aspects of the soul (Republic 435A f.). The modern terms - in-tellectual, emotional, and moral - cognition, affection, and cona-tion - suggest rough but somewhat inexact equivalents for his untranslatable expressions. In a celebrated passage (PhaedruF, 253D) he sketches a picturesque analogy which conveys a better ;notion of the fundamental difference: the first component he compares to a charioteer who holds the reins, and the other two a pair of horses who draw the vehicle; the former guides, the latter supply the power; the former is the cyberneiicl or directive ele-ment, the latter the dynamic. 'Intelligence,' said Sully, 'steers like a rudder; emotion and interest supply the steam.'
Aristotle makes a further contribution. He contrasts the actual or concrete activity with the hypothetical capacity2 on which it depends, and thus introduces the idea of an 'ability'. Plate's threefold classification he reduces to a two-fold. For him the main distinction is between what he calls the 'dianoetic' (cognitive or intellectual)capacities of the mind and the 'orectic' (emotional and moral). Finally, Cicero, in an endeavour to supply a Latin terminology for Greek philosophy, translated Aristotle's term for 'capacity' by facultas, and 'orexis' by appelitio (or sometimes conatus) respectively, while to designate dianoia he coined a new word, rendering the Greek work almost literally by the com-pound 'intelligentia' (inter-legentia).
Here then we have the origin of both the concept and the name. So far from being a 'word of popular speech', whose meaning has been restricted and distorted by the modern psychologist, in-telligence is a highly technical expression invented to denote a highly technical abstraction. From Aristotle and Cicero it de-scended to the mediaeval schoolmen, and the scholastic theories in turn became elaborated into the cut-and-dried schemes of the faculty psychologists and their phrenological followers.
(2) Biological. As Guilford has remarked, the modern notion of 'intelligence as a unitary entity' was 'a gift to psychology from biology through the instrumentality of Herbert Spencer'. Follow-ing Aristotle and the later Scottish school, Spencer recognized two main aspects of mental life - the cognitive and the affective. All cognition (he explains) involves both an analytic or discrimi-native and a synthetic or integrative process; and its essential function is to enable the organism to adjust itself more effectively to a complex and ever-changing environment.
----
1. As those who are familiar with the recent development of'cybernetics' will be aware, the Greek noun (from which our own word 'governor' is derived) means 'a steersman, or director'.
2. De Animn, Il, 3, 414a, 31. Eth. Nic., I, 13, 18, 1102b, 30. The usual rendering 'power' must not be taken to imply causal agency: Aristotle is simply describing what Professor Broad has called a 'dispositional property'.
----
During the evolution of the animal kingdom, and during the growth of tbe: individual child, the fundamental capacity of cognition 'progres- sively differentiates into a hierarchy of more specialized abilities' - sensory, perceptual, associative, and relational, much as the trunk of a tree sprouts into boughs, branches, and twigs. To designate the basic characteristic he revived the term 'intelli-gence'.' Evidence favouring Spencer's somewhat speculative theories was adduced by Lloyd Morgan and other pioneers of comparative psychology; and his views on intelligence were ac-cepted, not only by British biologists like Darwin, but also by continental writers like Binet and Claparede, and more recently by Piaget.2 Mendel's earliest disciples maintained that his whole doctrine of unit-characters was utterly irreconcilable with the in-heritability of a graded trait, such as intelligence; but, as we shall see in a moment, the later developments of the Mendelian hypo-thesis not only permit it, but actually suggest it.
(3) Physiological. The clinical work of Hughlings Jackson, the experimental investigations of Sherrington, and the micro-scopical studies of the brain carried out by Campbell, Brodmann, and others, have done much to confirm Spencer's theory of a 'hierarchy of neural functions', with a basic type of activity de-veloping by fairly definite stages into higher and more specialized forms. In particular, the examination of the cortex, both in men-tal defectives and in normal persons, suggests that the quality of the nervous tissue in any given individual tends to be predomi-nantly the same throughout. Defectives, for instance, exhibit a 'general cerebral immaturity'; their nerve-cells tend to be 'visibly deficient in number, branching, and regularity of arrangement in every part of the cortex'. After all, as Sherrington himself ob-serves, much the same is true of almost every tissue of which the - human frame is composed - of a man's skin, bones, hair, of muscles; each is of the same general character all over the body, although minor local variations are usually discernible. In the adult human brain marked differences in the architecture of dif-ferent areas and of different cell-layers are perceptible under the microscope; and these specializations emerge and develop pro-gressively during the early months of infant life. And, of course, such differentiation is precisely what the Spencerian theory would entail.
----
1. H. Spencer, Principles of Psychology (1870).
2. J. Piaget, The Psychology of Intelligence (1950).
----
The earlier histologists, whose maps of the brain still figure in elementary textbooks of physiology as Shell and others have recently pointed out, 'greatly exaggerated the definiteness in the localization of different functions in the brain, and over- looked the enormous amount of difference between different human brains.'l
The experimental study of the brain leads to the same con-clusion. The intact brain always acts as a whole.2 No part of the brain functions in total isolation from the rest, as the older champions of cortical localization originally assumed. The activity, in Sherrington's phrase, is 'patterned, not indifferently diffuse'; but the patterning itself'involves and implies integra-tion'. The evidence of neurology, therefore, suggests something very like a theory of general ability, which gradually differentiates into more specific functions, though we must beware of picturing such functions as separate 'faculties' located in certain centres or compartments of the brain, after the fashion of the older phreno-logists.
(4) Individual psychology. All these earlier writers were interested primarily in the working of the mind as such, that is to say, in problems of general psychology. The first to apply scientific methods for the problems of individual psychology was Sir Francis Galton, Charles Darwin's half-cousin. Darwin and Spencer had maintained that the basic capacities of the human mind were hereditary, transmitted as part of our common racial endowment. Galton went farther and maintained that individual differences in these capacities were also innate. As a result of his investigations into 'hereditary genius', he was led to discard the traditional explanation in terms of faculties and types, and to substitute a classification in terms of'general ability' and'special aptitudes.' Of the two he considered general ability to be 'by far the most powerful'.'Numerous instances,' he says,'recorded in this book, show in how small a degree eminence is due to purely special faculties: people lay too much stress on specialties, thinking that because a man shines in some particular pursuit, he could not have succeeded in anything else. They might just as well say that because a youth has fallen in love with a brunette, he could not have fallen in love with a blonde. It is as probable as not that the whole affair was due to a general amorousness.
----
1. D. A. Shell, The Organiratian of the Cerebral Cortex (1956), p. 21ff.
2. K. S. Lashley, Brain Mechanisms and Intelligence (1929).
----
It is the same with mental pursuits ... Without a special gift for mathematics, a man cannot be a mathematician; but without a high degree of general ability he will never make a great mathema- tician.' Much the same view was maintained by Dr Johnson. Robertson, the historian, had argued that it was by virtue of very different gifts Newton had become a great scientist, Caesar a great commander, Shakespeare a great poet.'No,' replied John- son,'it is only that one man has more mind than another, though he may prefer this matter to that. Sir, the man who has vigour may walk to the North as well as to the South, to the East as well as to the West.''
The definition implied
These converging lines of inquiry, therefore, furnished strong presumptive evidence for a mental trait of fundamental im-portance defined by three verifiable attributes: first, it is a general quality; it enters into every form of mental activity; secondly, it is (in a broad sense of the word) an intellectual quality - that is, it characterizes the cognitive rather than the affective or conative aspects of conscious behaviour; thirdly, it is inherited or at least innate; differences in its strength or amount are due to differences in the individual's genetic constitution. We thus arrive at the concept of an innate, general, cognitive ability. We cannot, how-ever, keep repeating a cumbersome phrase of twelve syllables every time we wish to mention it. And, since a name that sug-gests its own meaning seems preferable to a brand-new esoteric symbol, what better label can be found than the traditional term 'intelligence'?
Here then is a clearly formulated hypothesis, the outcome of centuries of shrewd observation and plausible conjecture - a psychological hypothesis fully in accord with the findings of the biologist and neurologist. Nevertheless, each of the three proposi-tions that I have just laid down has been vigorously challenged; and at this point, therefore, the need for ad hoc inquiries based on more rigorous experimental and statistical techniques becomes plain.
----
1. F. Galton, Hereditary Genius (1869); Boswell, Life ofJohnson, Vol.
II. Carlyle defends much the same conclusion: 'I have no notion of a truly great man who couldn't be all sorts of men - Poet, Prophet, Priest, King, or what you will' (On Heroes, Lect. III). Galton's usual term was 'general ability'. In French, however, the corresponding word (habiMtd) has a different meaning. Hence Binet, in adopting Galton's view, substituted the term 'intelligence'; and, largely as a result of the wide popularity of his 'intelligence tests' in this country, 'intelligence' became the more usual name.
----
II The Statistical Evidence
(1) The general factor. At the beginning of the century the prob-lem which chiefly exercised students of individual psychology was, in Bain's phrase,'the classification of intellectual abilities or powers'. (i) Were there, as the faculty psychologists maintained, a number of specialized abilities, each independent of the rest - observation, practical ability, memory, language, reasoning, and the like? (ii) Or was there, as Ward maintained, 'not a congeries of faculties, but only a single subjective activity' - a general capacity for cognition as such:, (iii) Were there, as Galton' be-lieved, both a general ability and a number of more or less specialized capacities? (iv) Or, finally, might there be, as the earlier associationists and most of the later behaviourists alleged, no discernible structure in the mind at all?
Each hypothesis entailed its own distinctive corollaries; and Galton's technique of correlation offered a ready-made method of checking them. Thus, the obvious plan for attacking such a many-sided issue was to devise and apply experimental tests to measure the main forms of mental activity, and then calculate the correlations between each test and the rest. If, for example, the behaviourist view is right, and there is 'no organized structure in the mind, no ground for classifying mental performances under one or more broad headings, no basis for inferring efficiency in one type of activity from efficiency in another', then we should expect all the inter-correlations to be zero or at least non-significant.2
----
1. Many contemporary writers, particularly in the field of education, attribute the antithesis between 'general' and 'special' abilities to Spear-man. Spearman himself, however, frankly admitted his own theories were prompted by those of Galton and Spencer. However, in his earlier writings he definitely rejected the notion of'special aptitudes', as merely a relic of 'the discredited faculties of the older school'.
2. Thomson's sampling theory, though expressed in language similar to that of the 'anti-structural psychologists', leads to very different corollaries. The Mind,' he says,'has little structure: unlike the body, it is not sub- divided into distinct organs, but forms a comparatively undifferentiated complex of innumerable elements.' These he pictures as 'bonds', i.e. interconnecting neural paths: they have the same character or quality throughout the brain. But, so far from the effects of specific stimuli being limited to specific neural paths (as the earlier opponents of structure assumed), 'any sample whatever of these elements can be assembled in the activity called for by a "test"'.
----
If, on the other hand, the mind consists of a number of specialized faculties or abilities, such as 'observation' (assessed by tests of sensory capacity), and 'practical ability' (assessed by tests of motor capacity), and so on, then we should expect that all the inter-correlations between the sensory tests would be positive and similarly that all the inter-correlations between the motor tests would be positive; on the other hand, we should expect that ah the cross-correlations between the one group and the other would be approximately zero. Lastly, if there were no specific faculties at all, but only 'a single cognitive activity' -'attention', as Ward believed,'sensory discriminations' as Sully, and later Spearman, maintained - then we should expect the entire table of correla- tions to exhibit what Spearman called 'a perfect hierarchical order', or (in the more precise language of the mathematical textbook) to form 'a matrix of rank one' - apart, of course, from minor aberrations due to sampling errors.
The results of the earlier inquiries revealed, almost without exception, positive and significant correlations between every form of cognitive activity. This disproves hypotheses (i) and (iv). Further, except when the sample was small and the sampling errors large, there were nearly always well-marked clusters of'augmented corre- lations confined to similar forms of cognitive activity, and leaving significant residuals after the general factor was removed. This rules out hypothesis (ii). We are thus left with hypothesis (iii) as the only alternative consistent with the facts. And, accordingly, the unavoidable inference is that both a 'general factor' and a number of'group factors' must be at work.l
But we are not yet justified in identifying this abstract 'general factor' with anything so concrete as 'general intelligence'. In Spearman's investigations 'general intelligence' was always re- presented by an external criterion, i.e. either by teachers' assess- ments for intelligence as popularly understood, or (in later re- searches) by standard tests, selected as furnishing accredited 'reference values'. In my own investigations, the 'general cogni- tive factor' formed an internal criterion, namely, what I called the 'highest common factor' in the battery of tests. And to deter- mine the concrete nature of such a factor, or rather of the pro- cesses that give rise to it, a supplementary investigation was used, based on observations and introspections, and on the correlation of the factor measurements with independent gradings.2
----
1. C. Burt,'Experimental Tests of General Intelligence', Brit. J. Psychol.,
3, 1909, pp. 94-177. Since Spearman did not believe in the existence of multiple factors, he never used 'factor analysis' as the term is commonly understood. This method was in fact devised by Karl Pearson to analyse bodily measurements : see C. Burt, 'Alternative Methods of Factor Analysis', Brit. J. Statist. Psychol., 2, pp. 98-121.
2. Actually teachers' gradings for 'intelligence' (as I showed in my 1909 research) are markedly biased on favour of memory or capacity to learn; Spearman, following Sully and the sensationalist school, originally equated intelligence with 'sensory discrimination' as the basic form of mental anab,sis. Ward, Stout, and others inclined to identify it with 'attention' or'apperception', i.e. mental or'noetic' synthesis. This early disagreement about the 'nature of intelligence' is no reason for repudiating the concept: after all, there is little agreement about the 'nature' of gravity: but that is no reason for discarding the principle. And, in point of fact, the conflict can easily be reconciled if we borrow the suggestion of the neurologists and suppose its function to be that of'integration', i.e. organization, which involves both analysis and synthesis.
----
Later investigators, notably Brown, Thomson, and more recently Thurstone, have argued that, if we accept the existence of group factors or 'primary abilities', we can dispense with the hypothesis of a general factor by assuming that the group factors overlap. Thurstone in his earlier researches used the same simpli- fied formula which I had proposedl - the so-called 'centroid formula' - but rotated the factors thus obtained so as to secure a 'simple structure' of'primary abilities'. But, to get a satis- factory fit, it was nearly always necessary to allow these'primary abilities' (the 'first-order factors') to be correlated with each other. Later, however, he put forward a method for deriving un- correlated 'second-order factors' which should include a 'general factor' accounting for the correlations between the 'primary abilities'. Thus his final scheme was practically identical with that which my co-workers and I had produced. In their more recent writings both Brown and Thomson eventually acknowledged that 'the evidence for a general factor now seems conclusive'. Thom- son indeed constructed numerous booklets for testing intelligence, notably the'Moray House Tests' which have been so widely used in the eleven-plus examinations.2
----
1. C. Burt, L.C.C. Report on the Distribution and Relations of Education Abiliries (1917), p. 53ff. Much of the work of Thurstone and his followers was carried out with students and older persons, who show far more speciali-zation than school children; moreover the older groups available for testing are nearly always partly selected for intelligence, and consequently exhibit a narrower range of variation. Thurstone and most American investigators ignore all the concurrent biological and physiological evidence, and rely almost exclusively on statistical analysis.
2. L. L. Thurstone, Multiple Factor Analysis, 1947, p. 421ff; W. Brown and W. Stephenson,'A Test of the Theory of Two Factors', Brit. J. Psychol., 23, 1933, p. 352ff, and G. Thomson, Factorial Analysis of Human Ability, 1948.
----
Granted the existence of a general factor, what is its relative importance? In nearly every study of cognitive abilities among children of school age, it appears that the general factor accounts for about 50 per cent of the individual variability - rather more in the case of the younger age-groups, rather less as their age in-creases and special abilities begin to mature. Hence, for pur- poses of prediction - forecasting what this or that particular child is likely to achieve in school or in after-life - the general factor is by far the most useful, though by no means the only, guide.
(2) The factor as cognitive. Merely to demonstrate the presence of a general factorcommon to all cognitive activities does not of itself suffice to prove that this factor is specifically cognitive. Impressed by this obvious ambiguity, several writers went on to argue that in all probability the factor common to all mental and scholastic activities was not cognitive but conative. Such an interpretation had a warm appeal for those who cherished the doctrine ofintel-lectual equality. When a pupil lagged behindhand in nearly every subject, many teachers laid the blame on what Dr Ballard dubbed the 'general factor of laziness'. Conversely, when a bright child forged ahead in all he undertook, he found himself applauded as a paragon of industry and held up to his fellows as a model of zeal; 'genius', said the apostles of the gospel of work, 'is just an infinite capacity for taking pains'. It was partly as a result of this alternative interpretation that Spearman dropped his earlier identifications (' sensory discrimination' in his first paper,'neural plasticity' in the second) and proposed instead a hypothesis of 'mental energy':
Accordingly, in our later experiments, my co-workers and I correlated assessments for intellectual performances with assess- ments for physical, temperamental, and moral qualities. This time most of the cross-correlations were still positive, but ex- tremely low: it seemed as if there was a small but far more com- prehensive general factor - a super-factor, as it were - making for excellence in every direction, while the older and more con- spicuous factor for cognitive efficiency now appeared simply as a broad group factor, confined to cognitive activities alone: in short, the so-called 'general cognitive factor' turned out to be merely one of the largest of a number of'group factors' varying in extent and size. At the same time, another broad group factor emerged underlying the temperamental and moral assessments: this was obviously identifiable with what we had previously called 'the general factor for emotionality'. No sharp division, however, was found which would separate affective (or emotional) characteristics from conative (or moral). And the so-called cognitive factor was found to be quite as prominent in tests of prac-tical efficiency as in tests of intellectual activity in the narrower sense. In the light of this further evidence, there was no need to surrender the idea of a cognitive factor. But it certainly appeared necessary to revise the implications conveyed by the word cogni-tion. The basic contrast seems to lie, not so much between cog- nitive processes and non-cognitive (i.e. affective or conative) in the old introspective sense of those terms, but rather between the capacity for adapting, guiding, or directing mental activities by means of discriminative and integrative processes, and the capa- city for responding promptly, actively, and energetically.
(3) The factor as innate. The evidence we have so far reviewed seems fully to vindicate the notion of a 'general cognitive factor'; and during recent years the most frequent target for attack has been, not so much the existence of such a factor, but rather the assumption that differences in this factor are to a large extent hereditary or at least inborn. The first to collect factual evidence in support of such a conclusion was, once again, Sir Francis Galton. In his earliest inquiries he relied mainly on an analysis of pedigrees.l His aim was to show that, if a child had a genius among his various relatives, the likelihood that he himself would be a genius was enormously increased. He collected family his-tories for nearly a thousand 'eminent men' in various walks of life. An 'eminent' person was defined as the ablest in 4,000: from the definition, therefore, it follows that the chance that a boy picked at random would achieve 'eminence' was 1 in 4,000. Galton found that, if the boy was the son of a genius, his chance would be increased to about 1 in 4; if the grandson, it would be 1 in 29; if the nephew, 1 in 40; if a first cousin, 1 in 100. He readily recognized that a gifted father would be likely to provide his son with a better education and perhaps use his influence in the boy's favour. This objection he met in various ways: first, by trying to find 'two classes of men with equal advantages, in one of which they have high hereditary gifts, while in the other they have not'; secondly, by noting cases in which the eminent relatives did not in fact influence the child; and thirdly, by noting cases in which the unmistakable genius appeared by a kind of'spontaneous varia-tion'.
----
1. F. Galton, Hereditary Genius, 1869.
----
Sir Isaac Newton's father, for example, was a simple yeoman-farmer who died before the boy was born; his mother took him from school at the age of 14 to help run the farm; instead, he devoted himself to mathematics and scientific experiments, showing (according to numerous anecdotes) an amazing precocity. He was accepted at Trinity College, Cambridge, as a subsizar, which meant that he paid his way by menial work. He became Lucasian Professor at the early age of 27. As Galton records, later investigations have discovered two exceptionally able scientists among his mother's relatives, one of whom later became secretary to the Royal Society; but the connection seems to have been unrecognized at the time. And Galton goes on to emphasize that 'abundant instances of this emergence from obscurity' are to be found in the pages of his book. D'Alembert was a foundling, afterwards shown to be well-bred in respect of ability, put out to nurse as a pauper baby; largely self-educated, he 'attained first rank as a celebrity by the time he was 24'. Galton's list of'in- born geniuses who owed little or nothing to their social circum- stances' could easily be extended. Gauss, still acclaimed as the greatest mathematician of all time, was the son of a bricklayer; Laplace, almost as brilliant, the son of a farm labourer; James Watt, Whewell, Opie, Lincoln, and Carlyle were sons of car- penters; Marlowe, Winckelmann, and James Mill, sons of cob- blers; Thomas Cromwell, David Cox, and Michael Faraday, sons of blacksmiths; Defoe and Wolsey, sons of butchers;
Luther, Zwingli, and John Knox, sons of peasants; Kant's father was a strap-maker, Tieck's a rope-maker, Bunyan's a tinker, Franklin's a soap-boiler, Kepler's a drunken inn-keeper; and so one might go on. Still more significant is the amazing precocity of most of those I have named. Many eventually entered a university; but it was generally by their intelligence and hard work that they gained their education, not their education that produced their high intelligence.l Galton fully recognized that these family histories could fur-nish no more than strong prima facie evidence; and for this reason he suggested various new and fruitful methods of investigation - the calculation of correlations, the application of standardized tests, the study of foster-children, and of identical and non-identical twins - many of which he later took up himself.
----
1. For my data I have relied chiefly on Havelock Ellis, Terman and Cox, as well as the Dictionary of National Biography, and various foreign biographies. The descriptions do not always agree, usually perhaps because the fathers changed their occupations. Many of the fathers, e.g. those of Watt and Gauss, became more or less well-to-do; but they commonly wanted their sons to follow in their footsteps, and 'learn a trade'. Indeed most of the sons might well have adopted Jacob Bernoulli's motto:'Invite patre sidera verso.'
----
As he pointed out, three distinct questions are really involved: (a) what evidence is there for the fact of inheritance, (b) what precisely is the mode in which intelligence is inherited, and © what is the relative importance of the genetic factor as compared with the environmental ?
(a) The fact. In controversies about the fact of mental inheri-tance most critics have tended to assume that the two causal agencies commonly invoked - heredity and environment - are not merely antithetical, but mutually exclusive. The environ-mentalists suppose that, once they have shown that intelligence tests are affected by environment, it follows that all differences in intelligence are due to nothing but environment. Similarly the thoroughgoing hereditarians are apt to talk as though they be-lieved that differences in intelligence were due to nothing but genetic constitution. This is the familiar fallacy which I am tempted to label 'nothing-buttery'. With a few rare exceptions, like eye colour or serological differences in the blood, every ob-servable characteristic that geneticists have so far studied has proved to be the product of the joint action of both heredity and environment. There are no such things as hereditary characters; there are only hereditary tendencies.
Now, where two inexplicable factors, such as heredity and en-vironment, are likely to be involved, the obvious procedure will be to keep first one and then the other as constant as possible, and observe the results in either case.
1. Uniform environment. As psychological consultant to the London County Council, I had free access to its orphanages and other residential institutions, and to the private files of case re-cords giving the history of each inmate. My co-workers and I were thus able to study large numbers of children who had been transferred thither during the earliest weeks of infancy and had thus been brought up in an environment that was much the same for all. We found that individual differences in intelligence, so far from being diminished, varied over an unusually wide range. In the majority of instances, they seemed to be correlated with dif-ferences in the intelligence of one or both of the parents. Some of the most striking cases were those of illegitimate children of high ability: often the father (as the case records showed) had been a casual acquaintance, of a social and intellectual status well above that of the mother, and had taken no further interest in the child. In instances like these it is out of the question to attri-bute the high intelligence of the child to the special cultural opportunities furnished by the home environment, since his only home has been the institution.
2. Uniform heredity To secure cases in which the children's genetic endowment is the same, we may turn to the assessments obtained for monozygotic or 'identical' twins. Not infrequently the mother is unable or unwilling to bring up two children at the same time, and one is consequently sent to a relative or to a foster home. Owing to the popular prejudice against separating twins, she not unnaturally tries to keep these arrangements secret. But patient and tactful inquiries show that cases of twins brought up in different environments almost from birth are in fact much commoner than is usually believed. We have now collected as many as 53 such cases.' I reproduce the more important correla-tions in Table I and for comparison have added corresponding coefficients obtained from other pairs, both related and unre-lated. As regards intelligence the outstanding feature is the high correlation between the final assessments for the identical twins, even when reared apart: it is almost as high as the correla-tion between two successive testings for the same individuals. On the other hand, with school attainments the correlations are much lower for twins reared separately than for twins reared to-gether in the same home.
Table I
Correlations for Mental and Educational Characteristics
| A. English children (Burt) | ||||||
| Identical twins reared together | Identical twins reared apart | Non-identical twins reared together | Siblings reared together | Siblings reared apart | ||
| Number of pairs | 95 | 53 | 127 | 124 | 151 | |
| Intelligence | ||||||
| Group test | 0.94 | 0.77 | 0.55 | 0.56 | 0.41 | |
| Individual test | 0.92 | 0.86 | 0.53 | 0.50 | 0.42 | |
| Final assessment | 0.93 | 0.87 | 0.54 | 0.53 | 0.44 | |
| Educational | ||||||
| Reading | 0.95 | 0.60 | 0.92 | 0.84 | 0.49 | |
| Arithmetic | 0.86 | 0.71 | 0.75 | 0.75 | 0.56 | |
| General attainments | 0.89 | 0.62 | 0.83 | 0.80 | 0.53 | |
| B. American adults (Newman) | ||||||
| Unrelated children reaerd together | Identical twins reared together | Identical twins reared apart | Non-identical twins reared together | |||
| Number of pairs | 130 | 50 | 19 | 51 | ||
| Intelligence | ||||||
| Group test | 0.28 | 0.792 | 0.773 | 0.62 | ||
| Individual test | 0.729 | 0.88 | 0.77 | 0.763 | ||
| Final assessment | 0.727 | |||||
| Educational | ||||||
| Reading | 0.755 | |||||
| Arithmetic | 0.748 | |||||
| General attainments | 0.754 | 0.789 | 0.758 | 0.770 | ||
Several of our critics - Maddox and Liam Hudson, for example - have cited the American study of twins, reported by Newman and his collaborators, as proving that intelligence is dependent on environment. Thus, to take an oft-quoted pair, 'Helen', who had been trained as a teacher, scored with the Stanford-Binet tests an I.Q. of 116; whereas her twin sister, 'Gladys', brought up for much of her childhood in an isolated district of the Canadian Rockies, scored only 92. But, says Newman, her score
was higher than we might expect considering her scant education; and ... it seems certain that the great deficiency in education had inhibited the development of the rather high grade of mental ability with which she was endowed by heredity."
Thus Newman's interpretation in no way conflicts with ours, as will be obvious on comparing his figures with ours (see last 3 columns of Table I).
----
1. C. Burt, 'The Inheritance of Ability', Amer. Psychologist, 13, 1958,
p. 3ff and refs.
2. H. H. Newman, Twins and Supertwins, 1942, p. 136ff. Newman's
tests, it should be noted, were mainly verbal.
----
It is sometimes alleged that, since twins are born at the same time, the intrauterine environment must have been the same for both before birth, even if later on their environments differ widely, and that it is the former that is crucial. As it happens, however, this gratuitous assumption reverses the actual facts. Embryological and obstetric records show that, particularly with twins developed from split ova, the position of each in the uterus, and the subsequent development, is liable to differ widely. As a result twins usually display poorer health and energy than normal children; and often one twin suffers more than the other.
I think, therefore, it may be safely said that, apart from the influence of some preconceived theory, few psychologists nowa-days would be inclined to deny the mere fact of mental inheri-tance: the most that can be plausibly alleged is that its influence is comparatively slight and distinctly elusive.
(b) The mode of inheritance. The majority of those who still question the importance of mental inheritance, and many of those who support it, seem to cherish rather antiquated notions of the way in which inheritable characteristics are transmitted. If, as is generally believed, mental capacities are dependent on the physi-cal characteristics of the brain (or, to speak a little more precisely, on the structural and biochemical qualities of the nervous system), then we should expect those capacities to be inherited in accord-ance with the same principles that govern the inheritance of other physical characteristics; and these principles (except for obscure and apparently exceptional instances of extranuclear heredity) are essentially those commonly associated with the name of Mendel. Many British psychologists, however, feel a strong and not unreasonable prejudice against applying 'atomistic theories like Mendel's' to explain the facts of mental life, and conse-quently, so far as they admit the possibility of mental inheritance at all, still cling to the old Darwinian principle of blended inheri-tance. On this view heredity means 'the tendency of like to beget like' (the definition quoted by one of them from the Oxford English Dictionary). As a result, they commonly assume that the arguments for inheritance must consist in demonstrating re-semblances between the parent and his children by means of corre-lations. When the two parents differ, then the child is expected to consist in an intermediate blend of both, much as Aristotle main-tained that the 'offspring of a leopard and a camel would be a "cameleopard" or giraffe'.
The approach of the modern geneticist is the reverse of all this. As he views it, the real problem is rather to explain why in so many instances 'like begets unlike'. Both for the environmentalist and for the believer in blended inheritance, one of the most puz-zling phenomena is the appearance, not only of extremely dull children in the families of the well-to-do professional classes, but also of extremely bright youngsters in families where both the cultural and the economic conditions of the parents would, one might imagine, doom every child to hopeless failure. With the Mendelian hypothesis these anomalies are just what we should anticipate. However, the few critics who are familiar with the Mendelian explanation seem, as a rule, to suppose that it can apply only to discontinuous variations, and point out that intel-ligence, like stature, exhibits not discontinuous but continuous or graded variation. Hence, so they contend (sometimes citing the experiments of De Vries on 'pure lines'), the apparent differences in intelligence between one individual and another I17ust be due almost entirely to differences in environmental conditions.
Mendel himself was the first to indicate how his theory could be extended to account for the problem of graded characteristics. When supplementing his experiments on the hybridization of peas by hybridizing beans, and (as before) crossing white-flowered plants with purple, he found that, whereas with peas the two types sorted out with no hint of any intermediate colour, with beans the offspring displayed 'a whole range of hues from white to deep purple'. This, he suggested, might be explained by postulating that with beans the colour was determined, not by a single pair of alternative factors, but by a number of such pairs, each positive factor, when present, contributing a small additional amount of colour. And if, as before, the recombinations are the effects of chance unions, then the resulting frequencies would inevitably approximate to those of the normal curve.
However, in our early surveys of London children, we found that, when complete age groups were tested, the distribution of intelligence departed significantly from that of a perfect normal curve: there was a swollen tail at the lower end, due to an excess of mental defectives, and a smaller enlargement at the upper end. This and other considerations led me to put forward the tentative hypothesis that innate variations in intelligence are due partly to unifactorial and partly to multifactorial inheritance: i.e. they result from Mendelian factors of two main kinds (no doubt over-lapping), (a) major genes responsible for comparatively large deviations, usually of an abnormal type, and (b) multiple genes whose effects are small, similar, and cumulative.
Karl Pearsonl endeavoured to test the Mendelian theory in its multifactorial form by comparing its implications with actual figures obtained for height, arm length, and similar physical measurements, collected from over 2,000 students and their rela-tives. The expected correlations which he deduced for various degrees of kinship were in every case far smaller than those actually observed. He therefore rejected the hypothesis of Mende-lian inheritance, and fell back on the older theory of blending. However, in deriving his formulae and his expected values, Pearson relied on an oversimplified model. Contrary to what we now know to be the case, he assumed that the effect of assortative mating - the tendency of like to marry like - could be ignored as negligible, and that dominance would in every case be perfect. Sir Ronald Fisher" has since deduced more appropriate formulae, which allow for these and other complicating factors: when the necessary adjustments have thus been made. The theoretical values fit Pearson's own figures as closely as could be wished.
My colleagues and I have applied Fisher's methods (suitably modified) to assessments for intelligence." The data were secured in the course of surveys of the entire school population in a repre-sentative London borough, and covered nearly 1,000 pairs of siblings, together with the ratings for parents, and (so far as they were accessible) grandparents, uncles and aunts, and first cousins. The final assessments for the children were obtained by submitting the marks from the group tests to the judgement of the teachers who knew the children best; where the teacher disagreed, the child was interviewed personally, and subjected to further tests, often on several successive occasions. The assessments for the adults were naturally far less accurate. Nevertheless, the correla-tions computed from the actual data agreed with the theoretical values deduced from the multifactorial hypothesis far better than the values deduced from any other hypothesis hitherto put for-ward. The only appreciable discrepancy occurred in the case of first cousins. Here, as for stature, the observed correlation for intelligence was larger than the theoretical; but the difference could readily be explained if (as suggested above) variations in intelligence are affected by a few major genes as well as by numerous minor genes. The figures for cousins of maternal, paternal, and mixed kinship also show some slight evidence sug-gestive of sex linkage.
----
1. K. Pearson, 'On a Generalized Theory of Alternative Inheritance with Special Reference to Mendel's Laws', Phil. Trans., 203, 1904, pp. 53-87.
2. R. A. Fisher,'Correlation between Relatives on the Supposition of Mendelian Inheritance', Trnns. Roy. Sec. Edin., 52, 1918, pp. 399-433.
3. C. Burt and M. Howard,'The Multifactorial Theory of Inheritance and Its Application to Intelligence', Brit. J. Sratis. Psychol., 9, 1956, pp. 93-131.
----
© The relative influence of heredity and environmerrr. In practical work, however, the question most frequently raised is, not whether differences in intelligence are inherited, nor even how they are inherited, but rather what is the relative influence of heredity as compared with environment. To such a question there can be no single answer. We can only try to determine, for this or that type of environment, for this or that population, and for this or that type of assessment, how far the observable results ap-pear to be influenced by each of the two main groups of factors.
Formulae analogous to those used to deduce the expected cor-relations from the theoretical variances can also be devised for comparing the amount of the constituent variances from the ob-served correlations. I have ventured to modify Fisher's methods so as to allow for unreliability and for the systematic effects of environment, i.e. for those environmental influences which are correlated with those of heredity, as well as for random effects. The genetic contribution may be regarded as comprising two dis-tinguishable portions: that due to the 'fixable' component (or, as Fisher expresses it, to the 'essential genotypes') and that due to the'nonfixable' part (i.e. deviations resulting from dominance and similar influences). The data analysed consist of (i) marks ob-tained from the intelligence tests of the ordinary type taken just as they stand and (ii) adjusted assessments obtained by the supplementary methods already described.
From Table II it will be seen that, with the crude test-results taken as they stand, nearly 23% of the total variance appears due to nongenetic influences, i.e., to environment or to unreliability, and about 77% to genetic factors; with the adjusted assessments only about 12% (or slightly more) is apparently due to nongenetic influences and 88% to genetic factors. The improvement implies that the practice of relying on tests alone - usually a group test applied once only - is by no means the best method of estimating a child's innate ability. More accurate assessments can almost al-ways be secured by submitting the test scores to the teachers for criticism or correction, and where necessary adjusting them as de-scribed above.
Table II
Analysis of Variance for Assessments of Intelligence
| Source | Unadjusted test scores | Adjusted assessments |
| Genetic component: | ||
| fixable | 40.5 | 47.9 |
| nonfixable | 16.7 | 21.7 |
| Assortative mating | 19.9 | 17.9 |
| Environment : | ||
| systematic | 10.6 | 1.4 |
| random | 5.9 | 5.8 |
| Unreliability | 6.4 | 5.3 |
| Total | 100.0 | 100.0 |
Environment appears to influence the test results chiefly in three ways: (a) the cultural amenities of the home and the educational opportunities provided by the school can undoubtedly affect a child's performance in intelligence tests of the ordinary type, since so often they demand an acquired facility with abstract and verbal modes of expression; (b) quite apart from what the child may learn, the constant presence of an intellectual background may stimulate (or seem to stimulate) his latent powers by in-culcating a keener motivation, an interest in intellectual things, and a habit of accurate, speedy, and diligent work; © in a few rare cases illness or malnutrition during the prenatal or early postnatal stages may, almost from the start, permanently impair the development of the child's central nervous system. The ad-justed assessments may help to eliminate the irrelevant effects of the first two conditions; but it is doubtful whether they can adequately allow for the last.
Class Differences. Since differences in 'intelligence' as we have defined it result from differences in genetic constitution, they are to a large and measurable extent transmitted from father to son. Moreover, as historical records amply testify, there has been, throughout the ages but more particularly during the last cen- tury, a natural tendency for individuals of high intelligence to rise in the social scale, and for those of low intelligence to drift down- wards; and their children tend usually to inherit the high or low intelligence of their parents. The cumulative effect is a marked difference in the average intelligence of the different social classes. Thus the average I.Q. of children from the higher professional and administrative classes is 120; that of children from the ranks of unskilled labour only 92. But individual differences within each social class are far wider than the differences between the various classes; and since the numbers in the manual classes are vastly greater than the numbers in the non-manual classes, there are far more gifted children among the former than among the latter. Nevertheless a large proportion of the gifted children in the manual classes still fail to attain the type of advanced education that their high intelligence might seem to deserve. This, it would seem, is due partly to temperamental differences (which are them- selves largely innate), and partly to social background, parental attitudes, and above all differences in interest, aims, and ideals.' Moreover, the birthrate is not only much bigger among the so- called 'lower' classes, but also among the less intelligent and more unstable families in every social class; and this has not unnaturally aroused considerable alarm lest, as a general result, the average level of intelligence among the nation as a whole might not be slowly but steadily deteriorating from one generation to another.
III Summary and Conclusion
The various lines of evidence we have now reviewed lead to the following conclusions.
1. The evidence from biology, from neurology, from intro-spection or self-observation, and from the impartial observation of the behaviour of other persons, suggests that there is a general mental factor which enters, with varying degrees, into all types of cognitive process (where the term 'cognitive' is to be inter-preted as covering any and every kind of mental activity that guides or directs our actual behaviour as distinct from those pro-cesses that supply energy or motivation). This conclusion is fully confirmed by statistical data, particularly by the results of so-called factor-analysis.
2. individual differences in this general factor are largely de-pendent on the individual's genetic constitution, and appear to be transmitted in accordance with the Mendelian theory of unifactorial and multifactorial inheritance. This implies that differences between members of the same family, as well as resemblances, are genetically determined.
----
1. For the data on which the conclusions are based see C. Burt,'Ability
and Income', Brit. J. Educ. Psychol., 13, p. 83h, 'Intelligence and Social
Mobility', Brit. J. Statist. Psychol., 15, p. 3ff, and Intelligence and Fertility,
Cassell & Co., 1952, and the references there cited.
----
3. These empirical findings suggest the concept of an innate, general, cognitive factor, which for convenience has been technically named 'intelligence' - a term which unfortunately is all too requently misunderstood.
4. Differences in 'intelligence' as thus defined can be measured with a fair degree of accuracy by so-called 'intelligence tests', pro-vided they are appropriately constructed and selected, and care-fully checked.
5. The degree of intelligence with which any particular child is endowed is one of the most important factors determining his general efficiency all throughout life. In particular it sets an upper limit to what he can successfully perform, especially in the educational, vocational, and intellectual fields. Nevertheless, intelligence is by no means the only factor. His progress and achievement will also be affected by his special abilities and dis-abilities, by his physical health and stamina, and above all by his emotional and conative qualities (i.e., by what is loosely termed his'temperament', his 'character', or his 'personality'). At every stage heredity and environment, genetic constitution and post-natal influences are continually interacting. As a result the influence of the innate general factor is greatest during early childhood, but tends to be increasingly overlaid or masked during later years.
6. There is no discernible difference between the average intel-ligence of the two sexes; but there are well marked differences be-tween the average intelligence of the socio-economic classes - that of the manual classes being decidedly lower than that of the non-manual and professional classes. However, within each class in-dividuals vary enormously, ranging in every case from the born imbecile to the born genius.
All this, it will be seen, presents an extremely complicated set of problems for the educationist, the politician, the social reformer, and the policy-makers of every truly democratic state. Too often both the facts and the origins of the various problems are ignored or misconceived. Here therefore is an urgent need not only for further research but also for a more general understanding of the results of such research.
