Genetic Determinism Refuted:
Excerpts from Human Diversity by Richard Lewontin

Introduction by Tani Jantsang

Much hoolah has been made over identical twins.  Despite the so-called "research" on twins, exposed and explained by many, the subject keeps popping up. 

The old version was that twins were one Soul in two bodies.  Books such as "The Curious World of Twins" by Vincent and Margaret Gaddis is one of the more interestingly entertaining books on twins, filled with anecdotal "evidence" and even a push for astrology.  But now, and for a long time, the focus seems to be on the fact that they share identical genes.  No problem there:  but that focus is stretched out, and backed by PHONY PROOFS, to try to show that genetic determination reigns supreme. 

This idea might be fine for some, but the following text exposes the shoddy and dishonest resrearch for what it is. 

But why is this being presented on this website?  Because some have fallen for genetic determination hook line and sinker.  Notably, it is mostly those who imagine they were BORN "the alien elite" or somesuch - when they were IN FACT born from the gene pool that has been Culturally Christian for 2000 or so years.  What?  A Christian gene pool? 

O.K.  Let us for a moment get into the notion of a Christian gene pool.  Well, what that would mean is that whole groups of people that married and mated and bore children, did so with Christian Cultural norms guiding them every step of the way.  Certain phenotypes (visible traits, physical or behavioral) would be selected against, others selected for.  They couldn't exactly select based on genotypes because they couldn't SEE the genotype:  they could only SEE the expression of the genotype - the phenotype.  Given that this would express within a given "NORM OF REACTION" determined BY the environment:  well they lived in a Culturally Christian ENVIRONMENT.  This would extend even into what they ate and did not eat:  metabolism. 

Genetic determination is COMPLETELY against any notion of Free Will, Free Choice, Freedom to like or dislike, and so forth. 

Please turn now to an expert, Professor Richard Lewontin, who shocked the world and won a prize for PROVING that heterozygocity is far superior to homozygocity.  A chapter from his book HUMAN DIVERSITY  follows.

This is presented for educational purposes and falls under "fair use."

Excerpts from Human Diversity by Richard Lewontin:

p.24:

... But we do not even know which environmental factors are relevant to the development of most human traits, much less how to control them for an experiment. Certainly, early nutrition has some important effect on the height to which people grow, but it is remarkably difficult to get reliable information on people's actual food intake. For behavioral traits, the things that make up "personality," we hardly know what in the social environment we should begin to measure.

The second difficulty is more fundamental. In order to establish the norm of reaction for a genotype, we need to have many individual organisms of identical genotype, since exposing the same organism to different environments in sequence will not tell us what we want to know. We would need sets of identical quintuplets (or decuplets, or some such) whom we could separate at birth (or even before) and raise in controlled environments. In fruit flies, special controlled breeding methods are available that allow us to produce large numbers of genetically identical organisms. In Achillea, identical triplets were produced by the simple expedient of cutting each plant into three pieces. In humans, producing large numbers of genetically identical individual organisms and then growing them in controlled environments is biologically very difficult and socially out of the question. The closest that investigators are reported to have come to finding human beings of identical genotypes raised in different environments are the reported cases of identical twins who have been raised apart. Identical twins come from the same fertilized egg, and so are genetically identical. If such a pair could be raised in different environments, then we would have at least two points in the norm of reaction of their genotype. We do not, however, live in a Gilbert and Sullivan operetta world, where children are separated r from their parents at birth or exchanged in the cradle and raised in radically different environments. The only published studies of identical twins raised apart that claimed completely random adoptions were those of Cyril Burt, but it has now been revealed that these twins were fictitious. The other published study with a substantial number of twin pairs (43) shows clearly that twins separated at birth are not separated very widely but are kept by sisters, aunts,

p.25:

grandmothers, or close friends, usually living in the same region or even in the same village, and often attending school together. We simply do not have the data necessary to plot even two points on a norm-of-reaction curve. Thus, our assessment of the interaction between genes and environment in influencing the development of human beings necessarily depends on historical, anthropological, and social information and on information about the molecular and developmental bases of different traits.

Although no one has ever determined the norm of reaction for human blood groups in relation to, say, temperature or nutrition, we are certain that blood type is insensitive to these variables because of information we have on the r molecular mechanism underlying the formation of red blood cell antigens. This confidence is supported, although by no means conclusively, by the fact that ~ people have never been observed to change their blood type in the course of a r lifetime. In itself, that stability is not evidence of a constant norm of reaction. Finally, the relation of blood types of children to those of their parents is particularly simple. For example, two parents of blood type M have children only of type M, whereas two type N parents have type N children only. Children of type MN are the only outcome of a mating of an M parent with an N parent.

The phenotype of an individual organism is not completely specified even when its genotype and its developmental environment are given. There is a third contributing cause of variation. Consider the Drosophila whose left and right sides are shown in the photographs above. The number of sternopleural bristles,

p.72:

environment their average expressions would differ. The lack of heritability does not reveal the cause of that lack. Conversely, a high heritability of a trait in one population at one time tells us nothing about its heritability at other times in other populations.

A second important fact about heritability---one that is widely misunderstood---is that knowing the heritability of a trait cannot help one to decide how to change it. It is simply not true that, if a trait has a heritability of, say, 90%, it is useless to change the environment because somehow "genes determine the trait." Our discussion of genetic variance showed that it is not possible to extrapolate from the amount of genetic variance to the consequence of a change in the environment. All that the heritability of a trait tells us is how much genetic variation exists for that trait at a particular time in a particular population. A measure of heritability contains no implicit prescription for change.

These cautions about heritability are not academic quibbles. They lie at the heart of important social issues. A great deal of effort has been devoted to trying to determine "the" heritability of such quantitative human traits as IQ performance, wealth, schizophrenia, mental retardation, blood pressure, and so on, in the belief that a knowledge of the heritability of these traits will somehow prescribe social action. A recent U.S. court decision held that claimed cures for baldness are necessarily fraudulent because baldness is "inherited". But there are examples whose import is much more serious: It has been asserted that compensatory education will necessarily fail because IQ is said to have a high heritability. Similarly, it has been asserted that psychiatric treatment of schizophrenia either is useless or must necessarily depend upon drugs and other physical manipulations because schizophrenia is "heritable". In each case, the same fallacy is operating: "Heritable" is taken to mean "insensitive to environmental change." In addition, in the case of mental disease, the implication is that, because it is heritable, there must be a molecular defect and that, therefore, only a molecular manipulation can treat it. In part, the problem is that a word like "heritability" carries with it an everyday meaning that has been confounded witt its technical meaning. Counters have become money. But the problem is deeper. It goes back to the false dichotomy between nature and nurture, to the belief that gene and environment are separate and separable determinants of organisms rather than interacting and inseparable shapers of development.

The Estimation of Heritability

All studies of inheritance, whether of quantitative traits or of simple qualitative traits, are studies of the resemblances between relatives. If genetic differences are implicated in differences between phenotypes, we expect relatives to look more alike than unrelated people, since relatives are likely to have some of the same genes, inherited from common ancestors. Because parents pass their genes

p.73:

on to their children, we expect those children to look more like each other and more like their common parents than they look like the children next door. The basic methodology of genetic investigation is always such family comparisons. The trouble is that members of the same family share more than some of the same genes; they share some of the same environments as well. To get at the role of genes in shaping family resemblances, it is necessary, somehow, to cancel out the effects of environmental similarities. In experimental organisms, that is no problem: Several offspring of the same pair of parents are raised in controlled, separated environments together with offspring of other parents. Cows and mice can be removed from their mothers at birth and nursed mechanically or given to foster mothers. Seeds of plants can be grown in rigidly controlled environments. But the nuclear or extended family and social class are realities of human life that cannot be so easily cancelled out or manipulated by experimenters. As a result, there is a complete confounding of the effects of genetic similarity and those of environmental similarity on the resemblances among parents and children, uncles and nieces, brothers and sisters.

It is essential to distinguish between family resemblance and genetic resemblance. Many traits are familial but not heritable: For example, in the United States, the greatest similarity between parents and offspring is in two social traits, religious sect and political party. Yet no serious person would suggest that the very high family resemblance for these traits is a result of genetic determinaton.

Familiality is often confused with heritability, when it is supposed that the resemblance of parents and children is a demonstration of the power of heredity. Familial similarity is the observation. It should not be confused with the explanation, which may involve genetic and environmental commonalities. The cases in which people confuse the two can be deeply revealing of prior social assumptions. Although no one takes a similarity in the political affiliation of parents and children to be evidence of a gene for political party, it is widely believed that a similarity between parents and offspring in IQ scores is prima facie evidence that genes influence---even determine---intelligence. Alcoholism is commonly thought of as inherited because it often happens that a father and his son are both known to drink excessively, but we never hear it said that Presbyterianism is in the genes. Yet the evidence is really the same. Assertions that alcoholism might somehow have a simple biochemical basis confuse physiological sensitivity to a given amount of alcohol, which does indeed appear to be heritable, with a complex social behavior---drinking compulsively---for which there is no evidence of heritability.

The only solution to the dilemma of environmental similarity is, in principle, to conduct adoption studies. Children separated from their parents and their

p.74:

siblings at birth can be compared with their foster relatives and their biological relatives. If a trait is completely heritable, we would expect adopted children to resemble their biological relatives closely with respect to that trait, while they ought to be no more similar to their foster parents than to randomly chosen people. On the other hand, we would predict that adopted children would resemble their foster parents, rather than their biological mother and father, in such traits as religious affiliation. Thus, adoption studies lie at the heart of human quantitative genetics. At first glance, the most seductive of these studies are those of identical twins raised apart. Because identical twins result from the separation of a single fertilized egg into two complete organisms, they are genetically identical. Therefore, if they are similar even when raised apart, their similarities must be the result of common genes; their differences will reveal the effects of environmental variation. The trouble is that there are practically no cases of identical twins who have been raised in truly different environments. In the next chapter, we will review the evidence about these romantic lives. But it is sufficient, for the moment, to observe that, when twins are separated at birth in real circumstances, they usually turn out to be given to very similar families.

p.76:

The Heritability of Some Human Traits

heights of the children and those of their biological parents, but there is no correlation between the heights of the children and those of their adoptive parents. Thus, we are entitled to conclude that height is highly heritable. However, all of the children are a full 10 centimeters taller than their biological parents. The explanation is that the superior nutrition the children received in their adoptive homes resulted in considerable growth, but equally in all the adopted homes. As a consequence, they showed no person-by-person similarities to their adoptive parents, but the general amelioration of their environment made them as tall, on the average, as their foster parents and much taller than their own malnourished biological mothers and fathers.

The phenomenon illustrated by the growth of these hypothetical children is directly related to the issue of heritability and environmental similarity: Height, in this example, is totally heritable. The heights of the children are perfectly correlated with those of their biological parents. Yet that perfect heritability does not contradict the possibility of making the children grow taller by giving them better nutrition.

It should come as no surprise that very little is known about the genetic variation underlying those human traits that show continuous variation. The kinds of adoption studies that would be needed to distinguish mere familiality from real heritability are not common, partly because the data are expensive to acquire but also because adoptions are usually not truly random with respect to family environment. There is a lot of folklore about the heritability of various human traits but not a great deal of good evidence.

An example of the difference between everyday wisdom and science is the case of musical ability. It is widely believed that musical ability is "inherited," yet, when the evidence is examined, it evaporates. First there is the problem of defining "musical ability." Is it the ability to compose, or to perform on an instrument, or simply to carry a tune, or just to be able to distinguish ascending from descending sequences of notes? The evidence offered in support of the potion that outstanding musical ability is inherited is usually the datum that composers run in families. After all, look at the seven generations of musical Bachs, of whom two were truly outstanding, the Mozarts (father and son), the Scarlattis, or the Haydn-brothers. But anecdotes are not evidence. For every Bach family, there are scores of outstanding musicians who were the first and last of their line. Mendelsohn's father was a banker, Chopin's a bookkeeper, Schubert's a schoolmaster, Haydn's a wheelwright. Much less is known about their mothers, but none were known composers or performers. It is not at all clear that the number of musical families relative to the number of musical singletons is greater than might occur by chance, because no one has ever com-

p.77:

piled the statistics of familiality of musical performance and composition. In reading about the lives of composers, one certainly gets the impression that a large number were the children of minor performers---mediocre court tenors like Beethoven's father or free-lance double-bass players like the elder Brahms. Even granting familiality, it is impossible to say whether the correlation between parent and offspring is genetic in any part or is entirely environmental. The familiality of musical performance certainly seems to have been greater in the eighteenth and early nineteenth centuries than it is now, which is what one would expect from a time in which music was a trade like tailoring, tinkering, or baking, passed on from father to son as a way of making a living. Unfortunately for science, the prolific Johann Sebastian Bach did not have the foresight to have ten of his twenty children fostered from birth by pastry cooks and peasant families so that we might estimate the heritability of musical genius.

p.78:

The table above shows estimates of heritability, H², for a number of physical and personality traits of white women in the United States. These estimates are from a number of different studies. The physical measurements are clearly defined. The "personality" traits, however, refer to scores on various standardized tests or the outcomes of structured interviews. The reader should not take too seriously the names given these tests. Whether "masculinity," "neuroticism," or "intelligence" has been assessed is a matter open to debate, inasmuch as there is no agreed upon definition or measure of such constructs. The values given in the table have not been estimated from adoption studies. Rather, they depend upon an alternative approach, using twins, that is meant to solve the problem of environmental correlation. Identical, or monozygotic, twins are genetically identical since they come from a single fertilized egg. Fraternal, or dizygotic, twins are simply siblings who happened to have been conceived simultaneously from the fertilization of two eggs. They are no more alike genetically than any pair of siblings: They share 50% of their genes. If we can assume that the environmental correlation between monozygotic twins is the same as that between dizygotic twins, since a pair of twins of either sort is brought up as a pair in the family, then we can use comparisons between the two kinds of twin pairs to estimate heritability. Identical twins will be identical for a trait with 100% heritability, while fraternal twins will not. In general, symbolizmg the correlation between monozygotic twins as rM and between dizygotic twins as rD, an estimate of herirability is:

H² = (rM - rD)/(1- rD)

p.79:

It was this formula that was used to obtain the values in the table on the facing page. It is exceedingly unlikely, however, that the assumption of the method is correct. Dizygotic twins, even when they are of the same sex (as they are in all such calculations), are not treated like monozygotic twins. The very similarity in physical characters between identical twins causes them to be treated alike and to see themselves as alike. In many ways, their similarities are reinforced: Often they are given names that begin with the same letter, are given the same hair style, or are dressed identically. Thus, whatever genetic similarities exist become the cause of enforced environmental similarities that spill over onto other traits. Physical traits, such as height or cephalic index, are less susceptible to this bias. But if arm length and foot length are influenced by sports activity or by shoe style, these traits may be more environmentally correlated in monozygotic than in dizygotic twins. So-called "personality" measures will suffer greatly from this confounding. As a result, we really do not know what the heritabilities of these traits are, or even if they are heritable at all. Without random adoption studies, we cannot know how much of the observed familiality of a given trait is a consequence of common genes and how much is a consequence of common environment. In the next chapter, we take a closer look at the explanation of a trait that has received a great deal of attention from human geneticists, psycholgists, and social theorists, but about which we know much less than is asserted: intelligence.

p.88:

Mental Traits

One of the most obvious facts of our social existence is the immense variation in status, wealth, and power that exists among individuals and groups. Some people have a lot of money, while others have little; some have power over the condition of their own lives and over the lives of others, while most are relatively powerless. In all advanced countries of the capitalist world, the poorest 20% of families have about 5% of the total income, while the richest 5% have 25% of the income. Nor have the proportions changed significantly in the past 50 years. If we consider wealth rather than income, the distribution is much more asymmetrical. About 2% of the population of the United States own 25% of the wealth. But, if we exclude from-consideration such commonly held property as cars and houses, that 2% own a much greater fraction (75% of the corporate stock, for example). In addition to individual variation in wealth and power, there is marked differentiation by race. The median family income of blacks in the United States is only 60% that of whites, but their infant mortality rate is 1.8 times as high, and their life expectancy is 10% shorter, as was true 50 years ago.

An outstanding feature of status, wealth, and power is that they run in families. The children of oil magnates tend to own banks, whereas the children of oil workers tend to be in debt to those banks. It was exceedingly unlikely that Nelson Rockefeller would have spent his life pumping gas in a Standard Oil station. Certainly, there is social mobility in our society, but rather less than is celebrated in song and story. The best-known study of social mobility in the United States showed that 71% of the sons of white-collar workers were themselves white-collar, whereas 62% of the sons of blue-collar workers remained in that category. These figures vastly overestimate the amount of mobility in status, wealth, and power, however, because most of the movement between white- and blue-collar jobs is horizontal with respect to income, status, control of working conditions, and security. Clerks are no less workers because they sit at desks rather than stand at benches, and salesclerks, who constitute one of the largest "white-collar" occupational groups, are among the lowest paid and least secure of all workers. Of American business leaders in 1952, 83% had fathers who were either businessmen or professionals, 10% more than in the first quarter of the century when farm families contributed significant numbers of children to upward social mobility.

The fact that there is familial variation in status, wealth, and power in our society is deeply troubling to many---perhaps most---Americans. We are the beneficiaries of social revolutions, extending over the seventeenth and eighteenth centuries, that were supposed to abolish inherited wealth and power. The founding fathers called for "liberty, equality, and fraternity" and assured us that "all men are created equal." Of course, they meant literally men---women's

p.89:

suffrage did not come until the twentieth century---but they did not mean literally all men: Slavery persisted in the United States (and in British and French dominions) until the middle of the nineteenth century. One cannot make a revolution, however, with the slogan "liberty and equality for some"; so the notion that we really are all born free and equal is the cornerstone of our national ideology.

How are we to reconcile the manifest contradiction between the ideology of equality and the fact of inequality? On the one hand, one might claim that the inequalities that have characterized our society since the eighteenth century constitute a structural property of social relations themselves, that we do not really live in a community designed to give equal psychic and material benefits to all its members but that, on the contrary, our social system is built on inequality. That is, we might claim that the Declaration of the Rights of Man and the Declaration of Independence were propaganda designed to legitimize the ascent to power of a new aristocracy, the aristocracy of money. Needless to say, that point of view has not been very popular.

An alternative explanation---the one that has been dominant for two centuries---is that our society is as equal as it can be, given the natural inequalities among people. According to this view, the political and social revolutions of the eighteenth and nineteenth centuries destroyed artificial hierarchies and allowed the natural differences in ability to manifest themselves: Ours is an equalopportunity society in which everyone starts the race of life together and with

p.90:

the same social opportunities, but some are simply faster runners than others. It r is not sufficient, however, to assert that there are intrinsic differences in ability, for that alone would not account for the passage of social power from one generation to the next. It must also be claimed that the differences are biologically inherited---that, for example, the Rockefellers of the present generation are rich not because they inherited money and power, like aristocrats of the age of Louis XVI, but because they inherited the ability to acquire money andr power.

The idea that variations in intelligence, morals, gentility, and acumen are~ biologically inherited was a prominent theme of nineteenth-century literature. r Dickens and Eliot were its greatest exponents in English. Oliver Twist, it will be remembered, was raised from birth in a parish work house, the most vicious and ~ degraded social institution of the nineteenth century, where, together withr "twenty or thirty other juvenile offenders against the poor-laws, [he] rolled about the floor all day, without much inconvenience of too much food or too much clothing." Yet, from earliest youth, he is the epitome of gentleness, honesty, and morality, not to mention perfect English grammar and pronunciation. In all this he contrasts sharply with young Jack Dawkins, the Artful Dodger, a person of similar uphringing who is as low and cunning a specimen of lower- class English ragamuffin as can be imagined. The reason for their difference, which is the central mystery of the novel, is that Oliver is of upper middle-class parentage: His life story is the perfect adoption study showing that blood will out.

A more extreme example is George Eliot's hero Daniel Deronda who, as the adopted son of an English baronet, spends his time in gaming and the other idle but genteel pursuits that were characteristic of young men of his class. But, mysteriously, at about the age of 21, he develops an interest in Hebrew philosophy and falls in love with a Jewish girl. The reader is not too surprised to learn, at the end of the book, that Daniel is really the son of a famous Jewish actress. Nor was this only an English preoccupation: The most widely read French authors of the late-nineteenth century, Eugene Sue and Emile Zola, used the same themes. Zola's entire cycle of Rougon-Macquart novels was explicitly designed to show the determining power of heredity over social differences.

In the twentieth century, the claims for the dominance of heredity in human affairs have been expressed less in literature and more in science. In 1905, in a scientific paper on twins, E. L. Thorndike, who was unquestionably the leading American psychologist of the day, declared that, "in the actual race of life, which is not to get ahead, but to get ahead of somebody, the chief determining factor is heredity." The firm scientific basis for this dictum may be judged from the fact that it was written a mere 5 years after the rediscovery of Mendel's

p.91:

paper, but 5 years before the chromosome theory of inheritance, 10 years before the development of the statistical theory of correlation coefficients, and 13 years before the foundation of the theory of inheritance of quantitative traits. In the three-quarters of a century that have since passed, the central effort of human behavioral and psychological genetics has been to put a firm foundation under Thorndike's claim.

Click here to see picture: "The supposed history of the Kallikak family is often cited in psychology textbooks as an exampleof the critical role of heredity in determining mental and moral traits."
of the critical role of heredity in determining mental and moral traits."

p.92:

Although there have been many studies of temperament, alcoholism, mental disease, spatial perception, and other mental characteristics, the core of human psychological genetics has been the problem of "mental ability." This concentration arises from the widespread conviction that social success in modern industrial society depends increasingly on the power of abstract reasoning. In this view, cognitive ability determines the order at the finish line in Thorndike's "race of life," an ability that must be inherited. Concentration on intelligence has been made possible technically by the creation of instruments that are said to measure differences in cognitive ability: so-called intelligence, or IQ, tests.

The IQ test was first introduced in France in 1903 by Alfred Binet in an attempt to identify those children who were not doing well in school but who would benefit from extra remedial work. The test emphasized memory, vocabulary, and the ability to discriminate among related items. This test was later modified in the United States by L. N. Terman to produce the Stanford-Binet IQ test, which became (and still is) the standard against which subsequent tests have been validated. In adapting Binet's test, Terman (and the proponents of the mental-testing movement in general) wrought a subtle but fundamental change in purpose. From being a test for singling out children who could profit from remedial work in school, the IQ test became a method for arraying all children on a scale of intrinsic mental ability, which was presumed to be independent of schooling and experience. The belief that IQ tests measure something that is intrinsic to the individual and beyond the influence of the environment is not an incidental feature of IQ testing; rather, it is basic to it. The very name IQ (intelligence quotient) is derived from the operation of dividing the actual scores on the test by a correction factor for age, thus restandardizing the test for each age group and cancelling out the major developmental change that occurs in mental functioning. Those who have worked on developing the tests have also cancelled out the effects of sex by weeding out those questions on which boys and girls, on the average, score differently. The tests are claimed to be without cultural bias and, in some earlier tests, to be without any linguistic bias either. This is patently untrue for verbal tests, but a considerable effort has gone into constructing nonverbal mental tests to cope with the problem of culture. The tests have not, however, been restandardized to cancel out class or race differences, inasmuch as these are the very differences the tests are meant to reveal. That is, if differential intelligence is the cause of differential social success, then a test that claims to measure intelligence had better discriminate between individuals with a larger probability of social success and those whose chance of making it is small.

A great deal of attention has been paid to the supposed fixity of IQ as opposed to the development of abilities during a person's life history. For the same per-

p.93:

son, the scores on an IQ test that is repeated within a short time are highly correlated (r = +.95); so the test is said to be reliable. Tests given increasing numbers of years apart become more and more poorly correlated, especially if the second test is given well into adulthood, but the correlation of tests 10 years apart is reasonably good (r = +.80). The different components of an IQ test, such as vocabulary, analogies, pattern recognition, and so on, are also reasonably well correlated with each other, as are different tests, including both verbal and nonverbal ones. This correlation between tests and parts of tests is regarded as a demonstration that they all measure some underlying general intelligence, the so-called "g factor," that is reflected in various ways but is itself a fixed feature of the organism that neither develops with age nor is susceptible to environmental influence. Thus, the theoretical superstructure of mental testing has a strong commitment to a biological explanation of variation in performance. The stage is set for the demonstration of the heritability of intelligence.

How do we know that a test called an intelligence test does, in fact, measure intelligence? When the first IQ test was created, it was designed so that children judged on some other ground to be intelligent would do well on it. If the test had given high marks to those children everyone "knew" to be stupid, it would have been rejected. The original IQ tests were culled and adjusted so that the scores corresponded to teachers' and psychologists' a priori judgments about who was intelligent and who was not. The tests were tinkered with to make them the best possible predictors of school performance.

IQ tests vary immensely in form and apparent content, but many of them contain a good deal of material that obviously depends upon social class, home environment, and quality of schooling. Children are asked to identify characters and authors from English literature ("Who was Wilkins Micawber?"); they are asked to make judgments about socially acceptable behavior ("What should you do if a child younger than you hits you?"); they are asked to conform to social stereotypes ("Which is prettier?" when given a picture of a child with Negroid features and one with doll-like European features). The "right" answers to the questions do, in fact, correlate highly with scholastic performance. On the other hand, nonverbal geometric tests correlate less well with school performance. Not surprisingly, the ability to sit for a long time concentrating on a series of apparently meaningless questions is itself a reasonable predictor of school performance.

IQ tests were not developed from some general theory of intelligence and then subsequently shown, quite independently, to predict scholastic or social success. On the contrary, they were carefully designed to be predictors of scholastic

p.94:

performance, and the notion that they measure some intrinsic human characteristic, intelligence, has been added on with no clear justification. Indeed, there is no general agreement even about what intelligence is, and at least one educational psychologist has defined intelligence as the quality that IQ tests measure. We do not, in fact, know whether there is normal variation in intrinsic "intelligence" because we do not know how that mysterious property is to be defined. What is clear, however, is that there is considerable variation in actual school performance and that there are short-cut tests whose scores are highly correlated with that performance. That these tests are called "intelligence" tests instead of "school-performance predictors" should not mislead anyone into accepting their implicit claims.

The important social claim of mental testing is not simply that IQ tests measure intelligence but that they explain the variation in social success. The equation is simple: Variations in status, wealth, and power are the result of variation in intelligence; IQ tests measure intelligence; therefore, IQ tests predict the distribution of status, wealth, and power. But do they?

The standard measure of social success used by American sociologists is not social class, a European concept whose current validity is denied by most English-speaking sociologists, but socioeconomic status (SES, for short). This is a numerical score compounded of the income, occupation, and years of schooling of the male head of household. The observed correlation between a man's childhood IQ score and his adult success, measured either as SES or as income alone, is reasonably high in most studies, about .85. Thus, IQ seems to be a good predictor of social success. There is a problem, however: Economic and social success may have many causes, including intelligence, and these causes are themselves causally linked to each other. The apparent correlation between income and IQ might be purely an indirect effect of the other causes. For example, suppose that family background were both the direct cause of good performance on an IQ test and the direct cause of success in later life. Then IQ would have a strong correlation with later success, not because IQ caused later success but because both IQ score and later success were effects of the same underlying cause. fomily background.

The case of IQ and success is an example of a general problem in the analysis of causes of variation. Whenever there are multiple and complex paths of causation. the simple observation that two variables are correlated does not identify the paths of causation. The figure on the facing page is a very much simplified set of possible paths of causation relating socioeconomic background, schooling, IQ, genes, and adult socioeconomic status. An observed correlation between any two variables in the figure is evidence only that there are one or more

p.95:

Click here to see picture: "A simple scheme of possible causal paths connecting family backgrounds, genotype, IQ, and social status."

paths connecting the variables. It is not evidence of the direction of the causation or of how many steps the path contains. Thus, childhood IQ might be correlated with adult income (1) because IQ is a cause of school success, which, in turn, is a cause of income (path f-c); or (2) because it is a predictor of adult IQ, which is a cause of income (path g-d); or (3) because it is a cause of school success, which is a predictor of adult IQ, which is a cause of income (path f-e-d). All of these paths make IQ itself a cause of adult success. Suppose, however, that all the paths were abolished except a and i. Childhood IQ would still be correlated with adult success, but without being a cause of that success. On the contrary, it would be an effect of socioeconomic status rather than its cause.

Clearly, if we wish to understand the causes of social power, simple correlations are not sufficient. We need to look at the individual links in the scheme of causation. This can be done by examining each variable alone while holding the others constant. Thus, we can ask, "How much of the variation in adult income is predicted by variations in childhood IQ if we consider only people with the same schooling and the same socioeconomic background?" Conversely, we could hold IQ constant and see how much variation in success is explained by variation in parental success. When this was done by economists Sam Bowles and Valerie Nelson, the results were quite dramatic. The first graph on the next page shows the probability of a man being in the top one-fifth of the population in income given various amounts of schooling. Schooling here is measured not in absolute years but in what proportion of the population had as many or fewer years in school. For example, a man who fell in the lowest 10% of the population in number of years of school had only a 3.5% chance of being in the top income group, whereas a man who fell in the highest 10% of schooling had a 45.9% chance of being at the top of the income distribution. Is this because high IQ causes both school success and economic success? No. The solid bars in the graph show the probabilities when we consider only people with IQ scores near 100, which is the average for the population as a whole. There is virtually no difference. Even a man with an IQ of 100 was ten times more likely to

p.96:

Click here to see picture: "Relation between the probability of high income for the population as a whole (light-color bars) and for only persons having average IQ's (dark-color bars) and years of schooling (upper graph) or social background (lower graph)." [caption]

receive a high income if he was in the top 10% of schooling than it he were in the bottom 10%. Holding IQ constant does very little to the relation between years of schooling and eventual success. The second graph shows a similar comparison when family socioeconomic score rather than schooling is considered. Men whose fathers were in the top 10% of the social hierarchy were ten times more) likely to receive high incomes than those who came from the poorest social~ stratum (43.9% compared with 4.2%). This changes only a little if we consider

p.97:

only men with average IQs. As the solid bars show, a man of average IQ from an upper-class family had an advantage of seven and one-half times over a man of the same IQ from the poorest family. If there is some intrinsic quality that differentiates the successful from the unsuccessful, IQ tests have failed to capture it. If such tests really do measure intrinsic intelligence, as they are claimed to do, then one can only conclude that it is better to be born rich than smart.

IQ and Genes

To postulate an inborn and unchanging basic intelligence is not the same as postulating genes for intelligence. The relation between the properties of being unborn, unchanging, and genetic are more complex than they appear. First, inborn does not mean genetic. Of the physical and physiological differences among individual people that are present from birth, many are caused not by genetic differences but by developmental noise. Small accidental alterations in the growth patterns of nerve connections in the fetal brain may produce considerable differences in mental functioning. Second, genetic does not mean unchanging, as this book has repeatedly stressed. Gene action is directly responsive to environmental signals, and the complex development and metabolism of the whole organism put it into constant interaction with the external world. Third, inborn does not mean unchanging. "Blue babies" born with an anatomical defect of the circulation can be made quite normal by a straightforward operative procedure to close off the connection in the blood supply that should have been closed off naturally during fetal development. Finally, differences between people may be unchanging without being either inborn or genetic, as those who have lost limbs, sight, or hearing in accidents can testify.

These facts have not been understood by psychologists, who have usually assumed that, if intelligence was really intrinsic, it must be genetic and that, if it was genetic, it must be unchanging. An example of this misunderstanding is a famous article by the educational psychologist Arthur Jensen, which posed in its title the question "How much can we boost IQ and scholastic achievement?" and concluded "not much" on the grounds that IQ is largely hereditary. To complete the circle of confusion, part of the evidence offered by Jensen in support of the notion that IQ is hereditary was its supposed constancy over a person's lifetime.

IQ scores are distributed as shown in the graph on the next page in each population for which the test was designed. The mean score is 100, and the standard deviation is 15 points. The distribution is symmetrical around the mean and has a bell shape called the normal distribution. There is nothing particularly revealing in any of these characteristics of the distribution of IQ scores, because the

p.98:

Click here to see the picture: "Distribution of IQ scores in a population for which the test has been standardized. The percentage of the population falling within each range of scores is given above the horizontal axis; so 17% of the people have an IQ between 80 and 90." [caption]

tests were designed and the scoring system was adjusted in order to produce ar normal distribution of scores with a mean of 100 and a standard deviation of 15. When the Japanese version of the Wechsler Intelligence Scale for Children was developed, for example, it was carefully tailored to produce that distribu-r tion among Japanese school children. In the belief that important issues of social practice would be determined by knowing the heritability of IQ scores, psychological geneticists have made a considerable effort to partition the variance of the IQ distribution into genetic and environmental fractions and to) establish a heritability ratio. Quite aside from the question of what use such information really is to social policy, the problems of the estimation itself arer enormous.

The question, as always, is to separate genetic from environmental similarity in families. The figure on the facing page gives observed correlations in IQ scores between various kinds of related and unrelated people. Both medians and ranges over various studies are given, together with the correlation expected if the heritability were 100% and the effects of genes on IQ were additive. Ther more closely related the people, the higher the correlation in their IQ scores. The median correlations between parent and child, between siblings, and between dizygotic twins are all close to the simple genetic expectation of .50. Unrelated people have a much lower correlation, and identical twins have a much higher one. An increasing correlation with increasing family relationship is predicted ~ by any theory of the causation of IQ, however; so the gross pattern is not very r informative.

The very large ranges for each class of relationship are disturbing. It is diffcult to have too much confidence in the studies of parent-child correlation when the results are evenly spread over a range of r = .20 to r = .80, or of the siblings who are, again, remarkably evenly spread between r = .30 and r = .80. Under the circumstances, the fact that their median values fall close to .50 seems more like a numerical artifact than the revelation of any biological reality. These ranges correspond to a range of heritability between H² = 40% and H² = 160%---not a very reassuring result.

Comparisons between people with the same genetic relationships but different environmental circumstances and those between people with different genetic relationships but the same environmental circumstances have been most

p.99:

Click here to see the picture: "Observed correlations in IQ between people of different degrees of relationship. Each point is a separate study and each vertical line shows the median value for each degree of relationship." [caption]

often used to estimate the heritability of IQ. The IQ scores of unrelated people are uncorrelated when they are reared apart, as indeed they must be under any theory of causation, but their scores have a median correlation of about .25 when they are brought up together. One study produced a correlation of .30 for the IQ scores of unrelated persons raised apart, which should caution us against accepting scientific results too readily. There is no way that a truly random sample of unrelated persons can be correlated in their IQ scores. Either the study was badly designed or the result was a statistical fluke. In either case, it gives one pause. Comparing the correlation between the IQs of foster parents and those of their foster children with the correlation between the IQs of biological parents and those of their children again show the effect of a common environment. At the other end of the scale, the IQ scores of identical twins reared together have a median correlation of .88, whereas those of identical twins reared apart have a correlation of .75. Again, the effect of a shared family environment is detected.

If we take the results in the figure above at face value for the moment, we can see that there are a variety of ways in which to estimate the heritability of IQ, and they all agree in giving a fairly high value. The .25 correlation between unrelated persons reared together might be regarded as a direct estimate of the effect of a common environment. Monozygotic twins raised together have both genes and environment in common, and their median correlation is .87. So, correcting for the effects of a common environment, we can get an estimate of heritability of .87 - .25 = .62. Alternatively, we can use the standard compar-

p.100:

ison between dizygotic and monozygotic twins. The median correlation for: monozygotic twins is .87 and that for dizygotic twins is .53. Therefore,

However, the trouble is that we cannot take any of these studies or theories at face value. We have already discussed (p. 79) the dubiousness of the assumption that identical and fraternal twins are really treated in the same way; so the estimate of .72 is too large by some undetermined amount. This problem also contaminates the first estimate. Can we really assume that the correlation of .25 between the IQ scores of unrelated persons reared together is also an adequate estimate of the strength of the effect of a common environment on identical twins ? If not, then we have undercorrected for common environment in the first estimate as well. Indeed, we ought to be suspicious of hanging too much generality on any study of what is an extraordinary human relationship.

Twins separated from birth have a fatal fascination for the human geneticist, just as they do for the romantic novelist. Alexandre Dumas's Corsican Brothers, separated by a knife at birth but feeling simultaneous pain and pleasure over a distance of many miles, are but the literary counterparts of Sir Cyril Burt's no less fictitious identical twins who have identical IQs although they have never seen each other. The scenario seems perfect for an examination of the genetic effects on human variation: Identical twins have identical genes; if, raised in unrelated environments, they nevertheless show similarities in their tests, these similarities must be genetic. In fact, the correlation between the IQ scores of identical twins raised apart is a direct estimate of the heritability of IQ. As the figure on page 99 shows, the estimate is .75.

A little rumination on the subject of identical twins raised apart begins to raise questions. Nineteenth-century novels aside, what do we imagine the circumstances to be that will separate identical twins in earliest infancy? Are the twins really raised in utterly different environments ? There are only four studies of IQ in separated twins reported in the literature on this subject, and these have been the subject of a detailed analysis and review, by the psychologist Leo Kamin, that produced disquieting results. As might be expected, none of the studies contains many twin pairs: Juel-Nielsen, 12 pairs; Newman et al., 19 pairs; Shields, 44 pairs; and Burt, 53 pairs. Newman and Shields got their pairs by advertising in newspapers or on television and then culling the respondents' mailed replies. Thus, it seems there was self-selection on the part of the respondents, each of whom was sufficiently similar to the other twin to regard the pair as identical and each of whom was in contact with his or her "separated" twin.

P.101:

Only Shields provided detailed life histories for his twins, and it turns out that they were not really separated at all. In real life, twins are separated at birth because the mother has died, or because the parents cannot afford to keep them both, or because they are too sick to do so. The children are typically given to aunts, sisters, or best friends, and are brought up in neighboring houses in the same towns. In Shields's study, every twin pair but 4 was raised by close relatives, close friends, or neighbors. Twin pairs separated at birth and raised in unrelated environments belong more to the realm of romance than to reality.

The largest and most widely used study of separated twins is actually a series of studies presumably carried out by Sir Cyril Burt and his collaborators over a period of 20 years. The researchers maintained that there was no significant correlation in economic status between the families raising the separated twins. No details were given. When Kamin examined the studies carefully, curious anomalies appeared. Sample sizes were reported differently, or sometimes not at all, in different reports. No details were given about the IQ tests. Test scores were adjusted to account for the interviewers' subjective perception of whether they adequately reflected the similarities of the twins. Correlation coefficients computed on different sets of twins nevertheless agreed over and over to the third decimal place. The original data, moreover, were supposed to have been lost in a laboratory fire.

Several years of investigation by Kamin and, later, by the medical writer Oliver Gillies finally revealed that Burt's twin studies were a complete fabrication. Burt's named collaborators did not exist, the test scores did not exist, and, it appears, the twins did not exist either. For this reason, all studies by Burt and his "collaborators" were left out of the figure on page 99. Burt apparently had a long history of fabrication, including laudatory reviews of his own work, published under fictitious names in a journal of which he was the editor.

The reactions of psychologists and human geneticists to the succession of revelations about Burt's work are in themselves revealing. Some said that Burt had merely "carelessly reported" his work. One can be careless, it seems, to the third decimal place. Others said it was a phenomenon of Burt's old age, a result of senility, but it became clear that the frauds went back to his early work as well. Most disturbing of all, some of his senior colleagues said that they had always doubted Burt's reports but that they had never challenged him because he "said it with such style."

Burt's fraudulent reports and the reactions of his colleagues to them are only the extreme of a general phenomenon in the study of human mental and temperamental variation. Most studies of the heritability of mental traits are marked by one or more serious methodological defects, including (1) very small sample size; (2) confusion of observed correlation between relatives with genetic correl-

P.102:

ation; (3) selective adoptions in fostering studies; (4) subjective ratings of similarities; and (5) after-the-fact statistical adjustments that bring the data more in line with genetic expectations. Any or all of these defects would automatically disqualify a research report for publication in a scientific journal if the subject were milk yield in cattle. Journals of psychology and behavioral genetics regularly publish them, and no progress in the rigor of such work is apparent. As late as 1979, the major scientific journal of behavioral genetics published an estimate of the heritability of human IQ based entirely on the observed correlation between parents and offspring in normally structured families, even though the editor knew, and has elsewhere stated, that there is no way of knowing in such cases how much of the correlation is a consequence of shared family environment. It is impossible to avoid the conclusion that there is a deeply established prejudice in favor of a genetical explanation of human behavioral variation.

Adoption Studies

In principle, it should be possible to assess genetic influence on IQ variation from adoption studies. The figure on page 99 includes data from studies comparing either the IQs of parents with those of their foster children or the IQs of siblings raised apart. On the face of it, the two sets of studies give very similar results, although the expected genetic correlation between the IQs of siblings raised apart is .5 and that between the IQs of foster parents and those of their foster children is zero. Thus, there is not much evidence here for heritability.

The ideal adoption study would compare the IQ of an adopted child with the IQs of its adopted parents and those of its biological parents. This ideal is difficult to realize, howevet, because it is usually impossible to obtain the data about the biological parents. A substitute is to use a different group of children and their biological parents, but to try to match the characteristics of the biological and adoptive families as closely as possible. In either case, it is vital that the adoptions have been made at random. That is, it is essential that children of parents whose IQs are high are not fostered into families with higher than average IQs. Otherwise, an IQ correlation may appear between foster children and their biological parents that may be a consequence not of the biological relation but of the environment in which they were raised.

There have been three large adoption studies that have used this comparative design. Two, the studies by Burks and Leahy, compared foster families with a different, but supposedly matched, set of biological families. The third, the study by Skodak and Skeels, used educational levels of the foster mothers, IQs of the children, and IQs of the biological mothers of the foster children. Unfortunately, no IQ tests were given to the adoptive mothers. The results shown in the table on the facing page seem very strong evidence for genetic effects, inasmuch as the correlation of children with their foster parents in each case is so tl~natelv no 1() tests were t!iven to the adontive mothers. The results shown in the table on the facing page seem very srong evidewnce for genetic effects, inasmuch as the correlation of children with their foster parents in each case is so

p.103:

much lower than the correlation of the children with their biological parents. A careful examination of these studies by Kamin, however, raises some serious questions about their design. The study by Burks included many severely retarded children, which, of course, greatly reduced the IQ correlation with the adoptive parents, who in general come from higher socioeconomic categories and have higher IQ scores than the average for the population as a whole. In both the Burks study and the Leahy study, the matching of biological and foster families was poor. Adoptive parents were older, had incomes that were higher by 50%, and had fewer children, as might be expected. They were, in general, much less variable than the biological families in nearly every respect. In the Skodak and Skeels study, there were selective adoptions, with children of highly educated mothers being placed in higher-status homes. So, from such studies we really do not know what the heritability of IQ is.

The most striking and consistent feature of adoption studies is not usually much commented upon by those interested in demonstrating genetic effects: Whatever the correlations may be between the IQs of children and those of their biological parents, the phenomenon of adoption raises children's IQ significantly. In the study by Skodak and Skeels, the biological mothers' IQs averaged only 86, one standard deviation below the average for the population. By contrast, the mean IQ of their children who were raised by adoptive families was 117, one standard deviation above the mean for the population. In a study of orphanage children in the United Kingdom, the same phenomenon was seen. Children taken into the orphanage in early infancy had an average IQ of 105 if they remained in the orphanage until they were about 5 years old, 100 if they were returned to their biological mothers, but 115 if they were adopted. These observations are just what is to be expected from the social characteristics of adopting parents: They are generally middle-class and upper-middle-class couples, with few or no children of their own, who have the financial power, the motivation, and the class background to produce "intelligent" children. There is, of course, no contradiction between this adoption effect and the possibility that IQ scores may be highly heritable. "Genetic" does not mean "unchanging." No matter how high the heritability of IQ might prove to be, upper-middle-class families tend to produce upper-middle-class children.

//END EXCERPTS//