The Definitive Word on Intelligence
Arthur R. Jensen, The g Factor, Praeger Publishers, 1998, 648 pp.
Arthur Jensen of U.C. Berkeley is one of the greatest social scientists of our time. He virtually single-handedly resurrected the scientific study of intelligence, and he has been at the center of many breakthroughs in this field. Needless to say, he is a courageous man, who has never let hysterical opposition or even death threats keep him from studying some of the most important and contentious issues we face.
The g Factor is only the latest of the many publications that resulted from what can now be seen as a watershed event: the 1969 appearance in the Harvard Educational Review of Prof. Jensen’s famous article on the heritability of IQ and how difficult it is to raise. This article not only reestablished the connection between genetics and intelligence but set the direction of Prof. Jensen’s career. He has since written countless articles in this field and three major books: Educability and Group Differences (1973), Bias in Mental Testing (1980), and now, The g Factor.
These books chart the recent remarkable progress in the study of intelligence. If Prof. Jensen had so dominated any less controversial field he would certainly be a candidate for the Nobel Prize. Unfortunately, his real stature is recognized only by a small number of specialists and professional colleagues, but the implications of his work continue to reverberate through the larger society. Whatever recognition he may ultimately receive, his work has gone far to set the study of mental ability once more on a firmly scientific basis.
The g Factor
This book is an investigation of the nature of intelligence, the extent to which it is under genetic control, and its uneven distribution between individuals and groups. The first part is a complete and sometimes technical treatment of “the g factor” itself, which appears to be a unitary mental ability underlying all activities we think of as requiring intelligence. “Factors” are the end result of a mathematical procedure called factor analysis, and the g factor is the “general” factor of intelligence, first hypothesized by the British psychologist, Charles Spearman (1863-1945). Spearman thought of g as a direct analogy to the “G” of physics, that is Newton’s gravitational constant. Spearman’s view, substantiated by almost a century of research, was that g is of central importance to psychology just as g was to Newtonian physics.
G can be thought of as the undifferentiated raw cognitive power of the brain. It cannot be directly measured, but it manifests itself in all types of cognitive activity, and people who are good at one kind of mental test tend to be good at all of them. To use the statistical term, a person’s different abilities are correlated, and similar abilities tend to correlate most closely with each other. For example, someone who is exceptionally good at any mathematical test is likely to be very good at all mathematical tests — but he is likely to perform well on verbal tests, too. As we will see, g is at work when even the smallest demands are made on the mind.
If people take enough different kinds of mental tests, their scores can be analyzed for factors, or the tendency of the correlations between similar abilities to cluster in groups. There will be factors for such things as verbal, musical, mathematical, and spatial manipulation abilities. Further analysis of these factors reveals a fundamental factor common to them all, which is the g factor.
We can therefore imagine a series of different factories in the brain, all powered by the same energy source. One of the factories manufactures solutions to mathematical problems, while another produces correct understandings of words and sentences. Other factories produce solutions to other kinds of mental problems, but all of them can be thought of as running off a common power source, which is g.
People differ in the efficiency of their individual factories, which is why smart people have different strengths in different areas despite being smart in a general sort of way. But people differ most significantly in the level of the general power source, or g. Someone with an IQ of 100 may have a math factory that is relatively more efficient than his verbal or music factory, but even in math he is likely to fall well behind someone with an IQ of 130 whose math factory is relatively less efficient than his verbal factory. It is the difference in levels of power available to all of a person’s factories that produce the marked differences in ability that characterize our species.
Many kinds of mental performance can be taught and people can show improvement, but what is improving is an ability that is not g. As Prof. Jensen explains, “At the level of psychometrics [mental testing], ideally, g may be thought of as a distillate of the common source of individual differences in all mental tests, completely stripped of their distinctive features, of information content, skill, strategy, and the like.”
Interestingly, Prof. Jensen reports that it is at the highest levels of g that people show the most variation in abilities that are independent of g. Thus, very intelligent people may have markedly different mental ability profiles despite similar levels of g. If all the factories are getting lots of power from their common source, some of the factories are likely to be unusually efficient so that the pattern of different levels of efficiency can differ considerably from one smart person to another.
Some critics have complained that g is not real because it cannot be measured directly and must be derived by a complex statistical process. Prof. Jensen shows that it is not, for this reason, artificial. If there were no g factor, sophisticated mathematics could not coax it into existence. Moreover, the same g factor is found in all human populations, and can be derived from the results of mental tests prepared by people who have never heard of g or who have even doubted there was such a factor. g can be calculated only because it exists, and in that sense is purely objective. Prof. Jensen believes that it reflects one of the basic functions of the brain, and that although all normal people share the same biological structures they differ greatly in the efficiency of certain neurological processes.
Direct assessment of brain functions gives strong evidence that g is a real, physiological phenomenon, and Prof. Jensen has been a pioneer in using what are called elementary cognitive tasks (ECTs) to study intelligence. The simplest sort of ECT involves a test device with two push-buttons (see illustration, below). The subject holds down the black button while he waits for a light to go on inside the smaller, white button. He then presses the illuminated button as quickly as possible. This measures two things. The first is reaction time: the time between the light going on and the subject taking his finger off the black button. The second is movement time: the time it takes the subject to move his finger from the black button to the illuminated button.
Obviously, this is a very simple (indeed, elementary) task, though tests of this kind can be made more complicated. For example, there can be a number of smaller buttons that can light up in different patterns, requiring the subject to make slightly more complicated decisions before moving his finger. We do not think of this sort of thing as mentally demanding — no one ever “fails” these tests — but the neurological processing that goes into these very simple tasks is closely related to intelligence.
Prof. Jensen has found that reaction speed is strongly correlated with g level, but that the highest correlation is between g and consistency of reaction time. With a set of scores from various different ECTs, it is possible to achieve a 0.7 correlation with g as calculated from conventional IQ tests. This approaches the g correlation (0.8) of Ravens Progressive Matrices, the IQ test that comes the closest to measuring g itself. Surprising as it may seem, careful monitoring of the processes that underlie ECTs can give results that are so reliable they rival pencil-and-paper tests.
ECT performance matches group differences in intelligence. It is worse in children than in adults, and better in gifted children than in normal children. Blacks have quicker movement times than whites while whites have quicker and more consistent reaction times. Asians do slightly better than whites, and performance for no group improves with practice; ECTs appear to measure something basic to the brain.
Another direct assessment of mental processing is the inspection time test. This uses an instrument called a tachistoscope to throw an image on a screen for a very brief period. Starting at the millisecond level, which is too quick for anyone to see the image, the exposure is gradually increased until a subject can just make it out. There is a correlation of .54 between speed of inspection time and IQ — remarkably high for a task that is so different from an IQ test. Once again, the test seems to be measuring a neurological process closely associated with mental processing.
Yet another direct assessment is the study of brain waves. Prof. Jensen explains that a wave pattern called average evoked potential can be analyzed in specialized ways that show a surprisingly high correlation with IQ.
Finally, researchers have devised something that is essentially a direct test of brain efficiency. The brain’s fuel is glucose, or simple sugar. When a radioactive isotope of glucose is injected into a subject’s blood stream it is possible to measure the rate at which the brain takes it up and metabolizes it. When rate of metabolism is measured while subjects are taking an IQ test, the high scorers use less sugar than the low scorers, with a remarkable correlation with IQ of around .7 or .8. The less powerful brains get wrong answers despite burning more fuel. If we return to the analogy of the brain as composed of factories, the common power supply simply appears to be less efficient.
If advances continue to be made in direct assessment of the brain, conventional IQ testing may be superseded. This would certainly silence any complaints about “test bias.”
Because the issue of whether education or environment can influence IQ levels is central to so much policy-making, The g Factor thoroughly covers the question of heritability. Kinship and adoption studies have provided some of the most illuminating data on this question, and Prof. Jensen reports them in detail.
Some of the most significant findings are the correlations of IQs of identical twins reared in the same family (.86), identical twins separated at birth and reared in different families (.75) and fraternal twins reared in the same family (.60). That identical twins separated at birth should have more similar IQs than fraternal twins reared by the same parents is perhaps the single most powerful argument for the view that genes have a greater effect on IQ than environment. As Prof. Jensen points out, “similarities in the MZA’s [monozygotic (identical) twins reared apart] environments cannot possibly account for more than a minute fraction of the IQ correlation of +.75 between MZAs.”
Studies of siblings and adopted children likewise confirm the power of heredity in determining differences in IQ, and it is now generally agreed among specialists that 60 to 80 percent of human IQ variation is due to genes. This does not mean, however, that the remaining environmental influences are well understood or can be used to raise IQ. As Prof. Jensen explains, “a large part of the specific environmental variance appears to be due to the additive effects of a large number of more or less random and largely physical events — developmental ‘noise’ — with small, but variable positive and negative influences on the neurophysiological substrate of mental growth.”
What is this developmental “noise”? “[S]uch effects as childhood diseases, traumas, and the like, as well as prenatal effects such as mother-fetus incompatibility of blood antigens, maternal health, and perinatal effects of anoxia and other complications in the birth process, could each have a small adverse effect on mental development.” These appear to be the kind of non-genetic factors that influence IQ, and they are not the sort of thing that can be easily manipulated.
As Prof. Jensen makes emphatically clear, the non-genetic influence comes only slightly, if at all, from what are called between-family differences: education of parents, social status, family income, school quality, etc. Liberals believe that these are the crucial factors that make people different from each other, but liberals are wrong. IQ (like other personality traits) is astonishingly impervious to any but the most degraded and unfavorable environments.
Prof. Jensen calls the environmentalist view “the sociologist’s fallacy.” It is true that children from wealthy homes tend to be smarter than children from poor homes, but wealth does not make them smart. They get genes for intelligence from their smart parents, and their parents are likely to be well off (and have homes full of books and speak in complete sentences) because they are smart. Of course, children do differ from their parents in intelligence, and these differences explain how families rise and fall. A person’s IQ has a correlation of .7 with his own adult socio-economic status but only about .4 with that of his parents.
Error though it be, the sociologist’s fallacy has driven not only an enormous number of government uplift programs but several well-publicized private efforts to raise the IQs of poor black children. Prof. Jensen reviews the results of the Milwaukee Project, Head Start, and the Abecedarian Project, some of which made extraordinary attempts to improve environments.
In some cases, the early results were very encouraging: gains of 20 or even 30 points compared to control groups. But as Prof. Jensen convincingly argues, what the children learned at intensive “infant stimulation centers” and the like was information and strategies that helped them take the tests. g very probably did not change. In most cases, administrators did not give a battery of tests and attempt to calculate g. Instead, they gave the same test at different ages and rejoiced to find improvement.
Professor Jensen gives a striking example of how training can improve test results without raising g. He notes many children’s IQ tests have a memory component: How long a string of letters or numbers can the child repeat back to the tester? Most adults can’t remember more than about seven numbers, but with lots of practice and training, people can remember as many as 70 or even 100 digits. They can do this because they develop a specific strategy or skill, not because their memory or g level has improved. The tricks a person uses to remember 70 digits are so specialized, in fact, that they do not even help the same person remember more than an average number of letters (rather than digits)!
Children who took part in these widely-acclaimed IQ-raising programs probably learned specific skills of this kind during the thousands of hours of instruction they received. But even the most intensive enrichment programs had virtually no permanent effect on school performance or IQ, which suggests that g itself was unchanged. Prof. Jensen concludes that IQ cannot be appreciably increased by specialized education.
It is true that the IQ test scores of children are affected to some degree by the environment their parents make for them. This is almost certainly because they learn more facts and absorb test-taking strategies and not because the love and care of good parents improves g. In fact, as children grow older they create environments that suit their own genetic endowments, and Prof. Jensen is categorical about what then happens: “By adulthood, all of the IQ correlation between biologically related persons is genetic . . . [T]he environmental contribution to the familial correlations is nil.” Surprising as it may seem, once a child grows up, his IQ score is similar to that of family members only because he is genetically related to them, not because they spent many years in the same household.
Prof. Jensen is equally forthright in explaining that genes account for the well-established IQ differences between the races. First, he points out that approximately half — or 50,000 — of the genes that vary in human beings play a role in brain functions, and that 30,000 affect the brain exclusively. It would be astonishing if genes did not play a central role in intelligence and if the races, which differ physically in so many ways, did not differ in brain function.
He also offers an arresting refutation of the fashionable view that race is purely a social construct and is not biological. Prof. Jensen likens race to the visible colors. A rainbow forms when the wave-length of light changes continuously and uniformly, but we do not perceive a continuous change. Instead, we see distinct bands of color. Though there may be some blurring of race at the edges because of cross mating, races are as distinct as the bands of visible color. Prof. Jensen also cites the increasingly persuasive genetic evidence for the biological distinctness of different populations (see figure, below).
A number of elegant demonstrations based on the principle of regression toward the mean strongly suggest a genetic origin for group differences. This principle is a biological law according to which parents who are at the extremes of any trait are likely to have children who are less extreme. Two very tall parents are likely to have children who are not quite so tall, and two very short parents are likely to have children who are not quite so short. In the children, these traits revert toward the average, or the mean. The same effect is found in intelligence, but the mean toward which the black IQ regresses is a full 15 points lower than the white mean.
Therefore, when black couples and white couples are matched for IQ, the black/white IQ difference in their children increases as parental IQ increases. In other words, high IQ is an anomaly in all races, but more of an anomaly for blacks than for whites, and the children of high-IQ blacks regress further because they are regressing toward a lower mean.
Prof. Jensen reports a study of high-IQ children in one school district that provides more evidence for the difference in means. When white and black students were perfectly matched for IQs of 120, the average IQs of the siblings of the whites was 113 whereas the average IQs for the siblings of the blacks was 99. Among blacks, an IQ of 120 is simply a much greater deviation from the norm than it is for whites, and this is reflected in the IQs of their more ordinary siblings.
Regression toward the mean explains something that has always baffled the “sociologists:” children of low-income whites (and Asians) get better SAT scores than the children of high-income blacks. If environment controls IQ, the children of wealthy blacks should be enjoying the benefits of good environment. They are, but those benefits are meager and do not make up for the effects of heredity and the lower mean toward which black children regress.
There is no non-genetic explanation for group differences that can account for phenomena of this kind, but they are perfectly consistent with widely accepted principles of genetics. Specialists understand the force of arguments of this kind, which is why the view that “racism” and other environmental factors cause the black/white IQ gap persists mostly among the ignorant — who are the great majority.
More strong evidence for a substantially different biological mean for IQ is found in studies of the low end of the IQ distribution curve as well. Mental retardation — IQs below 70 — is generally of two types, familial and organic. Familial retardation occurs in children who are otherwise normal but were simply dealt a very poor hand of the genes that affect intelligence. Given a normal distribution of intelligence, a few people are inevitably going to have very low IQs, just as a few will have very high ones. Organic retardation, on the other hand, is caused by clear biological defects, like Down’s syndrome (Mongolism) and children who suffer from it are obviously abnormal.
An important racial difference lies in the fact that half of whites with IQs below 70 are organic retardates but only 12.5 percent of the blacks are. The source of this difference is the racial disparity in naturally occurring distributions of intelligence. Given that the distribution curve for black intelligence is shifted approximately 15 points toward the left, a substantially larger proportion of otherwise normal blacks will fall below an IQ of 70.
The opposite is true at the high end of the curve. The percentage of whites with IQs higher than 130 is 20 times that of blacks. Because there are approximately six times as many whites as blacks in America, in real terms there are perhaps 120 times more whites than blacks with IQs at this level. This is why, without racial preferences, it is impossible to admit large numbers of blacks to competitive universities or to promote them to challenging positions.
Brain and head size studies likewise confirm the biological origins of group differences. It is now well established that brain size correlates with intelligence, and Prof. Jensen reports that the heads of black newborns are a full .4 standard deviation smaller than those of whites.
Likewise, it has long been known that near-sightedness, or myopia, is correlated with intelligence; children with IQs over 130 are three to five times more likely to be nearsighted than children with normal IQs. There seems to be no functional, cause-and-effect connection between myopia and intelligence, but a pleiotropic relationship exists in that some of the same genes affect both traits. Intelligence and myopia are somehow “side effects” of each other to some degree. Prof. Jensen finds that myopia is most common in Jews, next in Asians, then in whites, and least common in blacks — precisely the distribution one would expect. Moreover, reading does not cause myopia. An oculist can examine the eyes of children who are too young to read and who are not yet near-sighted, and accurately predict whether they will need glasses later in life.
It is well known that the test score gap between blacks and whites varies from one IQ test to another, and that the gap narrows on the least abstract, most information-laden tests. Prof. Jensen explains that the real difference lies in the extent to which a test measures g; the more g-“loaded” a test is and the fewer specific non-g abilities it measures, the greater the black/white gap.
Like many others who have studied the question, Prof. Jensen finds that the racial gap in IQ is increasing because of dysgenic birth patterns. In both races, less intelligent mothers are having more children than more intelligent mothers, but the disproportions are higher among blacks than whites. Also, since blacks have children, on average, two years earlier than whites, the generation time for blacks is shorter and dysgenic effects spread more rapidly.
One of Prof. Jensen’s most interesting racial findings is that the average IQ difference for blacks and whites in the same social class is 12 points — almost as great as the average difference between the two races (there is an average 17-point difference between any two people in the population picked at random). This is explained not only by preferential policies but also by racial differences in IQ distribution. If, for example, a demanding profession requires a minimum IQ of 125, blacks in that profession will tend to have IQs that cluster at the minimum, whereas whites will show greater variety. Because of this effect, the IQ gap between blacks and whites in the same social class narrows as one moves down the social scale.
Prof. Jensen finds that the geographic distribution of IQ is also uneven. For both blacks and whites, there is a continuous gradient that rises from the south towards the north and west. The gradient is sharper for blacks than whites, and both gradients are apparent in pre-school children, so regional differences in education do not explain it.
It has been widely reported that from infancy black children develop motor skills more rapidly than whites. Interestingly, Prof. Jensen finds that lower-class children (both white and black) develop more quickly than upper-class children, which suggests that slow maturation and high intelligence are correlated not just between races but within races.
For the most part, Prof. Jensen does not make policy recommendations; the facts alone are persuasive enough. He does point out, though, that life itself is a kind of continuous intelligence test, and that high g is one of the most important ingredients of success. He explains that scores on a highly g-loaded test are the best indicators of performance on any but the most specialized jobs. IQ is an excellent predictor for performance even on jobs that require manual dexterity and coordination. To a remarkable degree, g is the central mental characteristic of humans. Of course, intelligence is not everything. It takes more than brains to become a doctor — it takes persistence and discipline, too — but persistence is not enough. For many things, a certain level of g is indispensable, and low g cuts off desirable options at every stage of life. Low g is therefore a more accurate predictor of achievement than high g, since a lack of intelligence cannot usually be made up for by other qualities whereas high intelligence can be wasted.
When people with low g are scattered through otherwise normal communities it affects only individuals. Friends and relatives step in to help them. However, as Prof. Jensen points out, when people of low intelligence gather in large numbers, as they do in welfare housing, society falls apart. Prof. Jensen notes that in America there are now entire apartment blocks in which, even with welfare, the residents cannot get by without help from social workers. Dysgenic trends and increased immigration of low-g stock mean areas like this will only expand.
In this connection, Prof. Jensen makes some interesting observations about adult illiteracy. Most people assume that the cause is poor schooling, but he argues that the problem is usually not the process of decoding written language but understanding it. Most illiterates do no better on reading comprehension tests when the selections are read to them than when they try to do the reading themselves! Illiteracy, in Prof. Jensen’s view, is much more a problem of low g than of somehow not learning how to read.
There are a few points on which Prof. Jensen’s data differ from results AR has reported elsewhere. Some researchers have found that although the average IQs of men and women are the same, a greater standard deviation for men means that more of them are bunched at both high and low IQs. Prof. Jensen does not find sufficient evidence to draw this conclusion. He does confirm the standard sex differences in verbal and spatial abilities and even reports that some higher mammals show the typical male superiority in spatial ability. He also writes that in addition to their well-known advantage in verbal ability, one of the largest sex differences favoring women is in something called “speed and accuracy,” which is similar to clerical checking.
Prof. Jensen also takes up the question of why black women are so much more successful than black men. They are more likely to graduate from high school and college, pass high-level civil service tests, and enter skilled professions. This difference is not found among whites, and some researchers have wondered if black women may have a higher average IQ than black men. Once again, Prof. Jensen finds no such difference — but he offers no other explanation.
Prof. Jensen also differs from researchers who explain part of the black/white crime rate difference in terms of high black testosterone levels and an inability to defer gratification. He argues that population differences in g alone explain differences in crime rates. He notes that criminals of all races have IQs that are some 10 points below those of their siblings, and finds that within the same ranges of IQ, blacks and whites have essentially the same crime rates.
More than Generous
Needless to say, Prof. Jensen has spent his career disagreeing with others, and from time to time in The g Factor he must explain why his critics are wrong — and he is always a gentleman. Even with those who have disagreed with him in strong terms, he is more than generous in pointing out the parts of their theories that may be correct, and couches his own criticism in the gentlest terms. He treats his wildest, least scientific critics to nothing more than dignified silence: The names of Leon Kamin and Stephen Jay Gould do not even appear in an otherwise exhaustively researched and footnoted work.
The g Factor is not an easy book to read. Prof. Jensen writes clearly and repeats explanations when it would be unreasonable to expect perfect recall in his readers, but he writes for an informed, even specialist audience. He has already begun collaboration with a journalist on a more popular version of The g Factor. But those who are willing to invest the effort this book requires, will find that it is the monumental work of an extraordinary mind. A review can only begin to touch on its breadth and detail. This book is likely to become one of the landmark works in psychology, and it is the great good fortune of our society that a man of Prof. Jensen’s stature has made his career in this crucially important but thankless field.