Income, Intelligence, Inequality

Mankind Quarterly, Vol. 39 (Spring 1999) No. 3, 337-354 Edward M. Miller
University of New Orleans

The first question a potential reader has concerning this book is “What is the g Factor?” The g factor is what psychometricians such as Arthur Jensen use for what the rest of us call intelligence. Prof. Jensen is the world's leading expert on intelligence (over 300 papers published) and this book summarizes his life's work. This makes it a major work and one that will be used for years.

Jensen argues that the tests of mental ability reflect heavily a single factor, usually abbreviated g, and that this corresponds closely to what we call intelligence. By being very precise about what is being talked about, some of the semantic difficulties raised by those who don't want to admit that people differ in thinking power are avoided.

Unfortunately there are difficulties for the non-mathematician in even understanding what factor analysis is, or exactly what g is. Fortunately, the reader who hasn't mastered factor analysis can still understand and benefit from the book.

Here is a quick attempt to explain what g is, and what the debate about rotating factors is. The reader is probably familiar with tests such as the SAT which give verbal and mathematical scores, and sometimes other scores. The results of these are usually expressed by statements such as Jim earned 600 on the verbal part and 700 on the mathematical part. The reader would probably appreciate how one could also say that Jim had a total score of 1300, but did 50 points better on the mathematical part than the typical person with a 1300 total score. One can express the information either way. One could base admission decisions to an engineering program on achieving a certain total score with bonus points for doing especially well on mathematics, or one could base it on a weighted average of the two scores. With suitable weights, the two computational methods would select exactly the same set of applicants. Factor analysis is merely a sophisticated way for combining scores from different tests in different ways.

One of the most important facts about mental tests is that they are positively correlated. This makes it mathematically possible to condense the information on a scale into one measure, which is traditionally (among the specialists) referred to as g, for general intelligence. Factor analysis optimally combines the test scores to give a single measure. This measure captures in a single number as much as possible of the information in a battery of tests. That such a number can be calculated is a mere matter of mathematics. One of the main messages of this book is that most of the information in mental tests, and most of their practical utility, arises because the scores on such tests are correlated with the g factor. Ones level of g (i.e. intelligence) tells a lot about ability to learn and to function in a modern economy.

The reader may remember that I reviewed in this journal earlier (Miller 1996) a book titled, The g Factor. This was by intelligence researcher Christopher Brand, then a lecturer at Edinburgh University. This was briefly in the bookstores. But the publisher Wiley withdrew it, because it talked about racial differences. Subsequently, Edinburgh University fired Prof. Brand. Intelligence and racial differences in it are a controversial subject and powerful interests are fighting hard to keep the public from knowing what modern science has found out about the subject (Pearson 1997). Interesting, at the same time as the UK branch of Wiley was bringing out Brand's The g Factor, the US branch had Professor Jensen under contract for his book with the same title. Not surprisingly, after their action with regard to Brand's book, Wiley chose not to publish this book (in spite of its potential for selling well), and Professor Jensen had to choose another publisher.

For the readers of this journal, one of the first questions is how well g predicts economic variables. Intelligence is the single most useful variable for predicting job performance (p. 282), including in descending order of predictive validity skill testing, reference checks, class rank or grade-point average, experience, interview, education, and interest measures. Jensen examines the hypothesis that g accounts for most of the predictive validity of vocational tests. A military study of over 24,000 subjects training for 37 diverse jobs examined the predictive validity of the ten component tests of the Armed Forces Vocational Aptitude Battery and found a correlation of .75 between the test's g loading (roughly how important g was in determining job performance) and the test’s validity in predicting training success. A replication showed a correlation of .96 on a sample of 78,000 subjects in 150 military job training courses, leaving little doubt about the predictive validity of these tests arising primarily from their ability to measure g. The tests in this battery measure other things than g. By breaking the test's scores down into g and non-g factors, it was found that virtually all of the ability of the battery to predict success at training was due to g. "When an overall average prediction equation for all eighty-nine jobs was compared against using a unique optimal prediction equation for each job, the total loss in predictive accuracy was less than one-half of 1 percent." (p. 284).

In the same study the validity for predicting job performance was lower, but again most of the predictive validity of the tests arose from their ability to measure g, not from their ability to measure non-g factors.

In the civilian sectors, the best-studied test battery is the General Aptitude Test Battery developed by the US Employment Service. This includes paper and pencil tests for measuring verbal, numerical, spatial, and clerical aptitudes, and performance tests that measure such things as form perception, motor coordination, finger dexterity, and manual dexterity. This has been used to predict such measures of job performance as supervisor's ratings for over 500 occupations. Jensen reports a correlation of .65 between these tests's g loadings and the mean validity coefficients across 300 occupations, showing that the tests that are better measures of g also better predict of job performance.

By calculating optimal prediction equations using all the scales versus the predictions using just the g factor for 446 different occupations, Jensen shows that the average validity is raised from .27 to .36 (validity is the correlation between the test score and the criterion value), showing that one could improve on g for making predictions, but not by a vast amount. Other studies produce the same conclusion, that most of the predictive power in a battery of tests comes from the ability of the tests to measure g, rather than from their ability to measure more specific skills.

Of course, no book can be completely up to date. Murray (1998), using data on siblings (being from the same family whose members share social economic status and rearing) has shown that IQ is a very powerful predictor of education and income, with the smarter of two brothers typically emerging as the better educated, the one with the more prestigious occupation, and the higher income. Each IQ point produces an increment in annual income of $453 by the time the subjects are in their early thirties.

Racial Differences

The truly explosive part of Jensen's book arises from its discussion of racial differences. Jensen devotes two chapters to racial differences, the first one to documenting the differences, and the second to producing evidence that the differences are genetic.

The first chapter starts by summarizing the known differences in phenotypic intelligence. When IQ is scaled to a mean of 100 and a standard deviation of 15, the typical US mean for blacks is 85, with the values in different studies varying from 80 to 90. The black standard deviation is approximately 12, with the range from 11 to 14 (notice the black variability in IQ is less than the white variability).

Jensen includes a useful discussion (and a graph on p. 356) of how the percentages of blacks and white meet or fail to meet certain criteria which are of social and economic importance, assuming the white and black means differ by one standard deviation. With highly selective colleges admitting students from the segment of the population with IQ's above 115, the white-to-black differential will be about 7 to one. Gifted programs in high schools often require students to have IQs of 130, where the ratio will be about 20 to one. A traditional IQ cutoff for putting children into special programs for the educable mentally retarded is 70. Children below this level are 15.9% of the blacks, but only 2.3% of the whites, a ratio of about seven to one. The simple properties of the normal distribution combined with the differences in means, explain why the racial disparities in representation appear much larger at the extremes of the ability distribution (either very high or very low) than nearer the average. As Jensen points out, these extreme differences have led to much of the pressure for affirmative action.

Jensen argues convincingly that most of the racial difference in mental test scores is due to differences in g, but he also notes two differences in other factors. Whites show superiority on a spatial visualization factor, and blacks superiority on a short-term memory factor (both holding g constant).

Other fascinating aspects of the racial differences are discussed. The IQ difference between blacks and whites is greatest in the Southeast and diminishes as one moves north and west from that region of the US.

The racial IQ difference is not limited to the United States, but has been found in all countries where tests have been conducted. The biggest differences found in sub-Saharan Africa where the average from 11 studies is 1.75 standard deviations. In one large sub-Saharan study using Raven’s Matrices (a non-verbal highly g loaded test), the difference was two standard deviations (equated for schooling) for those with no apparent non-African ancestry, and l.1 for Africans of mixed ancestry.

British studies show about the same racial difference as studies in the United States. One interesting study showed that on arrival West Indian students (typically blacks from the Caribbean) and East Indian students (from the Indian subcontinent) were equally far below the native white population, but the East Indians caught up with the natives by graduation.

An interesting chart (p. 358) shows how the racial difference varies with socio-economic class (of parents). There is a clear tendency for the black-white difference to increase with the socioeconomic class. There is very little racial difference when the lowest tenths in both races are compared. Interesting, the average black IQ for students is below 95 for all socio-economic classes. Black infants score higher than whites on developmental scales that depend mainly on sensormotor abilities, but these do not correlate with school age IQ differences. The racial difference appears at three to five years of age. It is about .7 standard deviations by five to six years of age and approaches one standard deviation during the school years. After puberty it rises to about 1.2 standard deviations, where it stabilizes. This increase with age is important because the adult difference is a little larger than the one standard deviation difference usually quoted. Jensen notes the latest Stanford Binet norms show a black-white difference that is almost five IQ points less in pre-pubescent children than in post-pubescent children.

A frequent charge is that mental tests are biased against blacks. Actually, the only substantive evidence for this is that blacks score much lower on such tests. If one starts from the prior belief that there are no racial differences in ability, it must follow that the tests are biased. However, any evidence that tests are biased against a group must be based on something other than the group difference in performance on the same test.

One of the strongest pieces of evidence that tests are not biased is that they have similar predictive validity for job and scholastic performance in both the white and black populations. When the tests do differ in predictive accuracy, the black performance is usually overstated, not understated.

Various technical tests for bias have been conducted. One experiment had a panel of experts classify tests into two groups by how large their potential for bias was. Interestingly, the white-black difference was greatest on those considered the less vulnerable to bias. Also, in studies the relative difficulties of the questions are in the same rank order. This would not be true if the questions were measuring different things in the two races.

The races also show similarity in the wrong choices (called distracters) chosen in multiple choice exams if the subjects are of the same mental ages. "For example, the age ten whites and age twelve blacks are much more alike in distracter choices than are white and black children of the same chronological age" (p. 366). Similar effects are found on non-multiple choice tests. It appears that children of both races go through the same sequence of mental developmental stages, but the blacks lag the whites in passing through these stages. Their lag in attaining these developmental milestones is equivalent to an average cognitive development rate (beyond age two) of about 80% of that for white children. Thus, on average, a black-six year old, for example, performs about like that of a five-year-old on these developmental tasks" (p. 366).

Jensen devote considerable space to discussing the fact that the black-white test performance difference is far from constant, and depends on the test. Any good theory of racial differences must explain this variability. Inability to explain it destroys many otherwise plausible sounding theories.

A simple example of this is that the white-black difference on backward digit span (reciting number back from memory in the order they were read) is about twice as large as the racial difference in forward digit span (the same tests except that the numbers must be recited back in the opposite order). For instance if the racial difference was due to blacks being less motivated to do well on the tests, to low self esteem, to social economic differences, etc. one would expect the cause to affect both types of tests equally strongly, a prediction that is false. However, the g loading of the backward digits tests is much larger than that for the forward digits span, and the hypothesis that the races differ in g does predict what is observed, namely that the black disadvantage is greater on the backward digit span tests than it is one the forward digit span tests.

Jensen tests Spearman's hypothesis that the black-white difference is primarily in different tests’g component. He tests this by showing that the g loadings of different tests have a statistically significant positive correlation with the white-black score differences (in standard deviation form). This makes it much more likely that whatever determines the racial difference in test performance is a real difference in g, not merely an artifact. If there is some biological effect that causes black mental ability to develop at about 80% of the rate of whites on g, one would expect that the mental tests that showed the largest black-white performance gap would be precisely those on which performance was most sensitive to g.

Special interest attaches to the white-black gap in performance on elementary cognitive tasks. These are simple tasks which involve turning out lights when they come on, or telling which of three lights is further apart than the remaining two. These reaction time tests are very simple and virtually everyone can perform them with no errors. They seem the very opposite of the complex tasks that we normally think of as intelligence. However, performance on these tasks is correlated with intelligence (faster reaction times go along with higher intelligence) and the variability of the reaction times correlates even higher with g (smarter individuals have a lower standard deviation of reaction times). On very simple tasks, quick minds are indeed more intelligent ones. This observation suggests a biological basis for intelligence.

Because of their extreme simplicity, these reaction time tests appear unlikely to be influenced by culture, income, self esteem, or similar variables. Thus, the correlation of intelligence with such simple measurements is strong evidence for a biological basis for differences in intelligence.

Interestingly, blacks show slower reaction times and greater intra-individual variability, traits that are associated with lower intelligence. Even more interestingly, Jensen shows that Spearman's hypothesis applies even to such elementary cognitive tasks. The vector of g loadings for the tests correlates with the size of the black-white differences on the tests. This is what would be predicted if the blacks were really lower on biological intelligence, and g was primarily determined by biological intelligence. A biological theory also predicts (correctly) that the more sensitive the tests are to g, the greater the racial differences. Notice that many theories of the causes of black- white differences on mental test performance do not predict that the racial differences will vary with the tests, or that this particular pattern of differences will be found.

Genetic Differences in g

For the viewpoint of anthropology, the most controversial part of the book is the evidence it presents that the racial differences in g are genetic, rather than the result of other causes. Of course, it is virtually an article of faith that there are no genetic differences in behavior among the peoples of the world.

Jensen starts by explaining the political goals that he thinks that those who are denying the existence of races are trying to accomplish. He then presents the empirical evidence from modern genetics that the populations of the world can be grouped into a small number of groups, which correspond closely to the races traditionally identified. He draws heavily on Cavalli-Sforza, Menozzi, & Piazza (1994), whose conclusions had been summarized in this journal (Miller 1994), to show that this can be done solely on the basis of genetic data. At one point Jensen does his own analysis of their data to show how the traditional major races emerge from a purely mechanical analysis of the data.

Jensen also makes the point that all of the known polymorphism's differ in frequency among the world’s populations. It would be very odd if the frequency of the genes known to affect intelligence did not also differ in frequency among the world’s populations. A brief survey of human evolutionary history makes the point that human populations have been isolated for long periods of time, and that small differences in the strength of selection for g could easily have resulted in appreciable differences between the various population as of today. Jensen notes (p. 433) that "It is improbable that the evolution of racial differences since the advent of Homo sapiens excluded allelic changes only in those 50,000 genes that are involved with the brain."

Jensen notes that brain size has increased in the history of human evolution about three-fold, and that in many way a large brain is disadvantageous, consuming a large amount of energy (about 20% of the body's resting energy consumption is by the brain) and creating difficulties in birth. The only plausible reason for the evolution of a larger brain is that is that it facilitates thinking better about complex tasks. Jensen notes that after leaving Africa, the various human foraging populations could have been exposed to varying selection pressures depending on the climate. He notes that foraging is possible year round in tropical climates, while in colder parts of Eurasia food abundance varied with the season. He speculates that the movement into colder areas required the development of more sophisticated techniques for hunting large game. It also required foresight for planning ahead for the preservation, storage, and rationing of food in order to survive the severe winter months when foraging was practically impossible.

This leads to a discussion of the Beals et al. documentation that brain size varies in a statistically significant way with distance from the equator, especially in the Old World (the more recent arrival of humans in the New World results in a much weaker gradation of head size with latitude).

When attention is directed to races rather than to populations, similar differences in brain and cranial size are found, with the largest brains being found in East Asian populations (Europeans are a little smaller) and the smallest of the major races being found among sub-Saharan Africans. Earlier Jensen had presented the evidence from correlation studies and from magnetic imaging studies that g correlated with brain size. Jensen reports on his retabulation of data from the National Collaborative Perinatal study. This revealed racial differences in IQ at four and seven, and racial differences in head circumference, and that these were correlated. Also, among siblings, head size was correlated with IQ. Jensen calls such a correlation intrinsic, implying that a common factor, most likely larger brains, produces both the larger head circumference and the higher values for g.

Most importantly, Jensen examined head circumference with g statistically controlled, and found that the racial difference disappeared. When head size was controlled for, the racial difference in IQ was reduced, but was still present. This is what would be expected if a genetically determined difference in brain size was part of the explanation for the racial difference in intelligence, but only part of the explanation. He presents calculations suggesting that the racial difference in brain size could account for 6 IQ points of the average 16 point different in IQ.

Jensen demonstrates that the extent to which a test is g loaded correlates significantly with the extent to which that test's score is correlated with brain size. He also shows that the column vector of test by head-size correlations and a vector of standardized mean white-minus-black differences on each of the tests correlate .51. This result is not predicted by most theories that try to explain racial differences in ability.

Jensen suggests (p. 444) "the conventional null hypothesis of inferential statistics (i.e. no difference between populations) is so improbable in light of evolutionary knowledge as to be scientifically inappropriate for the study of population differences in any traits that show individual differences. The real question is not whether population differences exist for a given polymorphism, but rather the direction and magnitude of the difference."

He points out (p. 445) that it is rare in nature for genotypes and phenotypes of adaptive traits to be negatively correlated. When individuals are aggregated into populations, one would also expect that the observed phenotypic differences to correlate with genetic differences.

Jensen proposes that a reasonable default hypothesis is that the differences between groups reflect the same causes as the differences between individuals within the groups.

Jensen provides a long discussion of statistical methods for examining this default hypothesis, and reports on the new statistical methods that Rowe and Cleveland (1996) use to show that the data fits the default model very well (with a goodness-of-fit index of .98). A good fit in such studies would not be obtained if the differences between the races were due entirely to an environmental cause that did not operate within the races.

A well known phenomenon in genetics is regression to the mean. For instance, first degree relatives (such as children and parents) of persons who exhibit extreme values of a genetically influenced variable, such as intelligence, are closer to the population mean than those persons are. Those parents exhibit their extreme values because they just happen to inherit the genes that made for them, or were affected by one-time environmental effects. In the next generation, after the genes are reshuffled, the children are less likely to have inherited the genes that make for extreme values or to have been affected by these environmental factors, and hence have IQs closer to the population mean. When the populations are genetically different, the means that the children are regressing to are different. This explains the otherwise odd fact that the black-white IQ difference is smallest in the offspring of the lowest socio-economic class parents and increases steadily with the socio-economic class of the parents. This is predicted by genetic theory if the populations are genetically different, but is difficult to explain in theories in which racial differences are purely environmental.

Siblings are especially interesting to study in the context of regression to the mean because they are exposed to the same environments. Jensen (p. 471) describes the results of a study he did that involved 14 schools with over 900 white sibling pairs and over 500 black sibling pairs. "In this school district , blacks and whites who were perfectly matched for a true-score IQ of 120 had siblings whose average IQ was 113 for whites and 99 for blacks. In about 33 percent of the white sibling pairs both siblings had an IQ of 120 or above, as compared with only about 12 percent of black siblings." This is a predictable result of regression to the mean when the frequency of intelligence-enhancing genes differs in the two populations, and is predicted by a genetic theory of racial differences. There is no environmental theory that predicts such differences in the siblings of highly intelligent children.

Furthermore, Jensen points out the predictions of the genetic theory of regression are borne out over the full range of the study’s IQ. There is a linear regression of one sibling’s IQ on the other sibling’s IQ. However, the regression lines differ by race. Environmental models do not predict this result. This study involved sixteen different tests. The prediction from genetic theory was that the vectors of sibling correlations for the two races would themselves be correlated was borne out. More importantly, the white-black differences on these tests were positively correlated with the sibling correlations (sibling correlations are higher the more important genetic effects are). This is what one would predict if the racial differences were genetic; the larger the genetic influence on the text, the larger the racial difference. "A purely environmental hypothesis of the mean W-B differences would predict a negative correlation between the magnitudes of the sibling correlations and the magnitudes of the mean W-B differences. The results in fact showing a strong positive correlation contradict this purely nongenetic hypothesis" (p. 472).

The most powerful evidence that the racial differences are at least partially genetic comes from the adoption of black infants into middle class white families. An environmental theory would predict that the IQ of the black children would be the same as that of the white biological children of the same parents. However, while the white black gap was found to be reduced in childhood, by age 17 the white black gap with these middle class families was as large as the gap in the general population between blacks and whites. This is a result predicted from the theory that there is a genetic difference, but not from theories of an environmental difference. It is a discouraging finding for those who believe in the potential of the typical environmental intervention because adoption changed most of the environmental effects (neighborhood, school, family, church, peers) from that typical of blacks to that typical of middle class whites. Yet the effects on intelligence at age 17 (when adult levels have been achieved) proved minor. Most policy proposals concentrate on modifying just one of these variables, and would be expected to have smaller effects than result from placing the black infants in a white environment.

Sex differences in g

Jensen devotes one chapter to sexual differences in g. While there is a large literature on sexual differences in mental tests, most of it suffers from either using total score on IQ tests which have been carefully designed to show no sexual differences (most IQ tests), or suffers from examining particular abilities or using populations that are drawn from only part of the ability range.

After an analysis of the most representative of the general population samples available, Prof. Jensen concludes the sexes do not differ in g. However, he, like other workers, finds there is a sex difference on a spatial visualization factor (in favor of males) and on verbal fluency (in favor of females). He does note that males show more variability on most tests, especially those of spatial and quantitative abilities. Since many studies have focused on above-average individuals (university students, college applicants, Air Force officers) studies often show a larger male advantage than would be found if the whole population had been studied.

It is well established that female brains are smaller than male brains, and also that g correlates with brain size. Thus it is surprising that there is no sex difference in g. Jensen attempts to resolve this by noting that "the sex difference in brain size may be best explained in terms of the greater `packing density' of neurons in the female brain, a sexual dimorphism that allows the same number of neurons in the male and female brains despite their differences in gross size." (p. 541). The major problem with this theory is evolutionary; if one can obtain the same performance by packing the neurons closer together, this would presumably save energy and reduce birth difficulties. One naturally asks why such a superior design was adapted for female brains, but not for male brains. Another possibility is that the extra brain matter in males is used for some function males excel at, such as spatial visualization. To me this is far more plausible.

Although Jensen is widely accused of ignoring environmental effects, the book does include a fascinating account of various known environmental effects. The one that may be of the greatest practical importance is breast feeding. There is now considerable evidence that breast-fed children have higher intelligence. For a long time it was impossible to be certain this was not merely because the more intelligent mothers (whose children received good genetic and environmental backgrounds) were those who could breast feed, or chose to do so. However, research by Lucas et al. (1992) with premature babies fed through a tube and the type of milk randomly selected has shown that those given human milk do indeed have a higher intelligence when tested as children. The effect was an amazing 10 IQ points.

The g Nexus

To a general social scientist, one of the most interesting chapters is the one titled “The g Nexus”. This deals with the large number of economic and sociological variables that are correlated with g. While recognizing that correlation is not causation, there are good reasons for believing that causation runs from g to the other variables rather than from them to g. One reason is the large amount of the variance that is known (from twin and adoption studies) to be genetic in nature. This variance is not due to socioeconomic variables, while it is very plausible that the causation runs from g to the socioeconomic variables. Also, the large variability between siblings (who typically share the same socioeconomic and other aspects of family background) in g is inconsistent with causation running from socioeconomic variables to g. About one half of the total population variance in adult IQ exists in full siblings (who share the same family background). Studies that compare siblings differing in g typically show similar correlations to those that compare individuals from different families. Yet the IQs of full siblings (measured when they are children or adolescents) are positively correlated (+.30 to +.40) with measures of their educational, occupational, and economic status as adults. Causation is unlikely to run from adult status to childhood IQ, while an IQ-to-adult status link is quite plausible.

One of the interesting discussions is on the problem of education. It is pointed out that once g is controlled for, race and ethnicity make virtually no contribution to predicting scholastic performance in reading and mathematics. In an equation for predicting academic performance that includes a large number of socioeconomic variables, if g is left out, race still has considerable explanatory power. However, if g is included in the equation, race and ethnicity no longer are statistically significant. This suggests that race-specific effects of discrimination either have no effect, or exert their effects through effects on g.

A path analysis (p. 560) using college students showed that the effects of family variables on academic achievement appeared to be exerted through the effects on g.

Reading comprehension is the academic skill with the most economic importance. The significance of g for reading comprehension increases with age. Learning to decode (i.e. know which words go with which symbols) is heavily a function of memory and is usually mastered in the early grades. After the fourth grade (and setting aside those suffering from dyslexia) the correlation of reading comprehension with scores on intelligence tests is high, almost as high as the correlation of one test with another. Defects in reading comprehension in adults and older children seem to be essentially defects in g. Interestingly, there is a correlation of -.71 of reading comprehension in ninth graders with the standard deviation of reaction times (a simple measure of neuronal functioning). This suggests that reading comprehension ability is determined by the same basic feature of brain structure as determines reaction times and g. This is rather discouraging for those educators who believe that quality of teaching is the key variable in determining reading ability, but that is what the data shows.

Attention is drawn to the research of Gottfredson (1986) in which she calculates the ratio of blacks to whites with the IQ’s required for different professions. The blacks are at a disadvantage that is greatest for the high-IQ occupations such as physician or engineer and least (but still present) for occupations with lower IQ standards such as truck drivers or meat cutter. This arises from the fact that the mean IQ’s differ between the races, and that the difference becomes even greater for extreme values (a statistical fact that appears little-understood in popular discussions). Except for truck drivers, blacks appear under-represented in all these occupations. However, when the observed proportions are compared with the ratios that would be expected from the IQ data, blacks appear to be over-represented in all of these professions. For instance, given the IQ range physicians have traditionally been recruited from, the blacks would have only 5% of the representation that whites would have. But in the 1980 census blacks had 30% of the number of physicians as would be expected from their population. This under-representation, which is the cause of much hand wringing in the US, turns out to be an over-representation when IQ is controlled for. This data makes it very likely that blacks in the professions and skilled trades have recently been recruited from a lower IQ range than whites in the same profession. Since IQ contributes to job knowledge and job performance, a lower job performance should be expected from the blacks recruited under these lower standards.

In discussing the g nexus, Herrnstein & Murray’s (1994) correlations of IQ with other variables in social science (such as educational success, poverty, criminal behavior, unwed motherhood, welfare dependency, and even low birth weight) are used to show the power of including g in social science research. A fascinating table (p. 570) shows how the prevalence of various problems differs between whites and blacks, after controlling for intelligence. In most, but not all cases, the racial differences are greatly reduced. For instance, in spite of the efforts to interpret racial differences in income as being due to discrimination, after adjusting for IQ, the black with the average IQ for whites averaged $25,001 versus $25,546, not a major difference. For a fuller discussion of the Herrnstein & Murray (1994) research, see an earlier article in this journal (Miller, 1995).

Of course, there are a lot of other important variables besides g that impact on human success, and the book closes with a brief discussion of some of these (brief, because this is a book about g, not about everything that affects the human condition). One interesting concept is that of typical intellectual engagement, a personality trait which establishes the extent to which individuals engage in intellectual pursuits, and which correlates with the personality traits of conscientiousness and openness to experience. Some people engage in intellectually challenging activities more often than those with the same level of native ability. Not surprisingly, these people end up with more general knowledge, and a higher level of what the psychometricians call Gc, or crystallized intelligence.

The personality factor that plays the largest role in success across a wide range of occupations is conscientiousness.

For a high level of creativity and success in intellectual activity g appears to be necessary but not sufficient. An individual has to have a high enough g to deal with the complexity in his field, but the high- achieving individuals show a high degree of zeal and enthusiasm for their work, without which mere ability leads to little.

Conclusion

References

Cavalli-Sforza, L. L., Menozzi, P., & Piazza, A. 1994 The History and Geography of Human Genes. Princeton: Princeton University Press.

Gottfredson, L. S. 1986 The g Factor in employment. Special issue of the Journal of Vocational Behavior, 29, 293-450.

Herrnstein, R. J. & Murray, C. 1994 The Bell Curve: Intelligence and Class Structure in American Life. New York: The Free Press.

Jensen, A. R. 1973 Educability and group differences. London: Methuen.

Lucas, A., Morley, R., Cole, T. J., Lister, G. & Leeson-Payne, C. Breast milk and subsequent intelligence quotient in children born preterm. Lance, 339, 261-264.

Miller, Edward M. 1994 Tracing the Genetic History of Modern Man. Mankind Quarterly, Vol. 35 (Winter) No. 1-2, 71-108.

Miller, Edward M. 1995 Race, Socioeconomic Variables, and Intelligence: A Review and Extension of The Bell Curve, Mankind Quarterly, Vol. XXXV, (Spring), No. 3, 267-291.

Miller, Edward M. 1996 The g Factor: The Book and the Controversy. Journal of Social, Political, and Economic Studies, Vol. 21, (Summer) No. 2, 221-232.

Murray, C. 1998 Income, Inequality, & IQ. Washington, DC American Enterprise Institute for Policy Research.

Pearson, R. 1997 Race, Intelligence and Bias in Academe. Washington: Scott- Townsend.

Relethford, J. H. & Harpending, H. C. 1994 Craniometric variation, genetic theory, and modern human origins. American Journal of Physical Anthropology, 95, 249-270.

Rowe, D. C. 1994 The Limits of Family Influence: Genes Experience and Behavior. (New York: Guilford: 1994).

Rushton, J. P., 1994 Race, Evolution and Behavior: A Life History Perspective. New Brunswick: Transaction Publishers.