Arthur R. Jensen (2000) A Potpourri of G-related Topics. Psycoloquy: 11(087) Intelligence g Factor (48)

Volume: 11 (next, prev) Issue: 087 (next, prev) Article: 48 (next prev first) Alternate versions: ASCII Summary
PSYCOLOQUY (ISSN 1055-0143) is sponsored by the American Psychological Association (APA).
Psycoloquy 11(087): A Potpourri of G-related Topics

Reply to Kovacs & Pleh on Jensen on Intelligence-g-Factor

Arthur R. Jensen
Educational Psychology
School of Education
University of California
Berkeley, CA 94720-1670


Kovacs & Pleh (2000) offer brief comments on such a widely varied selection of issues related to "The g Factor" (Jensen, 1998, 2000) that I can only attempt to reply to those of them that seem to based on some misunderstanding of my views. Many of Kovacs & Pleh's points are not at all incompatible with anything in my book and often express sensible and widely accepted views in applied psychometrics (but I will pass silently over the few remarks that seem idealogically inspired, such as that g theory is "conservative" and "deterministic.")


behavior genetics, cognitive modelling, evoked potentials, evolutionary psychology, factor analysis, g factor, heritability, individual differences, intelligence, IQ, neurometrics, psychometrics, psychophyiology, skills, Spearman, statistics
1. Kovacs & Pleh (2000) seem to believe, contrary to anything I have ever written, that I believe in or advocate a "one-factor theory of intelligence". This erroneous view is expressed several times in their commentary. Neither do I throw out 'intelligence' and merely replace it with g. I assign distinctly different meanings to 'intelligence' and g; each term serves a quite different purpose. (I have fully explained all this in my reply to Bringsjord [Jensen, 2000].) Kovacs & Pleh follow a common misconception among some 'multiple intelligences' theorists that anyone who recognizes the existence of g must necessarily disclaim the existence of any other mental abilities.

2. In present-day psychometrics, the generally accepted factor model of the covariance structure of human mental abilities is what Carroll (1993) has called the 3-stratum model. This is a structure of abilities represented hierarchically by their degree of generality (i.e., correlation with other abilities), from the least general at stratum 1 (i.e., the first-order factors), to the next most general factors at stratum 2 (i.e., the second-order factors), to the single and most general factor that emerges at stratum 3 (the g factor). This factor hierarchy can be 'orthogonalized', which renders all of the factors uncorrelated, both within strata and between strata.

3. In a large test battery measuring many different abilities, the total variance of all the tests in the battery consists of the common factor variance attributable to all of the uncorrelated factors plus the variance that is 'unique' to each of the tests. (The unique variance consists of the true-score variance specific to each test in the battery plus each test's error variance, or unreliability.) In a large and diverse test battery, the g factor accounts for only a fraction of the total variance, but this is usually a larger proportion of the total variance than is associated with any other single factor. Also, the g factor has the psychometrically interesting and practically useful property of showing more and larger correlations with 'real-life' behavioral criteria, outside the realm of psychometrics, than does any other factor that can be extracted from any large battery of diverse tests (Gottfredson, 1997; Jensen, 1998, Chapter 9).

4. Even when g theory and factor analysis never entered into the construction of a test, as was the case for many of the most widely used IQ tests, those tests that happen to have the highest practical validity for predicting 'real-life' performance outside the testing room turn out to be among the most highly g loaded tests. The g factor also constitutes some substantial part of the variance in tests that were specifically designed to assess special aptitudes or talents, such as mathematical, literary, musical, and mechanical aptitudes. One can recognize and value these special abilities, which are crucial for the attainment of certain careers and achievements, without denying the existence and importance of g. It seems safe to say that none of the world's famous geniuses are noted for their level of g; each is acclaimed for some highly specialized talent or type of achievement. It is equally true, however, that none of them is low or even average in g. Even the leading "multiple intelligences" theorist, Gardner (1993) admits that a lower bound IQ of about 120 is required for individuals who are widely recognized for their exceptional achievements (e.g., as examples, Gardner mentions Einstein, Freud, Gandhi, Picasso, and Stravinsky, among others). But being above even this minimal threshold of IQ already excludes at least 90 percent of the general population.

5. Genetic variation, or hereditary individual differences in any character or trait is a necessary condition for evolution, providing the raw material upon which selection operates. Darwin recognized not only natural selection but also sexual selection as the mechanism of evolutionary change, and a recent book by Miller (2000) argues that sexual selection has been the more instrumental in the evolution of human mentality. But the fact that genetic variability is intrinsic to evolution does not negate the fact that it also exacts a cost for some members of a population, depending on the particular environmental demands and selection pressures that happen to be operating at the time. Though these costs are relatively relaxed in modern civilizations, they are not entirely absent with respect to many human characteristics, including mental abilities and particularly g, which has a large genetic component and conspicuous consequences in the contemporary spheres of education, occupation, and many other personally and socially valued assets in modern society. Although the negative correlations of individuals' levels of g with social variables such as delinquency, crime, welfare dependency, school dropout, drug abuse, child neglect, etc., are small and hence relatively unreliable for prediction of life outcomes for any particular individual, they are nevertheless of considerable social significance, as one can observe in many indices of the quality of life that exists in neighborhoods that differ significantly in the distribution of IQ (as a rough indicator of g).

6. Kovacs & Pleh unfortunately seem to be 20 or so years out of date in believing that personnel selection today puts more emphasis on specific abilities than on g. This happens nowadays only to the extent that educational institutions and employers feel constrained by political correctness to use selection tests with an especially strong appearance of job-specific face validity. A large body of research by Hunter and Schmidt (Schmidt, Ones, & Hunter, 1992) shows that the high 'validity generalization' of selection tests across many occupations is attributable to their g loadings and that if selection tests of all kinds were stripped of their considerable g variance, they would be practically worthless for selection. The chief 'active ingredient' accounting for the practical predictive validity of even the most job-related selection tests is psychometric g. In selecting for certain jobs, of course, one may need specific information besides an individual's IQ or g level, such as job-relevant knowledge, skills, and experience. An orchestra conductor, for instance, would not engage player without auditioning their instrument. These specialized requirements being equal, however, the higher IQ applicants will generally be preferred, and studies show that they are the best bet for proficiency on the job and for future advancement.

7. The importance of statistical significance, indicated by a statistical test's p value, is not the proper guide to the practical importance of a particular experimental finding or effect. A p value reflects sample size, which, if very large, can yield very small values of p, making them 'highly significant'. This tells us only that the observed effect is not very likely to be a fluke of chance and that we may not be entirely wasting our time by considering it. Much more informative than p concerning the potential usefulness of an observed effect is the so-called 'effect size', or d, which indicates the size of the effect relative to the standard deviation of the variable of interest in some specified population sample. In the behavioral sciences, many replications of substantial d values or correlation coefficients with respect to a given phenomenon are the best evidence we can have of their validity and potential scientific or practical importance.

8. Even among all the non-sequiturs I've come across in the rhetoric of the 'IQ controversy' the final sentence in Kovacs & Pleh's commentary is a gem.


Carroll, J.B. (1993). Human cognitive abilities: A survey of factor-analytic studies. Cambridge, UK: Cambridge University Press.

Gardner, H. (1993). Creating minds New York: Basic Books.

Gottfredson, L. S. (1997) . Why g matters: The complexity of everyday life. Intelligence, 24, 79-132.

Jensen, A.R. (1998). The g factor: The science of mental ability. Westport, CT: Praeger.

Jensen, A.R. (1999). Precis of: "The g Factor: The Science of Mental Ability" PSYCOLOQUY 10(023). psyc.99.10.023.intelligence-g-factor.1.jensen

Jensen, A.R. (2000). Artificial intelligence and g theory concern different phenomena. psyc.00.11.086.intelligence-g-factor.47.jensen

Kovacs, K. & Pleh, C. (2000). Evolution, individual differences and social commitments. PSYCOLOQUY 11(045) psyc.00.11.045.intelligence-g-factor.44.kovacs

Miller, G. F. (2000). The mating mind: How sexual choice shaped the evolution of human nature. New York: Doubleday.

Schmidt, F.L., Ones, D.S., & Hunter, J.E. (1992). Personnel selection. Annual Review of Psychology, 43, 627-670.

Volume: 11 (next, prev) Issue: 087 (next, prev) Article: 48 (next prev first) Alternate versions: ASCII Summary