Arthur R. Jensen (1999) Evoked Brain Potentials and g. Psycoloquy: 10(084) Intelligence g Factor (21)

Volume: 10 (next, prev) Issue: 084 (next, prev) Article: 21 (next prev first) Alternate versions: ASCII Summary
PSYCOLOQUY (ISSN 1055-0143) is sponsored by the American Psychological Association (APA).
Psycoloquy 10(084): Evoked Brain Potentials and g

Reply to Verleger on Jensen on Intelligence-g-Factor

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


IQ does indeed have a number of physical correlates, established in a great many studies of ERP (event-related brain potentials) and other brain measurements derived from PET scans and MRI. But IQ test scores embrace the variance of a number of independent ability factors. The hypothesis I tried to address with the method of correlated vectors was whether the g factor per se was related to individual differences in the ERP. In the two studies in the ERP literature that permitted application of the correlated vectors method (because they reported the correlations of the ERP with each of the various subtests on which the Full Scale IQ was based), I found that the ERPs were remarkably related to the g factor per se, and in one study the IQ/ERP correlation vanished entirely when g was statistically partialled out of the various IQ subtests.


behavior genetics, cognitive modelling, evoked potentials, evolutionary psychology, factor analysis, g factor, heritability, individual differences, intelligence, IQ, neurometrics, psychometrics, psychophyiology, skills, Spearman, statistics
1. Verleger's (1999) evident expertise in EEG research and his exceptionally detailed and fine-grained comments -- (and the recent journal references) on just one rather narrow topic in my book (Jensen, 1998, 1999), viz., the relationship of evoked potential to psychometric g -- are to be appreciated. Since my book was about g rather than IQ or other derived test scores in general, it was not my aim to provide an extensive review of the ERP literature on this topic. I cited two fairly recent and most extensive reviews of this large body of literature (Eysenck & Barrett, 1985; Deary & Caryl, 1993). And a most frustrating literature it is! But despite the seeming chaos apparent in this area of research, due in part to the ERP's liability to numerous sources of 'noise' and procedural artifacts, IQ and other mental measurements do show correlations with ERP far beyond chance, both in number of studies and in magnitudes of the correlations.

2. The main problem in this research, so far, seems to be insufficiently standardised procedures for achieving replications of promising findings by different laboratories. Hence there are many failures to replicate empirical outcomes or to zero in on those technical procedures that best yield correlations with mental test scores. Part of the problem is also on the psychometric side. Exceedingly few out of hundreds of studies have attempted to discover the psychometric locus of the correlation in terms of the various factors that compose the total score on IQ tests. Tests that are highly loaded on so-called 'fluid intelligence' (Gf), which is highly similar or even identical to Spearman's higher-order g, have fared better in ERP research than tests of 'crystallized intelligence' (Gc), which reflects recall of past-acquired information or skills rather than immediate learning or novel problem solving in the test situation itself.

3. To study ERP correlates in terms of psychometric factors, a battery of diverse tests and other psychometric conditions are needed to relate the factors in the battery to the external or non- psychometric variable of interest, in this case the ERP or, more precisely, some measure of individual differences derived from the ERP or average evoked potential (AEP). The studies provided the kinds of data which made it possible to test the hypothesis that it is specifically the g component of an IQ battery (the Wechsler Adult Intelligence Scale [WAIS] in the present examples), rather than any of the non-g variance in the battery, that is most highly related to the ERP measures. This hypothesis was strongly borne out by two measures derived from the ERP (habituation of the amplitude and complexity of the wave-form) in two independent studies. If we are looking for the physical basis of g, are such findings not of interest, and shouldn't they be replicated in other labs with sufficient frequency to establish the relationship of ERP phenomena to g beyond any doubt? The present findings, along with other physical correlates of g, indicate that studying the physical basis of g is a scientifically feasible and worthwhile endeavour. I hope that Verleger and other EEG experts (which I am not) will pursue this.

4. Unfortunately, I have probably caused Verleger to spend too much of his discussion questioning my Figure 6.1 on p. 154 in "TgF (Jensen, 1998). I cited the source of this figure (Eysenck & Barrett, 1985, Figure 2) in the text, but failed to attach it to the figure. It is described in the above reference only as a schematic representation of an ERP. I used it in my Figure 6.1 only for the purpose of illustrating such ERP terms as latency, amplitude, zero-crossings, wave complexity, and 'time-locked epoch'. The degree of wave complexity is more typical of high than of low IQ subjects (Eysenck & Barrett's Figures 3 and 4), although the origin of this figure is not mentioned in the above reference. In any case, it is not part of the data sets used in my demonstration of the correlated vectors analyses showing the high degree of relationship (correlations of .80 and .95) between the WAIS subtests' g loadings and the subtests' correlations with the ERP-derived variables (see Jensen 1999a).

5. The early pioneer studies of Ertl & Schafer (1969) from over 30 years ago have been superseded by so many methodological and procedural improvements in EEG research on IQ, etc., that they are hardly worth mentioning, except, of course, in an historical context.

6. No place can I find where I have ever written that the number of zero crossings of the ERP wave is a "primary criterion for the intelligence of a person," although I would agree it is one of numerous AEP-derived variables that show a correlation with IQ.

7. Nor have I ever said or thought that "ERPs provide a simple biological basis for the measurement of intelligence". Indeed, the complexity and technical difficulties involved in ERP measurement are anything but simple and have proved so difficult throughout the more than 30 years since researchers have been studying its relationship to IQ that I fear this line of research may be largely abandoned in the future in favour of imaging techniques such as PET, MRI and fMRI, which so far seem to have yielded more consistent and dependable results than have ERPs.

8. Yet, ERPs are physiological measurements, and the fact that they do show correlations with psychometric measures well beyond what could be expected by artifact or chance, I think, warrants their being more systematically studied in the psychometric context. The additional fact that at least two studies show ERP correlations specifically with g that are far beyond mere chance probability makes it all the more important that these results should be replicated and studied further. I believe Verleger's statements in the middle of his paragraph #11 are in clear agreement with my position on this point. But I believe that replication of the essential findings should precede speculations about differences in "cognitive stategy". This was once a popular kind of "debunking" speculation in research on the correlations found between IQ and choice reaction times and visual inspection time, but it has since been found that a person's use of some kind of cognitive strategy tends to reduce these correlations. Strategies apparently add "noise" to the measurement of individual differences in the non-psychometric measures we wish to correlate with IQ or other psychometric variables. Identifying different cognitive strategies that might affect the ERP-g correlation will be anything but easy, but if successful, could only afford further insight into the nature of g. One wonders if every physical correlate of g depends on the exercise of a single common cognitive strategy affecting all such variables, or on different strategies specific to every such variable with which g is correlated.


Deary, I. J., & Caryl, P. G. (1993). Intelligence, EEG, and evoked potentials. In P. A. Vernon (Ed.), Biological approaches to the study of human intelligence (pp. 259-315). Norwood, NJ: Ablex.

Ertl, J. P. & Schafer, E. W. P. (1969) Neural Efficiency and Human Intelligence. Nature 223: 421-423

Eysenck, H. J., & Barrett, P. (1985). Psychophysiology and the measurement of intelligence. In C. R. Reynolds & P. C. Willson (Eds.), Methodological and statistical advances in the study of individual differences (pp. 1-49). New York: Plenum Press.

Jensen, A. (1998). The g Factor: The Science of Mental Ability. Praeger

Jensen, A. (1999). Precis of: The g Factor: The Science of Mental Ability. PSYCOLOQUY 10(23) psyc.99.10.023.intelligence-g-factor.1.jensen

Jensen, A. R. (1999a). Correlated vectors, g, and the 'Jensen Effect.' Reply to Rushton on Jensen on Intelligence-g-Factor. PSYCOLOQUY 10(082) psyc.99.10.082.intelligence-g-factor.19.jensen

Verleger, R. (1999). The g factor and event-related EEG potentials. PSYCOLOQUY 10(039) psyc.99.10.039.intelligence-g-factor.2.verleger

Volume: 10 (next, prev) Issue: 084 (next, prev) Article: 21 (next prev first) Alternate versions: ASCII Summary