Arthur R. Jensen (2000) A "simplest Cases" Approach to Exploring the Neural Basis of g. Psycoloquy: 11(023) Intelligence g Factor (36)

Volume: 11 (next, prev) Issue: 023 (next, prev) Article: 36 (next prev first) Alternate versions: ASCII Summary
PSYCOLOQUY (ISSN 1055-0143) is sponsored by the American Psychological Association (APA).
Psycoloquy 11(023): A "simplest Cases" Approach to Exploring the Neural Basis of g

Reply to Ingber on Jensen on Intelligence-g-Factor

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


Ingber (1999) proposes EEG models and methods for exploring the brain physiology responsible for g. It could well be the case that the more ideal measurements of g that are based on a large battery of diverse mental tests is too broad, global, or spread through too wide an area of the brain to allow analytical study by neurological techniques. A successful use of the kinds of neuroscience techniques suggested by Ingber, and probably other techniques such as fMRI and PET, may depend upon a simplification of the behavioral side of the equation by using highly reliable measurements of individual differences in a number of test paradigms, each of which isolates a single pair of simple ability variables whose correlation is part of the g nexus.


behavior genetics, cognitive modelling, evoked potentials, evolutionary psychology, factor analysis, g factor, heritability, individual differences, intelligence, IQ, neurometrics, psychometrics, psychophyiology, skills, Spearman, statistics
1. Although I am not an expert in EEG technology and neurophysiology and therefore am not qualified to evaluate Ingber's (1999) theoretical and methodological proposals for advancing our understanding of the neural basis of g, I do appreciate his interest and expertise on the subject. They have enabled me to reconsider how an experimental psychologist might best contribute to the application of Ingber's approach to the neurological study of psychometric g. A number of points in Ingber's commentary remind me of certain aspects of my work as an experimental psychologist long before I became interested in g in its own right. Had the thoughts stimulated by Ingber's paper occurred to me at the time I wrote "The g Factor" (Jensen, 1998; 1999), I would have elaborated more on what I am about to say here. The seeds of this idea were present in the book but they had not sprouted until now. It might be a worthless idea in relation to Ingber's suggestions but my lack of expertise in his specialty area makes it impossible for me to reject the idea. It will be up to specialists in neuroscience to judge whether or not what follows here is heading up a blind alley.

2. Because g ideally represents the highest order distillate of the individual differences variance common to performance on many tasks representing many apparently different abilities (therefore implying various neural systems in different regions of the brain), it may be virtually impossible to discover the neural basis of the covariances between all the [n(n-1)]/2 pairs of n behavioral variables from which g is, so to speak, "distilled" by looking for g itself. What may have to be done in order to discover more than just correlations between psychometric g and individual differences in some neural measure, such as evoked potential variables, is to focus on the neurological aspect of certain well-chosen single pairs of simple variables in the whole correlational nexus that yields g. One would seek pairs of ability tasks on which individual differences are rather highly correlated -- tasks that theories or models borrowed from either cognitive psychology or cognitive neuroscience (or both) claim involve neural processes, pathways, regions of the brain, or different "cognitive components." The research question, then, would be: Why is there a high correlation between performances on two tasks that factually (or theoretically) involve different brain functions (although some functions may be identical for both tasks)?

3. I am reminded of a study I did some 30 years ago (Jensen, 1971), which illustrates the kind of paradigm I have in mind, although I know virtually nothing about the neurological circuits or processes involved in the performance since I was measuring only behavior. It was a study of individual differences in short-term memory (STM) in which the subjects (university undergraduates) were repeatedly tested on both auditory and visual memory span for digits. The tests (administered by a laboratory apparatus), as well as the recording of the subjects' responses and the pacing of the stimuli (digits), were identical across the auditory and visual paradigms. Through repeated testing, the reliability of the measurements was in the high .90s. The order of administering the visual and auditory tasks was counterbalanced.

4. I had expected to find that some individuals would perform better on visual than on auditory STM while others would show the opposite pattern. What I found, however, was a disattenuated correlation of unity between the two modes of presentation. After correcting for the very slight errors of measurement, subjects maintained the same rank order of ability on both tasks yet the overall performance for each of the modalities differed in their means and in their interactions with different experimental conditions (e.g., an interposed 10 seconds delay between presentation of the digit series and recall). So here we have two tasks with clearly different sensory input mechanisms and pathways which result in significant mean differences and interactions with experimental manipulations and yet show no interactions with individual differences. Consequently, they would have identical correlations with some other g loaded variable such as IQ with which, in fact, they are correlated. What is the locus of the visual/auditory STM correlation? What is the mechanism or process that makes them perfectly correlated? If we knew that, we would have accounted for one small bit of the g nexus. But that discovery might generalize to other behavioral measures.

5. Another laboratory study (Jensen, 1987b), in which individual differences could be measured with very high reliability, had similar properties. Using the Sternberg and the Neisser reaction time (RT) paradigms, the study involved scanning a digit series held in STM (Sternberg) and visually scanning a series of digits presented on the screen (Neisser) to detect either the presence or absence of a single "probe" digit. These paradigms are identical but opposite, since the single probe digit appears either immediately after (Sternberg) or before (Neisser) the 3-second presentation of the digit series (consisting of 1 to 7 digits). The subject merely presses a YES or a NO button to indicate the presence or absence of the probe digit in the presented series and the subject's RT is measured. Cognitive theorists have used these paradigms as examples of different cognitive components being involved in the two tasks. The Sternberg memory scan task puts a high premium on STM, whereas the Neisser visual scan task does not. Interestingly, the RT measures derived from the two paradigms show significant mean differences which also interact with experimental conditions, but individual differences (corrected for attenuation) are perfectly correlated across the two paradigms and both have the same correlation with IQ.

6. How can this be explained, given that these tasks supposedly involve different "cognitive components?". Of course, the Vocabulary and Block Designs subtests of the Wechsler scales must involve extremely different "cognitive components," yet they are highly correlated. But this is my point: there are too many ways that Vocabulary and Block Designs differ to allow the basis of their correlation to be easily identified. It should probably be easier to identify the locus of the common factor if the tasks differ in only one (or a very few) features.

7. Another equally simple paradigm is that of forward and backward digit span (FDS and BDS). This is surprisingly different from the two previous examples because individual differences in FDS and BDS are not very highly correlated, even when corrected for attenuation, and FDS is only about half as correlated with IQ as is BDS (Jensen & Figueroa, 1975). Everything is exactly the same in FDS and BDS tests except that in BDS the subject must recall the digits in the reverse order from that in which they were presented by the examiner. Individuals who can consistently recall, say, 8 digits in the FDS test, are found to differ reliably in the number of digits they can recall in BDS, and this FDS-BDS difference is positively related to their level of g derived from test batteries that contain no tests of STM.

8. Ingber (1999) views the Hick paradigm as a tool along the lines I have suggested. Its facets for experimental manipulation and the highly reliable measurement of g-related individual differences are advantages similar to those I have mentioned in connections with the STM paradigms (Jensen, 1987a). Other promising speed of information-processing paradigms that show correlations with broad psychometric tests are inspection time (for comprehensive references see Jensen, 1998, pp. 221-223) and -- the most recent candidate -- the modified blink reflex (Smyth, et al., 1999).

9. We might even hearken back to Francis Galton (1822-1911) himself. His supposedly discredited hypothesis that there is a general factor among various modalities of sensory discrimination and the idea that this common factor reflects general mental ability, or psychometric g, has proved to be correct after all (Acton & Schroeder, 1999; Li et al.,1998). What is the common cause of these correlations? We cannot rest until we know the answer at the level of neurophysiology, unless at this point someone can prove that the cause of these correlations does not involve the brain. Ingber sees a technical means for pursuing research on g in the direction that the Galton-Spearman paradigm is inevitably headed, and I hope he will continue to bring his expertise in neuroscience to help advance this endeavour.

    EDITOR'S NOTE: Owing to an enumeration error, the Ingber commentary
    to which professor Jensen refers here has been reassigned thread
    number 35 instead of the number 13 that it had been assigned upon
    publication in 1999.


Acton, G.S., & Schroeder, D.H. (1999). Color discrimination as related to other aptitudes: An analysis of the Farnsworth-Munsell100-huetest. Paper presented at the 9th Biennial Convention of the International Society for the Study of Individual Differences, Vancouver, B.C., Canada, July, 1999.

Ingber, L. (1999). Statistical mechanics of neocortical interactions: Reaction time correlates of the g factor. PSYCOLOQUY 10(68) psyc.99.10.068.intelligence-g-factor.35.ingber

Jensen, A.R. (1971). Individual differences in visual and auditory memory. Journal of Educational Psychology, 1962, 123-131.

Jensen, A.R. (1987a). Individual differences in the Hick paradigm. In P.A. Vernon (Ed.), Speed of information- processing and intelligence,(pp. 101-175). Norwood, NJ: Ablex.

Jensen, A.R. (1987b). Process differences and individual differences in some cognitive tasks. Intelligence, 11, 107- 136.

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

Jensen, A.R. (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., & Figueroa, R.A. (1975). Forward and backward digit span interaction with race and IQ: Predictions from Jensen's theory. Journalof Educational Psychology, 67, 882-893.

Li, S-C, Jordanova, M., & Lindenberger, U. (1998). From good senses to good sense: A link between tactile information processing and intelligence. Intelligence, 26, 99-122.Smyth, M., Anderson, M., & Hammond, G. (1999). The modified blink reflex and individual differences in speed of processing.Intelligence,27, 13-35.

Volume: 11 (next, prev) Issue: 023 (next, prev) Article: 36 (next prev first) Alternate versions: ASCII Summary