Mike Anderson (2000) An Unassailable Defense of g but a Siren-song for Theories of Intelligence. Psycoloquy: 11(013) Intelligence g Factor (28)

Volume: 11 (next, prev) Issue: 013 (next, prev) Article: 28 (next prev first) Alternate versions: ASCII Summary
Topic:
Article:
PSYCOLOQUY (ISSN 1055-0143) is sponsored by the American Psychological Association (APA).
Psycoloquy 11(013): An Unassailable Defense of g but a Siren-song for Theories of Intelligence

AN UNASSAILABLE DEFENSE OF G BUT A SIREN-SONG FOR THEORIES OF INTELLIGENCE
Review of Jensen on Intelligence-g-Factor

Mike Anderson
The University of Western Australia
Nedlands
WA 6019
www.uwa.edu.au/user/mike/mike.html

mike@psy.uwa.edu.au

Abstract

Jensen presents a compelling case for the reality of g. This represents a major scientific achievement. However, I consider that there are two serious mistakes in his choice of scientific strategy. The first is to adopt an operational stance -- that is, to drop all reference to the word 'intelligence' and define our object of scientific interest as the g-factor. This is a mistake because the operational stance does not work in science and in this particular case it strips the g factor of all of its psychological significance. The second mistake is in his pursuit of the g-nexus, where biological and psychological variables are related to social outcomes via g. Without a serious theory of mental structure, this attempt rings hollow. Indeed such a strategy impedes the more crucial scientific step of developing a theory of intelligence which, in the final analysis, is a theory of mental structure.

Keywords

behavior genetics, cognitive modelling, evoked potentials, evolutionary psychology, factor analysis, g factor, heritability, individual differences, intelligence, IQ, neurometrics, psychometrics, psychophyiology, skills, Spearman, statistics
1. The major impression to take from Jensen's (1998, 1999) "The g Factor" is that it represents a work of monumental scholarship. I have nothing but admiration for the way that Jensen has stuck to his task over the last 20-30 years in establishing the psychological reality of Spearman's g. In particular, he has developed, and in my view established, the notion that g might best be thought of as grounded in differences in speed of information processing. In many ways he has established this singlehandedly and doggedly with only sporadic, although valuable, help from others. It is not without regret, then, that in all probability my review will appear negative. This is only because my particular area of expertise is in psychological theories of intelligence, and here I am afraid the book has a hollow centre. This is not due to any lack in Jensen's scholarship -- as I have said this is clearly of the highest order. Rather it is a direct consequence of the scientific road Jensen has chosen. This road involves two important choices. The first is the operational stance and the second is reductionism, or more appropriately, consilience (Wilson, 1998).

I. THE OPERATIONAL STANCE

2. For Jensen, the term 'intelligence' has become scientifically useless because it means too many different things to too many different people (Jensen, 1999, pp.45-49). Hence, Jensen argues, we should drop the term and focus on the empirically sure core of intelligence, namely, the g-factor. Jensen attempts to distance himself from the much derided definition of intelligence as what the tests measure, by focusing not on IQ per se but on the undeniable empirical fact that 'cognitive abilities' covary. Providing independent evidence for the reality of g and developing a specific hypothesis (that speed of processing underlies g) form much of the early part of the book. Undoubtedly this serves to justify the claim that the g factor is a true discovery and not some necessary artifact of intelligence test construction. However, for me, much of the good work is undone by asking us to eschew the wider notion of intelligence and by operationalising intelligence as the g-factor. It is one thing to encourage explanations of the empirical regularity that g represents and quite another to claim that such an explanation is tantamount to a theory of intelligence. There appears little virtue (scientific or otherwise) in substituting the 'covariation of abilities' for 'what the tests test' as a working definition of intelligence. No other serious science relies on operational definitions (Layzer, 1976), and there is no clear argument as to why psychology should be any different.

3. Even within the confines of this book we can see the muddles that operational definitions lead to. On pp.52-53 the distinction between a 'mental ability' and a 'physical ability' is given an operational definition -- "The ability in question is not a mental ability if its largest factor loading (in a factor analysis of a wide variety of abilities) is found on a group factor defined as physical." Yet what defines mental and physical, other than an implicit theory? And explicit theories are always to be preferred in science. On this operational definition it is not possible to have a mental ability that is more correlated with a 'physical factor' than it is with the g-factor. This rules out the possibility of a mental function (this time defined theoretically) that could be related to physical parameters of the brain, for example, some aspect of linguistic ability being considered a mental ability. And if this is the case then much that is interesting about g for psychology (e.g., why are some mental functions related to g while others are not?) disappears. What Jensen has established is that any theory of intelligence must take general intelligence seriously. But the problem of what such a theory should look like is sidestepped by restricting scientific enquiry to 'general intelligence' -- he has thrown out the baby with the bathwater.

4. With an operational stance there is always the danger of an infinite regress and charges of tautology. This is especially so under the circumstances (as here) where the operational definition (the g-factor is intelligence) is set up after the empirical conditions that are supposed to constitute the crucial test of the construct (cognitive abilities covary) have been established. Empirical regularities provide the impetus for theories but cannot themselves provide the meaning of the constructs. Rather than providing the sure footing for scientific study, the operational stance has been the Achilles heel of Jensen's work. It has allowed the claim that g is a tautological consequence of test construction to survive much longer than it ought to have done. But in my view there is a more serious and potentially fatal mistake in Jensen's choice of scientific strategy.

II. THE CONSILIENCE STRATEGY

5. Having established that g is 'real' (i.e., it is not an artifact of test construction, power elites or factor analysis), and that a good working hypothesis is that g may be based on differences in speed of information processing (as measured by reaction times, inspection times and maybe even some brain-based measures), where does Jensen then lead us? First, he moves 'down' to argue that it is this same g that is the core of the heritability of IQ. Second, he moves 'up' to show how a 'theory of g' might make sense of race differences in IQ. In so doing, he bridges genetics, g, and race differences in one superstructure, thereby establishing what he comes to call the g-nexus. The g-nexus represents the variables that are linked through g -- "education, employment, income, poverty, crime and other social pathologies" (p.555) on the one hand, and IQ, speed of processing, brain physiology and genetics on the other. Herein lies the mistake. The attempt to link biological and psychological variables with social outcomes via g is a mistake, not because such a framework is considered by some as unsavoury (I, for one, don't particularly care whether or not race differences are based on genetic differences), but because, I believe, it is the wrong scientific strategy.

6. The strategy he has chosen might be called reductionism because his first step is to attempt to establish that it is g that is the basis of the heritability of IQ. In other words g is reduced to genes. The strategy can also be considered reductionistic because the social phenomena are to be explained, substantially, by differences in g which are in turn to be explained by differences in genes. However, given that the g factor itself sits in the middle of this causal chain, I prefer the term consilience (Wilson, 1998). Consilience is the search for the unity of knowledge involving, in theory at least, the linking of many different levels of explanation from quantum mechanics to social behaviour in an integrated causal chain. Wilson argues that consilience is the ultimate goal of science. I think the consilience strategy has been the wrong choice for the science of intelligence -- mainly because for consilience to work we need an adequate theory at the psychological level, and we are a long way from having that. Jensen's contribution is to establish the g-factor at the core of theories of intelligence. However, both his operational stance (i.e., the g-factor is intelligence) and his search for consilience have taken him away from developing the theory itself. In my view this has led to a number of unfortunate consequences:

    (i) Scientists in the field, taking the lead from Jensen, have
    concentrated too much on searching for physiological correlates of
    IQ.  In my view much of this effort will be wasted in the absence
    of a proper theory of mental structure with which to interpret
    these data.

    (ii) Even more problematic has been the encouragement of work that
    explores the g-nexus. While it is an interesting hypothesis for
    sociology and education that 'general intelligence' may mediate
    biology and social phenomena, the hypothesis is not particularly
    psychological.  With no flesh on its bones, that is to say yet
    again with no theory of mental structure, it can only ever be
    correlational in the same way that the relationship between say
    unemployment rates and mental illness is only of marginal
    psychological interest.

    (iii) The consilience strategy takes us into the murky and
    dangerous worlds of race and genes, and this has turned away a
    generation of potential researchers into the psychology of
    intelligence. Frankly, the g factor appears psychologically barren
    and not worth the heat it has generated. The g-nexus is important,
    I think, if you work in sociology or education. However, to attempt
    nothing more than the identification of variables that might
    explain group differences in social outcome is just scientifically
    uninteresting for theories of mental structure.  There is no more
    reason to investigate the association between groups divided on
    race than there is to investigate groups that differ in height.
    Either analysis is uninformative for a theory of mental structure
    and ultimately that is what a theory of intelligence should be.

7. The consilience strategy has taken priority over establishing or formulating a psychological theory of 'intelligence'. While having some sympathy with Jensen's exasperation with the foot-loose and fancy-free way that the word intelligence is bandied about, the point is that for psychology it is intelligence (and not the g factor) that is ultimately interesting. Surely the next step after establishing the reality of g (and I must stress that it is not a step we could reasonably expect Jensen himself to take -- there is only so much that can be done in a scientific lifetime) should be to show how it fits into a wider intellectual architecture or structure. In my view, it is not much of a psychological theory to point out that g is at the apex of a hierarchy of psychometrically defined abilities (Vernon pretty much said so 50 years ago, anyway). Examples of psychological questions are: What is the nature of non-g mental abilities? What is the mechanism through which speed of processing generates a g-factor? How do mechanisms related to g and more specific abilities change during cognitive development? Why do differences in speed of processing influence some mental processes and not others? How can a savant with low IQ master complex cognitive feats? There is none of this in Jensen's book. Indeed, such questions have been dismissed with a wave of the hand as early as pages 128-133 as part of the strategy of advocating the dreaded operational stance. And yet in my view such questions are at the core of a psychological theory of intelligence. Understanding mental architecture is the linch-pin for understanding many things that a science of intelligence should be able to inform.

8. At this point I do have to declare an interest. After acknowledging the reality of g, and owing mainly to Jensen's own work, I have adopted the opposite scientific strategy, and attempted to build a theory of intelligence that addresses psychological issues but has the g factor at its core (Anderson, 1986, 1992). I would have dearly loved to have someone of Jensen's formidable intellectual talent and work ethic challenge my theoretical conception of intelligence. Instead, I find that most people working in intelligence research are largely uninterested in theoretical development, and prefer to pursue the consilience strategy (with some notable exceptions, for example, Detterman, 1987, Nettelbeck, 1999). Inevitably this will sound like sour grapes on my part, but if it does I will have to endure that embarrassment. For it is not the absence of any reference to my theory per se but the absence of any concern to engage in a discussion about psychological theory that I find disappointing and ultimately disturbing.

9. The consequences of ignoring the development of theory are profound. Jensen pursues the strategy of reporting yet more correlations that are consistent with the 'g-is-speed' hypothesis. Often this involves crossing measurement, descriptive and explanatory levels in a haphazard and opportunistic fashion -- for example, correlating intelligence tests, school performance, reaction times, inspection times, evoked potentials and nerve conduction rates. Much of these data I find unconvincing because it is not clear what speed of processing means and consequently whether all these different techniques are measuring it (ironically, this lack of specificity reminds me of Gardner's strategy to establish the validity of multiple intelligences). But in the absence of a theory, where are all these data supposed to lead us beyond the already established conclusion that g is real and biological in origin?

10. It seems to me that 'The g-Factor' is best regarded as the rampart that others who would deny the reality of general intelligence as a biological constraint on cognitive abilities must assail. And as such I can imagine no more daunting stronghold. However, as a strategy for the scientific understanding of human intelligence it threatens to lead us into a wilderness from which there will be no return.

REFERENCES

Anderson, M. (1992). Intelligence and development: A cognitive theory. Oxford: Blackwell.

Anderson, M. (1986). Understanding the cognitive deficit in mental retardation. Journal of Child Psychology and Psychiatry, 27, 297-306.

Detterman, D.K. (1987). Theoretical notions of intelligence and mental retardation. American Journal of Mental Deficiency, 92, 2-11.

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). ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1999.volume.10/ psyc.99.10.023.intelligence-g-factor.1.jensen http://www.cogsci.soton.ac.uk/cgi/psyc/newpsy?10.023

Layzer, D. (1976). Science or superstition? A physical scientist looks at the IQ controversy. In N. Block & G. Dworkin (Eds) The IQ controversy. New York: Pantheon.

Nettelbeck, T. (1999). Savant syndrome: Rhyme without reason. In M. Anderson (Ed), The development of intelligence. Hove: Psychology Press.

Wilson, E.O. (1998). Consilience: The unity of knowledge. London: Little, Brown & Co.


Volume: 11 (next, prev) Issue: 013 (next, prev) Article: 28 (next prev first) Alternate versions: ASCII Summary
Topic:
Article: