February 23, 2009



That dreaded IQ again

I am continually amazed at people's capacity to fool themselves. And the smarter you are, the better you are at fooling yourself, it seems. Perhaps the most persistent fallacious belief is that there is such a thing as "healthy" food. Many people, particularly well-educated people, desperately want to believe that they can extend their lifespan by controlling what goes into their mouths. And what is the usual definition of "healthy" food? Low-fat food. And you cannot pick up a newspaper or magazine without seeing some comment about healthy food. Yet there is, as far as I am aware, no such thing. As you can see in the sidebar of my FOOD & HEALTH SKEPTIC blog, the biggest, best and most conclusive research study of the subject found that a low-fat diet had no effect on your health whatever! But that finding was like water off a duck's back. Although the U.S. government spent $400 million on making the study the very best piece of research that could be conducted on the subject, I have yet to see a single mention of it in the popular press and middle-class people still fuss about their food as if it will do them some long-term good. It is a form of religion I guess. People just NEED to feel that they have some control over their fate.

And people's understanding of IQ also seems to be needs-driven. Just about the worst thing you can be called in the USA today is a "racist", yet IQ tests clearly show that blacks are, on average, outstandingly dumb. And their high level of dysfunctional behaviour and low level of educational and economic success is exactly what you would expect of dumb people. It is clear that the tests do not lie. Prisons are full of dumb people and huge numbers of them are black.

But what I have just said is for most people very threatening. If they accept its obvious truth, they risk being seen both by themselves and others as "racist", and that will not do at all. So there are all sorts of mental wriggles that people use to avoid accepting what I have just said. And the truth of all those wriggles has of course been extensively tested and found to be lacking. But there are whole books written on that subject (start here if you want to look into it) so I will not go into details now about it.

What started me off on that little meditation was something very simple, the strapline (subheading) given for the article excerpted below. In answer to the question "Why men are more intelligent than women", the strapline gives an answer: "They aren't". Yet that strapline does not represent the findings reported in the article at all. The article is a generally accurate summary of our current knowledge about IQ and says that women DO on average have slightly lower IQs. It then makes a start on explaining why.

So why such an inaccurate strapline? Because the article was published in a semi-popular magazine and there would have been great cries of outrage from both feminists and real women otherwise. The fact that averages tell us nothing about any one individual within a given population would be no consolation. Many women would just not WANT to believe the conclusion so they had to be assuaged. So the conclusions were slanted to deflect attention from the actual averages to theoretical averages.

Why men are more intelligent than women

The answer is: They aren't.

By Satoshi Kanazawa

The orthodoxy in intelligence research for the second half of the 20th century had been that men and women had the same average intelligence, but men had greater variance in their distribution than women. Most geniuses were men, and most imbeciles were men, they said, while most women were in the normal range. This conclusion, however, was manufactured out of political expediency. Not wanting to discover, or a priori denying, any sex differences in intelligence, psychometricians simply deleted from the standardized IQ tests any item on which the performance of men and women differed.

More recently, however, especially since the turn of the millennium, there have been an increasing number of studies that cast doubt on this politically correct conclusion. Studies with large representative national samples from Spain, Denmark, and the United States, as well as meta-analyses of a large number of published studies throughout the world, all conclude that men on average are slightly but significantly more intelligent than women, by about 3-5 IQ points. So this has now become the new (albeit tentative) consensus in intelligence research.

However, these studies do not answer the ultimate evolutionary question of why men should be more intelligent than women. General intelligence likely evolved as a domain-specific psychological mechanism to deal with evolutionary novelty. However, unlike populations in different geographic parts of the world [translation: different races], men and women within a population have always faced the same level of evolutionary novelty throughout evolutionary history, because they have always migrated together. If general intelligence is a function of the evolutionary novelty of the environment, why then are men on average slightly more intelligent than women? My LSE colleague, Diane J. Reyniers, and I offer one possible explanation in our article, forthcoming in the American Journal of Psychology.

Psychometricians have known since the end of the 19th century that height is positively correlated with intelligence: Taller people on average are more intelligent than shorter people. And men in every human population are taller than women. So one possibility is that men are more intelligent than women, not because they are men, but because they are taller.....

More here

I am not sure why Kanazawa claims that items differentiating men and women have been excluded from IQ tests in the past. It is demonstrably and grossly untrue. He must have in mind some particular instance that is well outside normal practice. It is demonstrably untrue because many items in standard IQ tests DO discriminate men and women. For instance, women tend to do better on verbal puzzles and men do better on mathematical puzzles.

Normal psychometric practice is to select test items by the extent to which they hang together (correlate) with other items -- using such criteria as factor loadings or item-to-total correlations. Doing anything else would damage both the reliability and validity of the test.

It has been suggested -- correctly -- that one could construct an IQ test that did not discriminate men and women by adding in more verbal items and deleting some mathematical items but since the difference between men and women is small, it is hard to see what would be gained by that and, as mentioned, it would detract from at least the validity (construct validity, to be technical) of the test and probably the reliability also.

It would be nice to think that one could dictate the final composition of an IQ test or any other test by the initial item pool that one starts out with before test construction begins but that is not wholly true if one is really doing research rather than just making things up. I have found, for instance, that a subset of items that are not strongly represented in the initial item pool can be quite dominant in the test which emerges from the item analyses. See here for instance. So reality tends to defy intentions and expectations. In other words, there is every reason to believe that existing IQ tests give a correctly balanced picture of IQ. What is in them has emerged empirically.



Go to John Ray's Main academic menu
Go to Menu of longer writings
Go to John Ray's basic home page
Go to John Ray's pictorial Home Page
Go to Selected pictures from John Ray's blogs