« Carl on Hobbits - Update | Gene Expression Front Page | The Urban Sink »
June 15, 2005

Hard question for GNXP'ers

After the recent post, I notice that there still seems to be considerable debate on the Flynn effect, and I don't see any particular prevailing view among the participants here on GNXP.

So I'll ask this question flat out:

What makes you so sure (if you are sure) that IQ measurements of intelligence can be highly inaccurate across time but not across culture?

In the extended entry, I'm including a nicely succinct defense of IQ that appeared on Jerry Pournelle's site.

I'd appeal to data rather than analogies. Several kinds come to mind. The supporting data can be found in my articles and conference presentations with titles mentioning "life," "everyday life," "death," "practical," "why g matters," "health," and such. See also Herrnstein and Murray for number 1 below.


1. IQ (as long as it's a good measure of g) predicts a broad range of life outcomes better than does SES, from GPA to longevity. Corollary: You can wash out IQ's apparent predictive superiority only if you load your SES battery with additional surrogates for parents' or own g.

2. The phenotypic correlations between IQ and measures of social class (education, occupational prestige, income) are from a half to two-thirds genetic in origin.

3. SES cannot explain the big IQ differences among siblings growing up in the same household: They differ two-thirds as much in IQ, on the average (11-12 points), as do any two random strangers (~17 points). This is a glaring fact that SES enthusiasts have studiously ignored.

4. Adult functional literacy (e.g., see the fed's NALS survey) predicts life outcomes in exactly the same pattern as does IQ, though they won't tell you that. Functional literacy is measured by having subjects carry out everyday life tasks, such as using a menu to figure out the price for something. Persons scoring at levels 1-2 (out of 5) have been described as not having the ability to use their rights or meet their responsibilities in the modern world (40% of whites, 80% of blacks). Pick out a few NALS tasks at various levels and ask your critic what % of adults s/he thinks can perform them. They will be shocked and so will you when you see the data--go to my 1997 "Why g matters" article for NALS, or my 2002 "highly general and highly practical" chapter for health literacy items--e.g., on diabetes.

5. IQ predicts on-the-job performance better overall than any other single predictor (SES isn't even in the running), it predicts better when performance is objectively rather than subjectively measured, and when the tasks/occupations are more complex in what they require workers to do. At the same cognitive complexity level, IQ predicts job performance equally well in manual and non-manual jobs (e.g., trades vs. clerical. The exact same complexity pattern is found with functional literacy--the hardest items are the most complex (require more inference, are abstract rather than concrete, contain more distracting irrelevant information, etc.)

6. A large followup of Australian veterans found that IQ was the best predictor of death by age 40 (had 50+ predictors). Vehicle fatalities were the biggest cause (as is typical), and, compared to men with IQs of 100+, men of IQ85-100 had twice the rate and men IQ 80-85 had three times the rate. (Remember, SES could not explain this.) The US (and apparently Australia) forbid induction of persons below IQ 80 because they are not sufficiently trainable--found out the hard way.

7. Finally, if you succeed in describing g as a general learning and reasoning ability (one that gives high g people an increasing edge when tasks are more complex), then it is easy to show g's life and death relevance when you describe how health self-care and accident prevention are highly dependent on learning and reasoning. Consider what it takes to be an effective diabetic--lots and lots of judgment on a daily basis, or you're likely to lose your sight, your limbs, etc.

The more usual claim is that IQ is just an "academic ability," but it clearly is far more practical than that. It is far more practical than Sternberg's so-called "practical intelligence," whose existence rests solely on the chimerical evidence he conjures up.

You'll need a separate set of facts if the complaint is that IQ differences result mostly from nurture, not nature.

Linda S. Gottfredson
Professor, School of Education
University of Delaware
Newark, DE 19716 USA

Posted by Thrasymachus at 02:00 PM