Genes, Brains, and Human Potential The Science and Ideology of Intelligence

PRETEND INTELLIGENCE 87

Th e approach has been much admired in some quarters; indeed, it is
inventive. However, it has also been controversial and heavi ly criticized.
Th is is because of the large number of assumptions that enter into the
corrections and the poor quality of the original data and reports. Let us
have a quick look at some of these, just to indicate how the area has be-
come very murky indeed.

MAKE- DO TESTING (AGAIN)

A major prob lem with the Schmidt and Hunter validation studies is the

haphazard diversity of tests used in the original studies. Th ey have in-

cluded memory tests, reading tests, scholastic aptitude tests (SATs), uni-

versity admission tests, specialized employment se lection tests, and a vast

range of specialist armed forces tests: almost anything, in fact, that yields

numbers. Just calling these “general ability tests,” as Schmidt and Hunter

do, is like reducing a diversity of serum counts to a “general blood test.”

Moreover, the studies incorporated into the most- cited meta- analyses

cover a vast range of dates, some from the 1920s, with the majority of

them from before the 1970s.

Th e point is, of course, that such diverse tests are bound to tap diff er ent

abilities with vari ous sources of individual diff erences and with varying

distributions. Even todays’ attainment tests show little intercorrelation.

For example, as reported by the College Board in 2008, the correlation

between SAT scores and high school grade point average is only 0.28.

And the correlation of either with IQ tests is only around 0.2. Rolling

these up together as if they were expressions of the same “general ability”

seems little more than guesswork. As Kevin Murphy, himself a meta-

analytic expert, reported in a major review, it creates “lack of clarity in

what population pa ram e ter is being estimated.”^14

JOB PER FOR MANCE?

In contrast to the vast diversity of “intelligence” tests, only one mea sure

of job per for mance has been used in the majority of studies: supervisors’

ratings. It turns out that there are a host of prob lems with such ratings.

This content downloaded from 139.184.14.159 on Tue, 17 Oct 2017 13:52:12 UTC