Tags
IQ, Not Politically Correct, predictive validity, Race Reailist, social class, test construction, the BITCH test, the g factor
Commenter Race Realist (RR) wrote yet another article claiming that IQ tests are based on circular logic and just measure social class. He writes:
In sum, what these tests test is what the test constructors presume—mainly, class and racial bias—so they get what they want to see. If the test does not match their presuppositions, the test gets discarded or reconstructed to fit with their biases…At best, IQ test scores measure the degree of cultural acquisition of knowledge; they do not, nor can they, measure ‘intelligence’—which is a cultural concept which changes with the times. The tests are inherently biased against certain groups; looking at the history and construction of IQ testing will make that clear. The tests are middle-class knowledge tests; not tests of ‘intelligence.’
RR is right that IQ tests were originally designed to confirm existing prejudices of who was smart by deliberately selecting test items that so-called smart people did better on. This is ironic because the whole point of creating an IQ test was that teachers’ judgments were considered too biased to trust, so why did the first IQ testers rely on teachers to decide who was smart?
Psychometric tasks are great at being objective, but they’re not always great at measuring intelligence. By contrast teachers are great at judging intelligence, but they’re not always objective. Thus by selecting only those test items that most confirmed teacher judgement, they got the best of both worlds: An objective scale that was great at measuring intelligence.
Of course RR might argue that the teachers were just judging social class, not intelligence, and by extension so were the tests. Further he would argue that if the tests predicted socioeconomic success, it was not because smart people rise to the top, but rather because SES is all the tests were measuring in the first place.
However we now know that IQ tests predict life outcomes, not because they correlate with teacher’s judgments, but because they correlate with g; the general factor of IQ tests.
Thomas R. Coyle writes:
g is one of the best predictors of school and work performance (for a review, see [7], pp. 270–305; see also, [8,9]). Moreover, a test’s g loading (i.e., its correlation with g) is directly related to its predictive power. In general, tests with strong g loadings correlate strongly with school and work criteria, whereas tests with weak g loadings correlate weakly with such criteria. For example, Jensen ([7], p. 280) found that the g loadings of the Wechsler Adult Intelligence Scale (WAIS) subtests were directly related to their predictive power for school criteria (e.g., school grades and class ranks). WAIS subtests with stronger g loadings generally predicted school criteria well, whereas subtests with weaker g loadings predicted such criteria poorly. Consistent with these findings, Thorndike [10] found that g explained most of the predictable variance in academic achievement (80–90%), whereas non-g factors (obtained after removing g from tests) explained a much smaller portion of variance (10–20%). Similar results have been found for job training and productivity, which are robustly related to g but negligibly related to non-g factors of tests (e.g., rnon-g < 0.10, [7], pp. 283–285; see also, [9,11]).
From Non-g Factors Predict Educational and Occupational Criteria: More than g
g is whatever variable(s) causing all cognitive abilities to positively inter-correlate. RR will tell you g is circular logic because any cognitive ability that doesn’t correlate with g is excluded, but this is false.
As Arthur Jensen (1998) noted, there are very clear rules on a) what is an ability, and b) what is a cognitive ability, and none of them require a correlation with other cognitive abilities.
A test measures ability if it a) measures voluntary behavior, b) has temporal stability, c) has a clear standard of proficiency, and d) some generality. There is another set of criteria that determines whether a particular ability is mental or physical.
IQ skeptics can cite tests that don’t correlate with g, but these tests don’t qualify as ability measures. One example are so-called creativity tests where you’re asked to name as many uses for a brick as you can think of in two minutes. Such tests often lack a clear standard of proficiency because silly answers (i.e. use it to comb your hair) get the same credit as good answers (use it to smash a window).
No one to my knowledge has come up with a mental test that actually qualifies as an ability test yet does not correlate with g with the possible exception of the BITCH test (ironic name for a test that’s supposed to fight anti-black bias) however the BITCH test is clearly culturally biased. None of the major IQ tests are culturally biased against any of the founding racial subgroups of the United States (at least as defined by psychometric criteria).
