Reply to Chris Langan

Kevin Langdon

Once again we are treated to violent language from Chris Langan. I wonder if he thinks that this is helping his cause in this group.

The following is from Chapter 2 of The g Factor, "The Discovery of g":

Inspection of a great many factor analyses of the widest variety of mental tests imaginable reveals without exception that tests' g loadings are a perfectly continuous variable, raning from slightly greater than zero to slightly less than unity. . . . [Charles] Spearman invented a type of test that was entirely non-verbal, that was composed of fundaments (various geometric shapes) that are universally familiar to virtually all persons beyond three years of age in every culture, and in which every item calls for the eduction of relations and correlates. It also has the quality of "abstractness" in the sense that the fundaments (straight and curved lines, triangles, circles, squares, and the like) do not represent any real or tanglible objects, like animals, plants, furniture, or vehicles. . . . Spearman's test was further developed by one of his students, John Raven, and the eminent geneticist Lionel Penrose. . . . When the Progressive Matrices test is factor analyzed along with a variety of other tests, it is typically among the two or three tests having the highest g loadings, usually around .80. Probably its most distinctive feature is its very low loadings on any factor other than g.

The items on the LAIT (and on the nonverbal portion of the Mega Test) share this property of "abstractness." And an almost exclusive loading on g is exactly the pattern that Grady Towers found in his analysis of the LAIT. It is highly unlikely that this is a "different g"; there was no significant spatial factor at all.

Chris is wrong about the speed factor, too. The proof of this was provided in a passage from Dr. Jensen's Bias in Mental Testing reprinted in my reply to Chris Langan in Noesis #135. I reproduce only a portion of that passage here:

If the increase in score leaves unaltered the subject's rank order, the speed factor is of little importance. That is, the time or speed factor does not contaminate the scores with some ability or trait extraneous to what the test attempts to measure, in this case, intelligence. Usually the correlation between strictly timed and leniently timed administration is as high as the reliability of the test. When the correlation between the two timed conditions falls significantly below the reliability, the recommended time limit should be viewed with suspicion. It means that the speed factor is given too much weight in the test scores, when what we really want to measure is mental power rather than some kind of "personal tempo" factor. The personal tempo factor actually has little if any correlation with intelligence. E.L. Thorndike (1927, pp. 400-401) tried to determine the correlation between speed and altitude. To measure speed he used a large number of quite easy items and recorded the time that subjects required to complete a given number of such easy items without error. To measure altitude he gave subjects a succession of items steeply graded in difficulty; the altitude score was the difficulty level beyond which the subject failed 50 percent or more of the items. The correlations between the measure of speed (the reciprocal of time) and of altitude averaged about .40 in several groups (about .46 when corrected for attenuation).
This correlation suggests that the time that subjects require for the easy items that they all can do is not measuring the same thing as the number of steeply graded items that persons can get right without time limit. In other words, it indicates the presence of a speed factor that is independent of a power or altitude factor as a source of variance in test scores. The correlation of about .40 does not necessarily mean that the speed factor is correlated at all with the altitude factor, because the items used to measure speed still had some low level of difficulty so that these items were not a pure measure of the speed factor. We know that the average time required per item is correlated with the item's difficulty.

. . .

In what is probably the best experimental study of the matter, the correlation between subjects' speed scores and power scores, when difficulty level and response accuracy were controlled, is close to zero for all kinds of test items.