g Factor (psychometrics) - "Indifference of The Indicator"

"Indifference of The Indicator"

Spearman proposed the principle of the indifference of the indicator, according to which the precise content of intelligence tests is unimportant for the purposes of identifying g, because g enters into performance on all kinds of tests. Any test can therefore be used as an indicator of g. Following Spearman, Arthur Jensen more recently argued that a g factor extracted from one test battery will always be the same, within the limits of measurement error, as that extracted from another battery, provided that the batteries are large and diverse. According to this view, every mental test, no matter how distinctive, contains some g. Thus a composite score of a number of different tests will have relatively more g than any of the individual test scores, because the g components cumulate into the composite score, while the uncorrelated non-g components will cancel each other out. Theoretically, the composite score of an infinitely large, diverse test battery would, then, be a perfect measure of g.

In contrast, L.L. Thurstone argued that a g factor extracted from a test battery reflects the average of all the abilities called for by the particular battery, and that g therefore varies from one battery to another and "has no fundamental psychological significance." Along similar lines, John Horn argued that g factors are meaningless because they are not invariant across test batteries, maintaining that correlations between different ability measures arise because it is difficult to define a human action that depends on just one ability.

To show that different batteries reflect the same g, one must administer several test batteries to the same individuals, extract g factors from each battery, and show that the factors are highly correlated. Wendy Johnson and colleagues have published two such studies. The first found that the correlations between g factors extracted from three different batteries were .99, .99, and 1.00, supporting the hypothesis that g factors from different batteries are the same and that the identification of g is not dependent on the specific abilities assessed. The second study found that g factors derived from four of five test batteries correlated at between .95–1.00, while the correlations ranged from .79 to .96 for the fifth battery, the Cattell Culture Fair Intelligence Test (the CFIT). They attributed the somewhat lower correlations with the CFIT battery to its lack of content diversity for it contains only matrix-type items, and interpreted the findings as supporting the contention that g factors derived from different test batteries are the same provided that the batteries are diverse enough. The results suggest that the same g can be consistently identified from different test batteries.