What high-range tests measure

© March 2023 Paul Cooijmans

Latest thoughts on what high-range I.Q. tests measure

In the 1990s, the original motivation to occupy myself with intelligence testing was to find out whether and to what extent it is possible to measure intelligence in the high range. A question that seemed relevant was, does g, the common factor in mental testing, persist into the high range, or does it fall apart into group factors and specificity beyond some high I.Q. level? After more than a quarter of a century I analysed self-collected data in the report The differentiation hypothesis of g tested and concluded with some confidence that g does persist into the high range (albeit with some decrease), that g does not break up into other factors or specificity in the high range, and that the observed decrease of g in the high range is a mere statistical effect caused by the restriction of range that occurs when ever smaller (higher) segments of the high range are considered. When you restrict the range, you reduce variance, and thus reduce any correlations (which are based on variance), and thus reduce any factors (which are based on correlations).

Then, a next question is, is the g that is observed in the high-range the same as the g of mainstream psychometrics? This matter is sometimes brought up by people who seemingly believe that the answer is "no" and that mainstream g is the real g. This is a harder question that can currently not be answered through statistical analysis, mainly because of the (inherent) lack of mainstream test data with sufficient overlap with high-range test data. However, my latest thoughts on the possible answers to this question are as follows:

• Option (1): The observed gs are one and the the same, the only difference being that mainstream tests lack sufficiently hard problems to measure g into the high range;
• Option (2): There is a difference such that high-range tests give an undeserved advantage to systematizers, that is, those disposed for abstract-logical thinking styles. As a result, high-range tests are not good indicators of g; the real intelligence is g with the systematizing bias removed, so mainstream tests are better at measuring it.
• Option (3): There is a difference such that high-range tests give a deserved advantage to systematizers, while some mainstream tests are biased against systematizers and/or give an undeserved advantage to empathizers, that is, those who are disposed for dealing with people rather than abstract-logical matters. As a result, mainstream tests are less good indicators of g; the real intelligence is systematizing, or mainstream g with the empathizing bias removed, so high-range tests are better at measuring it. The bias that some mainstream tests have against systematizers may take the form of a lack of hard problems (which only systematizers can solve), of too short time limits, or of the inclusion of content types with low g loadings and possible bias in favour of empathizers. These content types may involve straightforward arithmetic, short-term memory, verbal fluency (for example, naming as many as possible words starting with a given letter within a limited time), reading, writing, grammar, spelling, perceptual speed (for instance, matching figures), speed and accuracy of clerical checking (like underlining certain letters in a text, or digit/symbol coding), and motor coordination and finger and manual dexterity. Such content types are present in a number of typical mainstream tests that are considered the "gold standard" of I.Q. testing, but are in actuality failing to measure g especially in the higher regions, and are notorious for the modest scores that widely accepted geniuses have achieved on them.

I currently think that (1) and (3) are more likely than (2). Also, (1) and (3) may be related in the sense that systematizers be the only ones able to solve hard problems to begin with, so that the advantage they have on high-range tests should be seen as rightful and deserved, and not as a "contaminant" of the measurement of g.

It is appropriate to remark here that systematizing should be regarded a healthy, adaptive trait, and not (part of) a disorder. This is relevant because systematizing is often said to be connected to autism, Asperger syndrome, or the "autism spectrum". But such disorders correlate negatively with I.Q. scores in the high range, according to several decades' worth of statistics. So if systematizing is the real g, it is either something separate from autism/Asperger, or it is only the adaptive aspect of autism/Asperger (correlating positively with I.Q.) without the disabling aspects (which correlate negatively with I.Q).

[More articles on intelligence]