The following shows how men and women compare on the latest edition of the Wechsler Adult Intelligence scales (Note: the index scores and IQ are expressed on a scale where the sex combined mean and standard deviation (SD) of Americans are 100 and 15 respectively, while the subtests use a scale where the mean and SD are 10 and 3 respectively):
If we take this at face value, the best measure of intelligence is full-scale IQ and here we see men are a little smarter than women (mean IQ 101.2 vs 98.9) and a little more variable (SD = 15.3 vs SD = 14.6).
Unfortunately we can’t take this at face value because at least in earlier versions of the WAIS, there were attempts to eliminate sex differences by removing (or at least counterbalancing) items or even subtests that favored one sex over the other.
So how then can we find the true sex difference?
One partial solution is to look at content-free subtests like digit span and block design. Unlike a general knowledge subtest where you can arbitrarily select items that favor men (knowledge of sports) or women (knowledge of fashion), you can not select a series of digits or an abstract visual diagram that favors (or fails to favor) one sex or the other. Thus content-free subtests are not amenable to arbitrary attempts to eliminate or create sex differences.
With the exception of the three subtests that makeup the Verbal Comprehension index (Vocabulary, Similarities, Information) all the WAIS-IV subtests are content-free. Thus by averaging the remaining three indexes that makeup the Full-Scale IQ (Perceptual Reasoning, Working Memory, Processing Speed), we might get a less biased measure of the sexes.
When we do this, men average 100.4 (SD = 15.2) and women average 99.6 (SD = 14.5). A trivial difference. Indeed rounded to the nearest whole numbers, both sexes are 100 with an SD of 15.
Ironically by looking only at subtests where they couldn’t remove sex differences, I found even fewer sex differences!
Does this mean sex differences in IQ are virtually non-existent? Not necessarily, since the subtests themselves may have been selected to have small or counterbalancing sex differences, even when sex differences on specific items within said subtests can not be removed.