A reader sent me the following email:

“I’ve always wondered if a person with strong drawing (or painting) abilities would score well on perceptual reasoning. Is there a correlation? If so is it a very strong and clear correlation?”

I don’t think there is. A study of prodigies documents a case of a painting prodigy who scored IQ 88 (U.S. norms) in spatial ability (bottom fifth of Americans her age)

At the time of testing, prodigy seven, a self-taught painter, was nineteen years of age. Although she had been interested in arts and crafts for as long as she could remember, she did not begin to paint until she was 13 years old. She said she was inspired by a young art prodigy she saw on television. After just eighteen months of painting, the prodigy won the National Gold Key award, the most prestigious art award given to high school students, at the age of 15. She was one of 50 recipients of the award across the country. At nineteen, her work is selling for thousands of dollars and is displayed in prestigious art museums around the world.

Testing results

Total IQ Score= 112; Fluid reasoning= 100; Knowledge= 128; Quantitative Reasoning= 100; Visual Spatial Abilities= 88; Working Memory= 138

It seems artistic ability requires certain special abilities, like being able to differentiate hues that have little to do with the perceptual reasoning tasks on tests like the WAIS or the Binet.
Another reader sent me the following email:

“So I’ve stumbled across one of your articles in your blog, and in it you said that WAIS IV has correlation with g of 0.9 (if I’m not mistaken?). Also somewhere in the comments I’ve read that Raven’s Advanced Progressive Matrices has correlation with g of 0.8 (if I’m not mistaken?).

Where do these numbers come from?
Also if these are correct, does that mean that WAIS IV should give you a better representation of your ‘overall’ intelligence?
So to clarify, does higher g-loading mean that the test is better approximation for your ‘true’ IQ?
Thanks in advance!”
The 0.9 figure for the WAIS-R (and by inference other editions) comes from Jensen.  He mentions it in his landmark book Bias in Mental Testing.
The 0.8+ g loading often cited for the Raven is probably too high (at least in today’s population).

In a 2017 study by Dimitri van der Lindena, Curtis S. Dunkelb, Guy Madison, the Raven was found to have a g loading of 0.609 in 900 American healthy young adults.


However in my opinion this sample likely suffered from range restriction, and the actual g loading of the Raven in the general U.S. population is about 0.7.  That’s roughly the g loading of the Matrix Reasoning subtest on the WAIS-IV in an extremely representative sample of Americans.

g is simply whatever is causing all cognitive abilities ever discovered to positively correlate in nationally representative samples, so the more an ability loads on g, the more predictive it is of one’s overall cognition (at least in the biologically normal population). There’s a lot more to intelligence and IQ than just g but g‘s perhaps the most scientifically precise measure of intelligence we have.