Around 1952, scholar Anne Roe had 64 extremely eminent scientists take a high cieling intelligence test created with the help of the Education Testing Service (ETS). They reportedly averaged a verbal IQ of 166 (range 121 to 177), a spatial IQ of 137 (range 123-164) and a math IQ of 154 (range 128 to 194).
I’m going to ignore the math section because the physical scientists found that test too easy and so their scores were excluded. Also, math is more of an academic skill than an underlying cognitive ability independent of verbal and spatial IQ, which is why the Wechsler IQ scales provide verbal and spatial IQs, but not math IQs.
Now the test these scientists took was normed by comparing their scores to the scores of students of known (Stanford Binet?) IQs, but since the students were presumably in their early twenties, and the scientists seemed to average in their late 40s, the scores should be adjusted for age. On the WAIS-R for example, people in their late 40s score the equivalent of 5 IQ points higher than people in their early 20s on Vocabuary (an excellent measure of verbal IQ), and 10 points lower on Block Design (an excellent measure of Spatial IQ), thus for the scientists, 5 points should be subtracted from the verbal IQ (166 – 5 = 161) and 10 points should be added to the spatial IQ (137 + 10 = 147).
Because the test the scientists took was probably normed using students who took the 1937 Stanford Binet, the score were likely already 15 years inflated by the time the scientists were tested circa 1952. It’s known that verbal IQ norms become inflated by 0.2 points a year and spatial IQ norms become inflated by 0.4 points per year, so their verbal IQs must be reduced by 3 points and their spatial IQs must be reduced by 6 points.
So on a scale where the average white American of their generation scored 100 (SD = 15), these 64 super eminent scientists averaged a verbal IQ of 158 (range 113-169) and a spatial IQ of 141 (range 127-168).
Assuming verbal IQ and spatial IQ correlate about 0.67 (which is the correlation between the verbal IQ and performance IQ on the WISC-R), we should expect these scientists to have had a full-scale IQ of about 155.
Would the most eminent scientists today average that high (relative to today’s white norms) on a high ceiling IQ test? I don’t know. The correlation between IQ and some measures of academic success (years of schooling) have dropped precipitously since the 1950s, so it’s possible that scientific eminence has also become much less related to IQ, causing elite scientists today to be much less exceptional.