According to this source , at age 8.3, Terence Tao scored 290 on the verbal section of the pre-1995 SAT(hat-tip to commenter Tenn for finding this data since my google searches for Tao’s verbal SATs turned up nothing) .
According to the book The Bell Curve (pge 694), if all American 17-year-olds took the old SAT in the 1980s (not just the college-bound elite) the mean verbal score would be 376 and the standard deviation (SD) would be 102.
Thus Tao scored 0.84 standard deviations below the average American-17-year-old (IQ 87 U.S. norms; IQ 84 U.S. white norms). But given that he was only 8.83, he deserves a huge age bonus.
On the WISC-R Vocabulary subtest (the subtest most similar to the verbal SAT), an 8.8-year-old who scores about 0.84 SD below American 16.8-year-olds (16-25 percentile), is actually about 2.17 SD above the mean for his own age (98-99 percentile), implying a verbal IQ of 133 (U.S. norms) or 132 (U.S. white norms).
A verbal IQ of 133 is extremely high but it’s still 47 points below Tao’s math IQ of about 180. I don’t know what the correlation between SAT verbal and math would be if the general population (not just the college bound segment) took the SAT, but let’s say it’s 0.67 (like the correlation between the verbal and performance scales on the WISC-R).
We would expect the average person with a math IQ of 180 to have a verbal IQ of 0.67(80) + 100 = 154, with 95% of all cases falling between 132 and 176.
Many people think it impossible that Richard Feynman could have obtained a valid IQ in the mid 120s as a child, but if the test was primarily verbal, we can see how it could have happened.