I have been told by a reliable source (prefer keeping private communications confidential) that Steven Pinker got a perfect score in both sections of the GRE in 1975. In that year, a perfect score on the verbal would be 2.4 standard deviations above the mean, while a perfect quant score would be 2.13 SD above the mean. However, this is coming from a highly selected sample of test takers - individuals who took the GRE had an average SAT score of 519, 1.23 SD above the mean.
To estimate his IQ, I simulated what the average IQ of somebody who score in both the top 3.36 SD of quant ability and 3.63 SD of verbal ability. He also has a PhD, so the datapoints also had to have an educational attainment
g-loading of GREV assumed to be .79. g-loading of GREQ assumed to be .77, g-loading of education assumed to be .55.
Simulation yields mean of 159 and standard error of 8.
set.seed(1)
g <- rnorm(60000000, mean=0)
iq <- 0.77*g + rnorm(60000000)*sqrt(1-0.77^2)
sc <- 0.79*g + rnorm(60000000)*sqrt(1-0.79^2)
ed <- 0.55*g + rnorm(60000000)*sqrt(1-0.55^2)
subby1 <- data.frame(iq, sc)
subby1$g = g
subby1$ed = ed
subby2 <- subset(subby1, (subby1$iq > 3.36) & (subby1$sc > 3.63) & (subby1$ed > 2.1))
mean(subby2$g)
sd(subby2$g)
mean(g)
sd(g)
sd(iq)
sd(sc)