Example: Hawking: 150ish IQ, John Sununu 190.
Many years ago there was a brief vogue among a few companies for psych testing potential employees. So I paid to have myself tested so I'd know what my potential employers "knew". Among other things, the tests informed me that I have an IQ that is 4.3 standard deviations above the mean.
This got me thinking. Which is more likely, that I'm smarter than 99.999% of the population, or that the test score was bogus? It should be obvious that it's far more likely that my test results were bogus!
Just because we can assign a single number to a person's intelligence the way we can to that person's height or weight doesn't mean that that number is as objective as height or weight is. What IQ tests purport to measure *cannot be observed directly*, and therefore cannot be measured directly. So we must not lose sight of the fact that IQ tests are *devised* by psychologists to correlate with something. How do they do this? By comparing a test's scores against something easy to measure -- rank in school for example. An IQ test that correlates poorly to performance in school would be considered "faulty", but one that correlates strongly to performancve in school would be considered "accurate".
In other words, IQ tests are only as meaningful as the outcomes they're deisgned to correlate with. An IQ test correlated to school success doesn't necessarily correlate precisely with "street smarts", many components of which are evolutionarily important (e.g. reading facial expressions).
Another thing to consider about how the test are calibrated is that the result is bound to be reliable ONLY near the mean, simply because confirmatory data out on the tails of the distribution is necessarily rare. So while I'd lend considerable credence to the 20 point spread between a 90 IQ and aa 110 IQ, I wouldn't lend the same credence to a difference between 140 and 160. I'd lend no credence whatsoever to the difference between a 140 and 160 IQ.
Basically, I consider distinctions betwen IQs over 125 unreliable, and distinctions between IQs over 135 as absolutely meaningless. There's no epistemological justfication for ranking people's intellectual abilities by IQ at that level. It's entirely possible that John Sunnunu would score 2.6 standard deviations higher than Stephen Hawking, but that's an artifact of the test, not reality.