Anders Ericsson, in his popular book Peak: Secrets from the New Science of Expertise, claims that though one may need a reasonably high IQ to even be a scientist, after that level, IQ is irrelevant to scientific success.
The average IQ of scientists is certainly higher than the IQ of the general population, but among scientists, there is no correlation between IQ and scientific productivity. Indeed, a number of noble prize winning scientists have had IQs that would not even qualify them for Mensa, an organization who's members must have a measured IQ of at least 132, a number that put's you in the upper 2 percentile of the population.
(From chapter 8: "But What About Natural Talent", of Peak: Secrets from the New Science of Expertise)
My understanding is that this is completely wrong, and the best scientists tend to have higher IQs, at all levels of performance. But this is a background belief that I haven't concretely verified.
Is there a canonical reference regarding the impact of IQ on scientific contribution?