Re: shows that if any correlation exists, it's extremely weak
Hypothetically: even though the correlation is demonstrably weak in a population sense; for a small subset of somehow suceptible individuals, it could be strong. And with the strength of the "suceptible" correlation being drowned out by the noise and other confounding factors (eg lead, as mentioned elsewhere in this thread) in the dominant non-suceptible population. Whether or not this might be true I do not know, but it's not implausible.