[schoen]

I should also have been more careful to more clearly distinguish spurious correlations from correlations where both things are related to a third factor (which I think is probably often the case in John Robbins's graphs and somewhat plausibly here).