Comment by civilized

I don't know. If the experiments were perfectly random, the Buttondown-Substack difference is a 3-sigma effect, so the difference is unlikely to be random chance. But why? What's so special about one vs the other?