In this paper (Rubin, 2022), I make two related points: (1) researchers should halve two-sided p values if they wish to use them to make directional claims, and (2) researchers should not halve their alpha level if they’re using two one-sided tests to test two directional null hypotheses.