During multiple testing, researchers often adjust their alpha level to control the familywise error rate for a statistical inference about a joint union alternative hypothesis (e.g., “ H1 or H2 ”). However, in some cases, they do not make this inference.