One-sided tests provide actual type I error probabilities that are consistent with specified alpha levels in situations where differences in only one direction will ever be the basis for a consequential claim of significance. A drug-placebo comparison presented to the FDA in support of a new drug application is clearly a case in point. It is important to use the correct one-sided tests in all cases where they are appropriate in order to avoid the suspicion that they are only used to relax criteria so that marginal results can be reported as significant. The choice of a one-sided test should be a design consideration with consequent reduction in sample sizes commensurate with greater power against consequential alternatives.