SUMMARY A method of devising stepwise multiple testing procedures with fixed experimentwise error is presented. The method requires the set of hypotheses tested to be closed under intersection. The method is applied to the problem of comparing many treatments to one control and to ordered analysis of variance.