&NA;In evaluations of new drugs, even when a random assignment, double‐blind, controlled clinical trial design is used, an array of methodological problems associated with treatment of those who do not respond (treatment nonresponders) can lead to unclear or misleading conclusions. Four such problems are reviewed and illustrated: unwitting diagnostic heterogeneity of depressive subtypes in study samples, inadequate dose and duration of medication, selection of inappropriate criteria for determination of efficacy, and nonreporting of global outcome measures for individual patients. The solutions offered to these problems are not novel. However, the frequent occurrence of these errors in recent reports of antidepressant trials suggests that a review of these problems and possible solutions would be timely. In addition, the authors describe an alternate, comparatively underutilized strategy, the discontinuation design, that is particularly suitable for new drug trials and that circumvents the problem of treatment nonre‐sponders.