Formal compliance with Level-I criteria (randomization and a control group) may yield data subject to limitations. Loopholes include lack of-, or ineffective blinding, or qualitatively different treatments for the active versus control arms. However, subjects and evaluators CAN be blinded using straightforward and ethical methods. Review boards may debate whether controls should receive no therapy or sham therapy. The best answer will depend upon the hypothesis under investigation, and multiple studies, or multi-arm trials may be required. A more important ethical issue is to settle questions about efficacy as accurately and as harmlessly as possible. Patient interactions (unintentional cues) and lack of investigator equipoise are additional sources of bias that may be confused with placebo effects. Finally, investigators should refrain from non-study use of the therapy under study.