A placebo may be a pharmacologically active or an inert substance, a procedure, or a patient-doctor interview. Placebos work best in symptoms or disease which vary over time and between patients. The placebo works best in behaviour disorders, somatic autonomic disorders like pain, and neurohumoral disorders like hypertension. However, placebo action is incompletely defined in its molecular pharmacology. The endogenous brain systems of opioid, antiopioid, and gamma-aminobutyric acid polypeptide transmitters and neuronal receptors account in part for placebo analgesia. Non-painful stress may be mediated through other neurohumoral systems. A separate neural system might control these subsystems. Confidence based on the doctor's empathy commonly evokes the placebo effect. How the symbolic input of thought or emotion is translated into neuronal events is unknown. Double-masked 'controlled' clinical trials use placebo to reduce bias; overuse of placebo here may harm some patients. Oral placebos for routine use include thiamine at low dose. Potent drugs like glucocorticoids cannot be justified as placebo in mild disease or non-disease. Both patient and doctor are usually unaware of the placebo effect during interviews. Doctors may increase placebo efficacy by improving interpersonal skills.