Researchers disagree whether placebo effects are clinically important. The controversy is partly due to conceptual and methodological problems. The objective of this paper is to analyze the main methodological problems in the estimation of placebo effects. Variations in the meaning of the concept of placebo effect are described, and the main methodological problems are analyzed. The notion of placebo effect has at least two main meanings: effect of placebo intervention, and effect of patient-provider interaction. When the terms are defined pragmatically, effects of placebo can be estimated as the difference between placebo and no-treatment in randomized trials. The effect of patient-provider interaction can similarly be assessed by comparing manipulation of the patient-provider interaction with no manipulation. In both cases bias due to lack of double-blinding is a potential problem.