We review three possible theoretical mechanisms for the placebo effect: conditioning, expectancy and endogenous opiates and consider the implications of the first two for clinical research and practice in the area of pain management. Methodological issues in the use of placebos as controls are discussed and include subtractive versus additive expectancy effects, no treatment controls, active placebo controls, the balanced placebo design, between- versus within-group designs, triple blind methodology and the double expectancy design. Therapeutically, the possibility of shaping negative placebo responses through placebo sag, overservicing and the use of placebos on their own are explored. Suggestions for using conditioned placebos strategically in conjunction with nonplacebos are made and ways of maximizing the placebo component of nonplacebo treatments are examined. Finally, the importance of investigating the placebo effect in its own right is advocated in order to better understand the long-neglected psychological aspects of the therapeutic transaction.