Sixty house officers and 39 registered nurses in a university teaching hospital were surveyed to ascertain their knowledge of placebo action and their patterns of placebo use. The majority of physicians and nurses greatly underestimated the percentage of patients who experience pain relief when given placebo. Placebos typically were given to disliked patients who were suspected of exaggerating their pain or had failed to respond to usual medical regimens, or both. Positive responses to placebo medication were then interpreted by the physicians as evidence that the pain had no physiologic basis. Many studies have shown that overdemanding and complaining patients are, if anything, less likely to respond to placebo than patients well liked by the hospital staff. Nevertheless the results of our survey suggest that this is precisely the type of patient "at risk" for placebo treatment.