The relative performance of three analgesic rating scales--visual pain analog, verbal pain intensity, and verbal pain relief--was assessed in clinical trials with 1,497 patients and a variety of pain models. The scales correlated strongly with one another, with inconsistent and generally minimal differences in sensitivity. Overall, the verbal relief scale tended to be slightly more sensitive than the pain analog rating, which in turn showed a small advantage over the verbal pain intensity assessment. When the scores derived from the categorized ratings 1 hour after drug dosing (generally the time of peak effect) were analyzed, there was little difference whether a parametric or nonparametric approach was taken. When the cumulative measures of overall effect over 6 hours were considered, however, the nonparametric approach was decidedly more powerful. There was a similar pattern when the analog scores were analyzed. This unanticipated finding appears to be due to the cumulative measures (from all three scales) being more skewed toward the lower end of their respective ranges than are the 1-hour scores. A composite efficacy variable was defined, incorporating data from the three primary scales; this measure was found to be generally comparable in sensitivity to the individual scales and may be useful as a global summary of response. While our investigation provides evidence that any of the ratings considered will accurately reflect analgesic response, the verbal relief scale was the most sensitive and might be the best choice if a single measure is desired.