To be considered well designed, a study of analgesic efficacy must be conducted under double-blind conditions with patients randomly allocated to each treatment group. Ideally, the study should include both a placebo and an acceptable standard or reference analgesic. Extraneous variables should be kept to a minimum. The most firmly established pain models have used pain caused by cancer, parturition (including episiotomy), surgery postoperatively, oral surgery, and headache. The cancer pain model is most amenable to crossover or multiple-dose studies because of the chronic nature of the pain. Appropriately designed clinical studies have provided conclusive evidence that aspirin and acetaminophen are equianalgesic and, milligram for milligram, equipotent. There also is some evidence that acetaminophen can reduce swelling in inflammatory conditions other than arthritis (eg, in patients who have had oral surgery).