The effect of triamcinolone subacromial bursa injection versus naproxen therapy was compared in a randomized, double-blind, placebo-controlled study of 100 patients who had painful shoulders. Outcome was compared using degree of active abduction, pain, limitation of function, and a clinical index that combined equally weighted measures of all of these. In a time-adjusted analysis, triamcinolone was superior to placebo in all clinical variables. Naproxen was superior to placebo in all variables except pain. Triamcinolone was superior to naproxen in the relief of pain (P = 0.04) and the clinical index (P = 0.04). Multiple linear regression analysis showed that naproxen and triamcinolone treatment accounted for only 16% of the variation in outcome, compared with 44% accounted for by the clinical index prior to treatment. Thus, patients with a poor pretreatment clinical index (those with the most room for improvement) were least likely to improve. We conclude that both triamcinolone (P = 0.00005) and naproxen (P = 0.02) are superior to placebo in the treatment of the painful shoulder.