Background: Comparative responsiveness data are needed to inform choices about pain outcome measures. Objectives: To compare responsiveness of pain intensity, pain-related function, and composite measures, using data from a randomized trial and observational study. Research Design: Analysis of responsiveness. Subjects: A total of 427 adults with persistent back, hip, or knee pain were recruited from primary care. Methods: Participants completed Brief Pain Inventory, Chronic Pain Grade (CPG), Roland disability, SF-36 bodily pain, and pain global rating of change measures. We used the global rating as the anchor for standardized response mean and receiver operating characteristic curve analyses. We used the distribution-based standard error of measurement to estimate minimally important change. To assess responsiveness to the trial intervention, we evaluated standardized effect size statistics stratified by trial arm. Results: All measures were responsive to global improvement and all had fair-to-good accuracy in discriminating between participants with and without improvement. SF bodily pain was less responsive than other measures in several analyses. The 3-item PEG was similarly responsive to full Brief Pain Inventory scales. CPG and SF bodily pain were less responsive to the trial intervention and did not perform well among participants with hip/knee pain. Agreement between anchor and distribution-based methods was modest. Conclusions: If a brief measure is desired, the 3-item PEG is more responsive than the SF bodily pain scale. CPG and SF bodily pain scales may be relatively poor choices for trial outcome assessment. Both anchor and distribution-based methods should be considered when determining clinically important change.