Cohort study of patients with low back pain (LBP) receiving physical therapy. To examine the responsiveness characteristics of the numerical pain rating scale (NPRS) in patients with LBP using a variety of methods. Although several studies have assessed the reliability and validity of the NPRS, few studies have characterized its responsiveness in patients with LBP. Determination of change on the NPRS during 1 and 4 weeks was examined by calculating mean change, standardized effect size, Guyatt Responsiveness Index, area under a receiver operating characteristic curve, minimum clinically important difference, and minimum detectable change. Change in the NPRS from baseline to the 1 and 4-week follow-up was compared to the average of the patient and therapist's perceived improvement using the 15-point Global Rating of Change scale. The majority of patients had clinically meaningful improvement after both 1 and 4 weeks of rehabilitation. The standard error of measure was equal to 1.02, corresponding to a minimum detectable change of 2 points. The area under the curve at the 1 and 4-week follow-up was 0.72 (0.62, 0.81) and 0.92 (0.86, 0.97), respectively. The minimum clinically important difference at the 1 and 4-week follow-up corresponded to a change of 2.2 and 1.5 points, respectively. Clinicians can be confident that a 2-point change on the NPRS represents clinically meaningful change that exceeds the bounds of measurement error.