The purpose of this study was to evaluate the moderate term (5 weeks) reproducibility of Borg scale ratings of the effort to breathe (Borge) and the degree of discomfort evoked by breathing (Borgd) in patients with COPD during exercise. Six subjects with moderately severe COPD (FEV1, 1.42 +/- 0.50 L) underwent progressive incremental exercise (15 W/min) on a cycle ergometer to a symptom-limited maximum every week for 6 weeks (first week used as practice session). Minute ventilation (VE), oxygen consumption (VO2), and Borg ratings were obtained every minute during exercise. Borge and Borgd were highly correlated in each subject (r = 0.99 +/- 0.01). Borg scores were not significantly different across study days during both maximal and submaximal exercise. The within-subject coefficient of variation (CV) for Borge during maximal exercise was 13.9 +/- 9.0% (range, 6 to 31%) which was not significantly different from that observed for the physiological indices: 8.2 +/- 4.1% (range, 4 to 15%) for VE and 5.2 +/- 3.4% (range, 1 to 10%) for VO2. In contrast, at 66% of the maximum workload, the within-subject CV for Borge was 25.0 +/- 13.6% (range, 12 to 50%) which was significantly greater than that observed for the physiologic indices: 5.8 +/- 2.0% (range, 3 to 9%) for VE and 4.6 +/- 1.1% (range, 3 to 6%) for VO2. In every subject, Borge was linearly correlated with VE, VO2, and workload. However, within an individual subject, the slope of these relationships varied between trials; within-subject CV for the slope of the Borge/VE relationship was 20.2 +/- 8.0% (range, 12 to 32%). In conclusion, during incremental exercise Borg ratings of dyspnea are not as reproducible as physiologic indices in patients with COPD.