The evaluation of dyspnea and its responsiveness to therapy in COPD should consider the multidimensional nature of this symptom in each of its sensory-perceptual (intensity, quality), affective and impact domains. To gain new insights into mechanisms of dyspnea relief following pulmonary rehabilitation (PR), we examined effects on the major domains of dyspnea and their interaction with physiological training effects. This randomized, controlled study was conducted in 48 subjects with COPD. Subjects received either 8-weeks of PR or usual care (CTRL). Pre- and post-intervention assessments included: sensory-perceptual (i.e., exertional dyspnea intensity, dyspnea descriptors at end-exercise), affective (i.e., intensity of breathing-related anxiety during exercise, COPD self-efficacy, walking self-efficacy) and impact (i.e., activity-related dyspnea measured by the Baseline/Transition Dyspnea Index, Chronic Respiratory Questionnaire dyspnea component, St. George's Respiratory Disease Questionnaire activity component) domains of dyspnea; functional performance (i.e., 6-minute walk, endurance shuttle walk); pulmonary function; and physiological measurements during constant work rate cycle exercise at 75% of the peak incremental work rate. Forty-one subjects completed the study: PR (n = 17) and CTRL (n = 24) groups were well matched for age, sex, body size and pulmonary function. There were no significant between-group differences in pre- to post-intervention changes in pulmonary function or physiological parameters during exercise. After PR versus CTRL, significant improvements were found in the affective and impact domains but not in the sensory-perceptual domain of dyspnea. In conclusion, clinically meaningful improvements in the affective and impact domains of dyspnea occurred in response to PR in the absence of consistent physiological training effects.