Serial submaximal treadmill tests are often used to evaluate the efficacy of therapy in patients with atrial fibrillation. Since the response to serial tests can be influenced by a 'learning phenomenon', we performed maximal exercise tests on 9 patients (mean age 63 +/- 4 years) with chronic atrial fibrillation. Points of analysis for the initial and follow-up treadmill exercise tests were 3 mph/0% grade, the gas exchange anaerobic threshold, and maximal exertion. Significant (p less than 0.05) reductions in ventilation (l/min) and oxygen uptake (ml/kg/min) were observed on follow-up at a standard submaximal work load of 3.0 mph/0% grade and at the gas exchange anaerobic threshold. There was no significant alteration in these variables at maximal exertion. A reduction in heart rate was observed throughout exercise during the follow-up test with the most marked reduction (21 beats/min) occurring at 3.0 mph/0% grade. There were no differences in respiratory exchange ratio or systolic blood pressure at any point. The reduction in submaximal heart rate and gas exchange variables without a significant change in these variables at maximal exertion is consistent with a learning effect. Therefore, studies comparing consecutive submaximal exercise test responses in patients with atrial fibrillation can be misleading.