The effect of gender and training on repeatability of cardiopulmonary exercise performance has not been well defined. Therefore, we performed two bicycle exercise tests 1 wk apart in each of two groups: In 19 normal subjects (age 24 to 64 yr, 10 females), with a mean maximal oxygen uptake (VO2max) of 42 mL.kg-1.min-1, who had been in an ongoing training program including bicycle exercise, and in 19 untrained volunteers (23 to 54 yr, 11 females) with a mean VO2max of 36 mL.kg-1.min-1 (P < 0.05). Mean differences in physiologic variables measured during tests 1 and 2 were calculated. Repeatability coefficients were defined as 2 SD in percent of the means. In untrained subjects mean (+/- SD) maximal heart rate decreased by 4 +/- 5 beats.min-1 from the first to the second test (P < 0.05). VO2max and maximal work rate (Wmax) remained unchanged. No significant changes in these or other variables occurred in trained subjects. Repeatability coefficients for VO2max were 8 and 13% in trained and untrained subjects, respectively (P = NS). For Wmax the repeatability coefficient in untrained (11%) exceeded that in trained subjects (4%, P < 0.05). Repeatability coefficients did not differ among males and females. Our study provides normal values for repeatability of various parameters assessed during exercise testing and demonstrates that interpretation of performance during repeated tests has to account for training of the subjects.