Abstract : OBJECTIVE: To test whether ventilatory thresholds, measured during an exercise test, could be assessed using time varying analysis of respiratory sinus arrhythmia frequency (f(RSA)). METHODS: Fourteen sedentary subjects and 12 endurance athletes performed a graded and maximal exercise test on a cycle ergometer: initial load 75 W (sedentary subjects) and 150 W (athletes), increments 37.5 W/2 min. f(RSA) was extracted from heart period series using an evolutive model. First (T(V1)) and second (T(V2)) ventilatory thresholds were determined from the time course curves of ventilation and ventilatory equivalents for O(2) and CO(2). RESULTS: f(RSA) was accurately extracted from all recordings and positively correlated to respiratory frequency (r = 0.96 (0.03), p<0.01). In 21 of the 26 subjects, two successive non-linear increases were determined in f(RSA), defining the first (T(RSA1)) and second (T(RSA2)) f(RSA) thresholds. When expressed as a function of power, T(RSA1) and T(RSA2) were not significantly different from and closely linked to T(V1) (r = 0.99, p<0.001) and T(V2) (r = 0.99, p<0.001), respectively. In the five remaining subjects, only one non-linear increase was observed close to T(V2). Significant differences (p<0.04) were found between athlete and sedentary groups when T(RSA1) and T(RSA2) were expressed in terms of absolute and relative power and percentage of maximal aerobic power. In the sedentary group, T(RSA1) and T(RSA2) were 150.3 (18.7) W and 198.3 (28.8) W, respectively, whereas in the athlete group T(RSA1) and T(RSA2) were 247.3 (32.8) W and 316.0 (28.8) W, respectively. CONCLUSIONS: Dynamic analysis of f(RSA) provides a useful tool for identifying ventilatory thresholds during graded and maximal exercise test in sedentary subjects and athletes.