1. In vivo ATP synthesis rates were measured in the human medial gastrocnemius muscle during high intensity exercise using localized 31P-magnetic resonance spectroscopy (31P-MRS). Six-second localized spectra were acquired during and following a 30 s maximal voluntary rate exercise using a magnetic resonance image-guided spectral localization technique. 2. During 30 s maximal voluntary rate exercise, ATPase fluxes were predominantly met by anaerobic ATP sources. Maximal in vivo glycogenolytic rates of 207 +/- 48 mM ATP min-1 were obtained within 15 s, decreasing to 72 +/- 34 mM ATP min-1 by the end of 30 s. In contrast, aerobic ATP synthesis rates achieved 85 +/- 2 % of their maximal capacity within 9 s and did not change throughout the exercise. The ratio of peak glycolytic ATP synthesis rate to maximal oxidative ATP synthesis was 2.9 +/- 0.9. 3. The non-Pi, non-CO2 buffer capacity was calculated to be 27.0 +/- 6. 2 slykes (millimoles acid added per unit change in pH). At the cessation of exercise, Pi, phosphomonoesters and CO2 were predicted to account for 17.2 +/- 1.5, 5.57 +/- 0.97 and 2.24 +/- 0.34 slykes of the total buffer capacity. 4. Over the approximately linear range of intracellular pH recovery following the post-exercise acidification, pHi recovered at a rate of 0.19 +/- 0.03 pH units min-1. Proton transport capacity was determined to be 16.4 +/- 4.1 mM (pH unit)-1 min-1 and corresponded to a maximal proton efflux rate of 15.3 +/- 2.7 mM min-1. 5. These data support the observation that glycogenolytic and glycolytic rates are elevated in vivo in the presence of elevated Pi levels. The data do not support the hypothesis that glycogenolysis follows Michealis-Menten kinetics with an apparent Km for [Pi] in vivo. 6. In vivo -measured ATP utilization rates and the initial dependence on PCr and glycolysis were similar to those previously reported in in situ studies involving short duration, high intensity exercise. This experimental approach presents a non-invasive, quantitative measure of peak glycolytic rates in human skeletal muscle.