We show how quantitative information about proton handling in human skeletal muscle in exercise and recovery can be obtained by 31P MRS and illustrate this with data from metabolic disorders. Proton production, proton efflux and passive buffering can be distinguished by comparing changes in [phosphocreatine] and pH at the end of exercise and by calculating ATP turnover during ischaemic exercise and in the 'natural experiment' of myophosphorylase deficiency (McArdle's disease). We calculate the effective buffer capacity to be 20-30 mmol/L/pH unit (slykes), somewhat lower than published measurements made in vitro but similar to other values obtained in vivo. This analysis is applied to data from normal muscle and a variety of disease states to estimate proton efflux during recovery and ATP production during exercise: (i) proton efflux during recovery is pH-dependent, reaching a 10 mmol/L/min at pH 6.2, and is increased in some cases of mitochondrial myopathy and in hypertension; (ii) glycogenolytic ATP production during exercise can reach 25 mmol/L/min in normal muscle and correlates approximately with [Pi] at the start of aerobic exercise and throughout ischaemic exercise; (iii) oxidative ATP production can reach 20-25 mmol/L/min and (as during recovery) correlates approximately with [Pi].