In exercising muscle, creatine kinase ensures that mismatch between ATP supply and ATP use results in net phosphocreatine (PCr) splitting. This,inter alia, makes31P magnetic resonance spectroscopy a useful tool for studying muscle ‘energy metabolism’ noninvasivelyin vivo. We combined this with near–infrared spectroscopy (NIRS) to study ATP synthesis and oxygenation in calf muscle of normal subjects and patients with peripheral vascular disease. Experimental and clinical details and basic data have been published elsewhere (G.J. Kemp et al.,Journal of Vascular Surgery34 (2001), 1103–10); we here propose an analysis of interactions between metabolic ‘error signals’ and cellular PO2(estimated from NIRS changes, provisionally assumed to reflect deoxymyoglobin). Post–exercise PCr recovery is monoexponential, and the linear relationship between PCr resynthesis rate (= oxidative ATP synthesis) and the perturbation in PCr (conceptually the simplest error signal) is consistent with negative feedback. In patients the inferred ‘mitochondrial capacity’ (= oxidative ATP synthesis at ‘zero’ PCr) is decreased by 53±6%, leading to reduced oxidative ATP contribution in exercise, because of increased deoxygenation. Increased PCr perturbation partially outweighs cellular hypoxia, but as low cellular PO2is required for capillary–mitochondrion O2diffusion, rate–signal relationships may overstate maximum oxidative ATP synthesis rate.