Fast kinetics of muscle oxygen consumption (VO2) is characteristic of effective physiological systems integration. The mechanism of VO2 kinetic control in vivo is equivocal as measurements are complicated by the twin difficulties of making high-frequency direct measurements of VO2 and intramuscular metabolites, and in attaining high [ADP]; complexities that can be overcome utilising highly aerobic canine muscle for the investigation of the transition from rest to contractions at maximal VO2. Isometric tetanic contractions of the gastrocnemius complex of six anaesthetised, ventilated dogs were elicited via sciatic nerve stimulation (50 Hz; 200 ms duration; 1 contraction s(−1)). Muscle VO2 and lactate efflux were determined from direct Fick measurements. Muscle biopsies were taken at rest and every ∼10 s during the transient and analysed for [phosphates], [lactate] and pH. The temporal VO2 vs. [PCr] and [ADP] relationships were not well fitted by linear or classical hyperbolic models (respectively), due to the high sensitivity of VO2 to metabolic perturbations early in the transient. The time course of this apparent sensitisation was closely aligned to that of ATP turnover, which was lower in the first ∼25 s of contractions compared to the steady state. These findings provide the first direct measurements of skeletal muscle VO2 and [PCr] in the non-steady state, and suggest that simple phosphate feedback models (which are adequate for steady-state observations in vitro) are not sufficient to explain the dynamic control of VO2 in situ. Rather an allosteric or 'parallel activation' mechanism of energy consuming and producing processes is required to explain the kinetic control of VO2 in mammalian skeletal muscle.