The mechanisms controlling skeletal muscle oxygen consumption (V(o)₂) during exercise are not well understood. We determined whether first-order control could explain V(o)₂kinetics at contractions onset (V(o)₂(on)) and cessation (V(o)₂off)) in single skeletal muscle fibres differing in oxdidative capacity, and across stimulation intensities up to V(o)₂(max). Xenopus laevis fibres (n = 21) were suspended in a sealed chamber with a fast response P(o)₂ electrode to measure V(o)₂ every second before, during and after stimulated isometric contractions. A first-order model did not well characterize on-transient V(o)₂ kinetics. Including a time delay (TD) in the model provided a significantly improved characterization than a first-order fit without TD (F-ratio; P < 0.05), and revealed separate 'activation' and 'exponential' phases in 15/21 fibres contracting at V(o)₂(max) (mean ± SD TD: 14 ± 3s). On-transient kinetics (τV(o)₂(on)) was weakly and linearly related to V(o)₂(max) (R² = 0.271, P = 0.015). Off-transient kinetics, however, were first-order, and τV(o)₂(off) was greater in low-oxidative (V(o)₂max < 0.05 nmol mm⁻³s⁻¹ than high-oxidative fibres (V(o)₂(max > 0.10 nmol mm ⁻³ s⁻¹; 170 ± 70 vs. 29 ± 6 s, P < 0.001). 1/ τV(o)₂(off) was proportional to V(o)₂(max) (R² = 0.727, P < 0.001), unlike in the on-transient. The calculated oxygen deficit was larger (P < 0.05) than the post-contraction volume of consumed oxygen at all intensities except V(o)₂(max). These data show a clear dissociation between the kinetic control of V(o)₂at the onset and cessation of contractions and across stimulation intensities. More complex models are therefore required to understand the activation of mitochondrial respiration in skeletal muscle at the start of exercise.