A computer model of oxidative phosphorylation (OXPHOS) in skeletal muscle is used to compare state 3, intermediate state and state 4 in mitochondria with rest and work in skeletal muscle. 'Idealized' state 4 and 3 in relation to various 'experimental' states 4 and 3 are defined. Theoretical simulations show, in accordance with experimental data, that oxygen consumption (V'O2), ADP and Pi are higher, while ATP/ADP and Δp are lower in rest than in state 4, because of the presence of basal ATP consuming reactions in the former. It is postulated that moderate and intensive work in skeletal muscle is very different from state 3 in isolated mitochondria. V'O2, ATP/ADP, Δp and the control of ATP usage over V'O2 are much higher, while ADP and Pi are much lower in the former. The slope of the phenomenological V'O2-ADP relationship is much steeper during the rest-work transition than during the state 4-state 3 transition. The work state in intact muscle is much more similar to intermediate state than to state 3 in isolated mitochondria in terms of ADP, ATP/ADP, Δp and metabolic control pattern, but not in terms of V'O2. The huge differences between intact muscle and isolated mitochondria are proposed to be caused by the presence of the each-step activation (ESA) mechanism of the regulation of OXPHOS in intact skeletal muscle. Generally, the present study suggests that isolated mitochondria (at least in the absence of Ca2+) cannot serve as a good model of OXPHOS regulation in intact skeletal muscle.