The characteristics of the exercise hyperpnoea in humans are generally agreed upon. During moderate-intensity exercise (i.e. below the lactate threshold, θ_L), ventilation (V_E) correlates closely with pulmonary CO₂ output (VCO₂) resulting in stability of arterial PCO₂, PO₂ and pH. Thus, following an initial short ‘cardiodynamic’ (or phase 1) component, V_E increases with mono-exponential (or first-order) kinetics (phase 2) to the new steady state. However, at work rates associated with a metabolic acidaemia (i.e. above θ_L), V_E response kinetics become nonlinear and steady states may be unattainable. The development of respiratory compensation causes V_E to increase out of proportion to VCO₂, with the ensuing hypocapnia constraining the fall of arterial pH. Ventilatory control models have traditionally included proportional feedback (central medullary and peripheral carotid chemosensory) and feedforward (central and/or peripheral neurogenic) elements (e.g. Waldrop et al. 1996). However, the precise details of the control process remain unresolved, reflecting technical and interpretational challenges associated with isolating putative control mechanisms in intact humans, and also the challenges to linear control systems theory presented by multiple-input integration (reflective, in part, of the ventilatory and gas-exchange complexities encountered within supra-θ_L exercise intensities). The rapid V_E increase at exercise onset has been argued to be neurally mediated via muscle reflexogenic drives and/or ‘central command’ (e.g. Waldrop et al. 1996). More recently, cardio-circulatory mechanisms influenced by alterations in central circulatory pressures and intramuscular vascular tissue pressure and/or conductance have also been advocated (Haouzi et al. 2004). Control of V_E in phase 2 has traditionally been ascribed to chemosensory mechanisms, argued to provide a ‘fine tuning’ for arterial blood-gas and acid-base regulation. The carotid body chemoreceptors have been shown to exert an important modulating influence on the phase 2 V_E kinetics in humans, interestingly (from a control perspective) with maintained exponentiality of response. Some investigators, however, favour central neural mechanisms of short-term potentiation in phase 2 (e.g. Waldrop et al. 1996). The extent to which the central chemoreceptors are involved in the control process is uncertain, as cerebrospinal pH during moderate exercise is reported to remain reasonably stable. Above θ_L, the carotid chemoreceptors appear to be largely responsible for mediating respiratory compensation, although having surprisingly slow kinetics, with additional involvement of the central chemoreflex through a constraining influence on V_E mediated by the hypocapnia. Interestingly, the ventilatory control process during exercise behaves as if it has appreciable redundancy, with selective inactivation of any one of several putative mechanisms seeming to have little impact on the magnitude of the exercise hyperpnoea. Such observations have fuelled the formulation of innovative control schemes that reflect both spatial interactions and temporal interactions, such as (a) optimisation of humoral and respiratory-mechanical ventilatory ‘costs’ (Poon, 1983) and (b) memory or long-term potentiation (Mitchell & Babb, 2006). Ventilatory control can be challenged at very high work rates, as are encountered in elite endurance athletes (e.g. Dempsey et al. 2003). For example, the ventilatory demands to clear metabolically-produced CO₂ and to effect respiratory compensation for the metabolic acidaemia can become so great that they approach, or even exceed, the mechanical limits of the lungs and chest wall (an index of which is the maximum voluntary ventilation). In addition, the perfusion ‘costs’ of the high respiratory-muscle power generation can, at limiting levels, constrain perfusion of the locomotor muscles and thus predispose to earlier fatigue. Furthermore, arterial PO₂ and O₂ saturation can evidence a decrease (‘exercise-induced arterial desaturation’), consequent to the compromised V_E response and to pulmonary gas exchange inefficiencies (e.g. truncating of pulmonary capillary transit times; regional ventilation-to-perfusion mismatching). In conclusion, the challenge is therefore to discriminate between robust competing control models that not only integrate such structures within plausible physiological equivalents, but also account for both dynamic and steady-state system responses over a range of exercise intensities, recognising that respiratory system limits can over-ride the control at very high work rates.