In the Olympics 2008 Special Issue, Amann and Dempsey (Amann & Dempsey, 2008) reported a study on the effects of pre-fatigue trials at two different intensities on peripheral locomotor muscle fatigue and performance during subsequent 5 km time trials. According to the authors, the data presented support their theoretical model that central neural drive to the locomotor muscles (and thus performance) during high-intensity endurance exercise in normoxia is limited to a significant extent by inhibitory afferent feedback related to metabolic stress in the locomotor muscles. This negative feedback regulation of endurance performance ensures that peripheral fatigue does not trespass a critical threshold associated with potential locomotor muscle damage and pain. Amann and Dempsey's concept of peripheral fatigue as a variable carefully regulated by the CNS (Amann & Dempsey, 2007) is extremely interesting because it challenges the traditional view that endurance performance is primarily and directly limited by peripheral locomotor muscle fatigue. However, in our opinion, there are three major concerns with their interpretation of the data. Noakes and Marino have already argued that the significant increase in power output measured at the end of time trials is incompatible with a negative feedback system based on inhibitory sensory inputs from fatigued locomotor muscles (Noakes & Marino, 2007). We argue that such a negative feedback system is also incompatible with the normal levels of power output measured at the beginning of the 5 km time trials preceded by the pre-fatigue trials. Indeed, it is not clear why inhibitory sensory inputs from fatigued locomotor muscles should reduce central motor drive and power output (i.e. induce central fatigue) only in the middle part of the time trial. The other potential mediators of central fatigue proposed by Amann and Dempsey (disturbances in brain neurotransmitters, depletion of brain glycogen, increases in core and brain temperature, and inhibitory supraspinal reflexes originating from fatigued respiratory muscles) should also reduce power output from the very beginning of the 5 km time trials preceded by the pre-fatigue trials, and restrain the CNS from greatly increasing power output at the end of all 5 km time trials. The concept of central fatigue also implies that reduced central neural drive to the locomotor muscles is a direct cause of reduced power output (Gandevia, 2001) during the 5 km time trials. However, competitive cyclists maintain a high and relatively stable pedalling cadence during time trials (Lucia et al. 2001). Therefore, we assume that the statistically significant reduction in vastus lateralis EMG amplitude (which the authors use as an index of central motor drive) measured during the 5 km time trial preceded by the exhaustive pre-fatigue trial is the effect of the conscious decision to shift into a lower gear with consequent reduction in force required by each pedal stroke. Such choice (which, clearly, cannot be subconscious) and the voluntary act to change gear fall in the realm of behavioural regulation of homeostasis (Woods & Ramsay, 2007), not fatigue defined as an exercise-induced reduction in the force or power-generating capacity of the neuromuscular system (Amann & Calbet, 2007). Although both processes reside in the brain, there is a fundamental difference between voluntary changes in central neural drive to skeletal muscles (voluntary behaviour) (Carlson, 2004) and a subconscious reduction in central motor drive during maximal voluntary contractions (central or, more precisely, supraspinal fatigue) (Gandevia, 2001). Therefore, in our opinion, the two concepts should not be mixed up. Indeed, the very concept of central fatigue during submaximal voluntary contractions like the ones required by cycling has been questioned (Presland et al. 2005; Taylor & Gandevia, 2007). Furthermore, behavioural regulation of homeostasis, but not central fatigue, can generate complex and often anticipatory changes in behaviour that go beyond negative feedback regulation (Woods & Ramsay, 2007) such as the end-spurt in power output observed by Amann and Dempsey during the 5 km time trials. As in other forms of homeostatic behaviour (e.g. drinking, eating and putting the coat on), voluntary actions beneficial to homeostasis are motivated by conscious sensations such as thirst, hunger and coldness (Denton, 2006). In Amann and Dempsey's model, the conscious sensation motivating the voluntary reduction in endurance performance (the so-called ‘sensory tolerance limit’; Gandevia, 2001) is the leg discomfort caused by stimulation of muscle nociceptors by fatigue-related metabolites, i.e. exercise-induced muscle pain. At first glance, such a proposition is highly attractive because of anecdotal evidence that muscle pain limits endurance performance (Cook, 2006), and because the afferent pathways linking peripheral nociceptors to various subcortical and cortical structures are well characterized (Almeida et al. 2004). However, the results of several scientific studies suggest that most people can tolerate the muscle pain associated with intense cycling, and that exercise tolerance is limited by the conscious sensations of leg effort and dyspnoea, the two main components of perception of effort/fatigue experienced during dynamic whole-body exercise (Borg, 1998; Jones & Killian, 2000; Presland et al. 2005; Cook, 2006; Marcora et al. 2008). Furthermore, the uncomfortable sensations of intense leg and respiratory effort are primarily efferent perceptions generated by the corollary discharges of central neural drive to the locomotor and respiratory muscles (Jones & Killian, 2000) with little or no influence from peripheral afferent feedback. Indeed, in a recently published study, we demonstrated that reduced locomotor muscle force increases the ratings of perceived exertion (RPE) during intense cycling exercise independently of metabolic stress (Marcora et al. 2008). Furthermore, spinal blockade of sensory inputs from locomotor muscles using epidural anaesthesia has no effect or may even increase RPE during cycling exercise because of its side-effects on motoneurons (Kjaer et al. 1999; Smith et al. 2003). Therefore, we predict that the epidural studies proposed by Amann and Dempsey in their paper will result in either no change or a decrease in endurance performance. These predictions are contrary to the hypotheses derived from their theoretical model because blockade of inhibitory afferent feedback from metabolically stressed locomotor muscles should allow the CNS to significantly increase power output during time trials and time to exhaustion during constant-power rides (Amann et al. 2008) with potentially dangerous consequences for the locomotor muscles. So if we argue that the negative feedback system proposed by Amann and Dempsey does not play a major role in limiting endurance performance during high-intensity exercise in normoxia, how can we then explain the ‘critical threshold’ of peripheral fatigue they measured in this and previous experiments (Amann & Dempsey, 2007)? In our opinion this ‘striking’ finding does not necessarily imply that the development of peripheral fatigue during endurance exercise is a variable carefully regulated by the CNS. It may simply be due to fatigability and the ‘unusual’ recruitment pattern of different fibre types during intense cycling. Indeed, fatigue at whole-muscle or muscle-group level is mainly due to selective metabolic fatigue of type IIx fibres with little contribution from the other two types of fatigue-resistant fibres (type IIa and type I) (Sargeant, 2007). Because all types of available muscle fibres are recruited from the very beginning despite submaximal force requirements (Greig et al. 1985; Altenburg et al. 2007), fatigue of type IIx fibres may occur early during intense cycling, and no further peripheral fatigue would occur regardless of duration and/or intensity of additional exercise because the remaining type IIa and type I fibres would not be greatly affected by metabolic stress (Sargeant, 2007). These neuromuscular processes may provide a simpler and biologically plausible alternative explanation to Amann and Dempsey's ‘astonishing’ finding that, at the end-exercise, the level of peripheral fatigue was identical between the three 5 km time trials independently of the level of pre-existing fatigue and/or marked differences in endurance performance.