The aim of this thesis was to investigate the extent of peripheral and central components in the development of fatigue during maximal exercise. Fatigue during maximal and supramaximal exercise has traditionally been modelled from the peripheral context of an inadequate capacity to supply metabolic substrate to the contracting muscles to meet the increased energy demand. However, there are a number of observations that are not compatible with the peripheral fatigue model but which support a reduced central drive during exercise acting to prevent organ failure that might occur should the work be continued at the same intensity. Candidates for the role of “exercise stopper” have been identified as mechanical forces, teleoanticipation, cardiovascular capacity and dyspnoea. We explored these various possibilities in order to determine the most likely cause of exercise cessation during high intensity exercise.The development of a plateau in oxygen consumption during maximal incremental exercise has traditionally been used as evidence that an oxygen deficiency in the exercising muscles causes the termination of exercise. However, the incidence of this “plateau phenomenon” depends largely on mode of exercise, testing protocol and sampling frequency. The aim of this study was to examine whether the development of the “plateau phenomenon” is an artefact of pedalling cadence. In the first study nine healthy individuals performed in random order a maximal incremental ramp test (0.5 W.s-1) on four occasions at a fixed cadence of 60, 80 or 100 rpm and at a self-selected cadence. Oxygen consumption (VO2), CO2 production (VCO2), minute ventilation (VE) and heart rate were measured throughout each trial and averaged over 30 s. Cadence was recorded every second. Neither VO2max nor peak power output were different between trials. Submaximum VO2, VCO2 and VE were not influenced by cadence. A plateau in oxygen consumption was observed in 14% of the trials. Cadence declined significantly towards the end of the self-selected cadence trial (p < 0.05). This ramp protocol produces a low incidence of the “plateau phenomenon” and the measured physiological variables are unaffected by cadence. Furthermore, only one subject displayed this phenomenon on more than one occasion. This confirms that the “plateau phenomenon” is an artefact of the testing protocol. The significant fall in cadence in anticipation of exercise termination during the self-selected cadence trial indicates the presence of a neural regulation, which would lead to a “plateau phenomenon” in those cycle tests in which the work rate is cadence-dependent.The purpose of the second study was to assess whether pacing strategies are adopted during supramaximal exercise bouts lasting longer than 30 s. Eight healthy males performed six Wingate Anaerobic Tests (WAnT). Subjects were informed that they were performing four 30 s WAnT and a 33 s and 36 s WAnT. However, they actually completed two trials of 30, 33 and 36 s each. Temporal feedback in the deception trials was manipulated so that subjects were unaware of the time discrepancy. Power output (PO) was determined from the angular displacement of the flywheel and averaged over 3 s. The peak power (PPI), mean power (MPI) and fatigue (FI) indices were calculated for each trial. Power output was similar for all trials up to 30 s. However, at 36 s the PO was significantly lower in the 36 s deception trial compared to the 36 s informed trial (392 ± 32 W vs 470 ± 88 W) (p < 0.001). The MPI was significantly lower in the 36 s trials (714 ± 76 W and 713 ± 78 W) compared to the 30 s trials (745 ± 65 W and 764 ± 82 W) although they were not different at 30 s (764 ± 83 W and 755 ± 79 W). The significant reduction in FI was greatest in the 36 s deception trial. In conclusion, the significant reduction in PO in the last six seconds of the 36 second deception trial, but not in the 36 second informed trial, indicates the presence of a pre-programmed 30 second “end point” based on the anticipated exercise duration from previous experience. Furthermore the similarity in pacing strategy in all informed trials suggests that the pacing strategy is centrally regulated and is independent of the total work to be performed.Athletes adopt a pacing strategy to delay fatigue and optimise athletic performance. However, many current theories of the regulation of muscle function during exercise do not adequately explain all observed features of such pacing strategies. We studied power output, oxygen consumption and muscle recruitment strategies during successive 4km cycling time trials to determine whether alterations in muscle recruitment by the central nervous system could explain the observed pacing strategies. Seven, highly trained cyclists performed three consecutive 4 km time trial intervals, each separated by 17 minutes. Subjects were instructed to perform each trial in the fastest time possible, but were given no feedback other than distance covered. Integrated electromyographic (iEMG) readings were measured at peak power output and for 90 s before the end of each trial. Subjects reach a VO2max in each interval. Time taken to complete the first and third intervals was similar. Peak power output was highest in the first interval but average power output, oxygen consumption, heart rate and postexercise plasma lactate concentrations were not different between intervals. Power output and iEMG activity rose similarly during the final 60 s in all intervals but were not different between trials. The similar pacing strategies in successive intervals and the parallel increase in iEMG and power output towards the end of each interval suggests that these pacing strategies could not have been controlled by peripheral mechanisms. Rather, these findings are compatible with the action of a centrally regulated that are recruited and de-recruited during exercise. The extent to which peripheral feedback influences recruitment patterns could not be determined from these experiments.The fourth study examined whether the supplementation of inspired air with a hyperoxic mixture results in a dose-dependent increase in peak work rate and maximal oxygen consumption (VO2max) during a ramp test to volitional exhaustion. To avoid the methodological disadvantages associated with breathing the gas mixtures from mixing bags, the trials were performed in a sealed chamber in which the oxygen fraction (FIO2) in the ambient air was altered and subjects were able to inhale directly from the environment. The three oxygen fractions in which the subjects exercised were 21% (room air), 35 or 60%. Arterial blood sampling occurred at rest and every 3 min during the trial. The blood was analysed for the partial pressure of oxygen (pO2), and carbon dioxide (pCO2); pH; oxygen saturation (sO2); haemoglobin saturation (O2Hb); and lactate concentrations. Expired gas and heart rate were measured continuously. Arterial sO2 and O2Hb were elevated in both hyperoxic conditions and did not fall throughout either trial. However in the normoxic trial sO2 and O2Hb declined over the duration of the trial. Lactate concentrations and pH were similar between all trials. VO2max was significantly higher with an FIO2 of 35 and 60% but was not different between hyperoxic conditions. Maximal ventilation (VEmax), carbon dioxide production (VCO2max) and heart rate were similar for all trials. Peak power output was increased in the trained athletes in the 60% FIO2 trial. Since the plateau phenomenon occurred infrequently in all trial (~9%) and the effect of hyperoxia on performance was less than the changes in blood oxygen carrying capacity, we conclude that hyperoxia improved exercise performance not solely by increasing oxygen delivery to the exercising muscles.In order to be able to directly compare the results from studies using different equipment it is important to know the interchangeability of the results from the machines. The fifth study tested the reliability and interchangeability of the two automated metabolic gas analyser systems that would be used in this series of studies at a range of submaximal workloads. Eight highly trained cyclists performed two incremental submaximal cycle ergometer tests. For each session either a Schiller CS-200 or a Vmax Series 229 automated gas analyser was used for expired gas analysis. Data for oxygen consumption (VO2), CO2 production (VCO2), minute ventilation (VE) and respiratory exchange ration (RER) were averaged for each of the five stages (200, 250, 275, 300 and 325 W). The VO2, VE and RER were similar between trials at all workloads. However, VCO2 was significantly lower in the Schiller trial at workloads above 200 W (p < 0.05). Although there was a significant correlation between the two automated systems for the measured parameters (VO2 = 0.78; VCO2 = 0.80; VE = 0.82; RER = 0.72) (p < 0.05), a Bland-Altman plot revealed that the limits of agreement between the two systems were unacceptably large (VO2 = 0.53 to 1.30 L.min-1; VCO2 = 0.55 to 0.64 L.min-1; VE = -22.3 to 30.3 L.min-1; RER = - 0.03 to 0.13). The co-efficient of variation within the analysers was insignificant for both systems. Both the systems provide reliable measures of expired gas parameters. However, care should be taken in directly comparing studies that have used the two different systems due to the poor agreement between the systems.The factors causing the termination of maximal exercise at sea level are unknown. A widely held view is that skeletal muscle anaerobiosis consequent to an inadequate oxygen delivery to the exercising muscles limits exercise. However, there is also evidence that respiratory muscle fatigue at the high ventilatory volumes achieved during maximal exercise delivery and respiratory muscle work on maximal exercise performance, we exercised 8 highly trained cyclists in a pressure-sealed chamber in which O2 concentrations were manipulated and helium (He) was substituted for nitrogen in the ambient air in order to reduce the work of breathing during exercise. This system ensured that external inspiratory and expiratory resistance was minimised and identical in all experimental conditions and approximated conditions present during usual exercise. During trials with O2 enriched ambient air the peak work rate increased (451 ± 58 W vs. 429 ± 59 W). Neither maximum nor submaximal oxygen consumption was altered in FIO2 of 35% (5.0 ± 0.6 l.min-1) compared to 21% (4.9 ± 0.7 l.min- 1). Substituting helium for nitrogen had no additional effect on work (453 ± 56 W) or VO2max (4.9 ± 0.7 l.min-1) beyond those observed for the hyperoxic conditions. Although submaximum VE was reduced with helium, VEmax was unchanged. Since exercise was terminated at the same peak work rate (± 5 W) in the two hyperoxic conditions we postulate that the actual work rate may be the sensed variable that determines maximal exercise performance. The findings from these studies suggest that the maintenance of physiological homeostasis and the avoidance of organ and cellular damage are of fundamental importance during maximal exercise. This is achieved through central regulation of work output based, possibly, on afferent information from the mechanoreceptors in the exercising skeletal muscles or alternatively, the extent of motor unit recruitment during maximal exercise may be hardwired in the central nervous system in a system of feed-forward control.