A number of complex frameworks have been developed for considering the metabolic energy supply and various energy demands during cycling. None of these, however, adequately account for the cost of moving the limbs against inertial and gravitational forces, i.e., the internal mechanical work. Previously, internal mechanical work, or its time derivative, internal mechanical power (IP), has been determined using either a physiological approach or a biomechanical approach. Both approaches have their advantages and disadvantages. The physiological approach is valuable because it quantifies the energy state of the entire system where changes to whole-body energy metabolism with changes in energy demands are relatively incontrovertible. This approach, however, determines IP by subtracting external mechanical power (EP), the power applied externally on the pedals, and the metabolic cost of rest from the total energy expenditure. This simple relationship between metabolic and mechanical powers has been rebuked by biomechanists, who have instead argued that there is some degree of energy transfer from IP to EP, i.e., they do not have separate metabolic costs. A notable advantage of the biomechanical perspective is that it can determine the potential and kinetic energies of the separate body segments and can, with the use of inverse dynamics, largely distinguish the muscle groups from which the energy was sourced. The greatest limitation, however, is that the biomechanical perspective does not account for the dissipation of energy to heat, a significant destination of metabolic energy in the human body. This thesis has adopted a multidisciplinary approach to include a collection of investigations that aim to incorporate a term for IP into an existing framework for delineating the demands for energy during cycling. The first study demonstrated the additional information that can be obtained from current incremental protocols in cycling. This included a physiological estimate of IP and the amount of energy transferred from the downstroke leg to the contralateral limb via the cranks and bottom bracket, which has implications for the biomechanical estimate of internal mechanical power. A series of physiological models for estimating IP were evaluated in the second investigation. It was demonstrated that the IP for the same cadence calculated by two already established models was more than 5 W different, and therefore exceeding the acceptable limit of agreement, at 50-55, 80-85 and 110-115 rev · min-1. Three studies were then conducted to determine the effects of 1) posture during measurement of resting metabolic rate, 2) the methods for determining delta efficiency, and 3) the inclusion of the metabolic cost of 0 W (unloaded) pedalling, on the physiological IP calculation. The results from these studies were used to further refine the physiological model for estimating IP in the subsequent investigations of the thesis. Changes to IP, as a result of changes to cadence, were shown to affect the temporal expression of EP in the sixth study. In the seventh study, it was proposed that both the physiological and biomechanical models be applied to determine a range of values for IP, within which the real value of the energetic cost of IP would lie. Bearing in mind the assumptions and limitations of the respective models, such a range allows for the description of energy during cycling that considers metabolic and mechanical parameters together. At the upper limit of this range is the estimation of IP calculated using a physiological model in which IP and EP are considered independent and without the capacity to transfer energy between one another. Such an estimation yields the maximum possible cost for IP. At the lower limit of this range is the estimation of the cost of IP calculated using a biomechanical model, specifically where complete inter-compensation of muscle power among biarticular muscles and no antagonistic muscle actions occur. The range for IP at 80 rev · min-1 for the group of highly-trained cyclists was 28 to 56 J · s-1 and at 110 rev · min-1 was 34 to 96 J · s-1. These ranges from the seventh study were incorporated into a previously established model for predicting cycling performance to evaluate the contribution of IP in predicting energy expenditure (E˙) in cycling outdoors on level ground. The difference in predicted E˙ using the established model and measured E˙ in cyclists riding on the duck-board of an outdoor, concrete velodrome, could be accounted for by the energy expended to move the limbs, i.e., the IP. The results of this thesis demonstrated the practical importance of calculating IP, enabling more accurate predictions of performance. Moreover, if it were possible to reduce the energy required for this component of mechanical power, which does not directly contribute to forward motion, the energy available to meet the demands that do result in forward movement would increase. Several methods for decreasing IP have been suggested previously, including the use of non-circular chainrings, which provide opportunities for future research. The outcome of decreasing the energetic cost of IP would be an increase in a cyclist’s velocity and ultimately an improvement in performance.