The macaque middle temporal area (MT) is exquisitely sensitive to visual motion and there is a large amount of evidence that neural activity in MT is tightly correlated with the perception of motion. The mechanisms by which MT neurons achieve their directional selectivity, however, have received considerably less attention. We investigated the motion-energy model as a description of motion mechanisms in macaque MT. We first confirmed one of the predictions of the motion-energy model; macaques-just like humans-perceive a reversed direction of motion when a stimulus reverses contrast with every displacement (reverse-phi). This reversal of perceived direction had a clear correlate in the neural responses of MT cells, which were predictive of the monkey's behavioral decisions. Second, we investigated how multiple motion-energy components are combined. Psychophysical data have been used to argue that motion-energy components representing opposite directions are subtracted from each other. Our data show, however, that the interactions among motion-energy components are more complex. In particular, we found that the influence of a given component on the response to a stimulus consisting of multiple components depends on factors other than the response to that component alone. This suggests that there are subthreshold nonlinear interactions among multiple motion-energy components; these could take place within MT or in earlier stages of the motion network such as V1. We propose a model that captures the complexity of these component interactions by means of a competitive interaction among the components. This provides a better description of the MT responses than the subtractive motion opponency envisaged in the motion-energy model, even when the latter is combined with a gain-control mechanism. The competitive interaction extends the dynamic range of the cells and allows them to provide information on more subtle changes in motion patterns, including changes that are not purely directional.