We examined how the relationship between neuronal activity and behavior evolved over time during a motion-detection task. Recording from two regions of visual cortex that process motion, the middle temporal (MT) and ventral intraparietal (VIP) areas, we used the time it took subjects to detect a motion stimulus to evaluate the dynamics of the underlying neuronal signals. Single-neuron activity was correlated with stimulus detection and reaction time (RT) in both areas. The rising edge of the population response from both areas was highly predictive of RT using a simple threshold-detection model. The time course of the population responses, however, differed between MT and VIP. For MT, the onset of the neuronal response was relatively constant, whereas for VIP the onset of the neuronal responses increased with RT. In contrast to previous studies, we found that single neurons were not reliable detectors of the motion signal when constrained by realistic detection times.