To define predictive behavior and mechanisms in visual smooth pursuit, various target motions were presented to 2 monkeys. Target stimuli included: single sinusoids (1's), triangle waves (T's), sums of 4 nonharmonically related sinusoids (4's), bandpass limited white noise (B's), and wideband white noise (N's). Velocity error was least for 1's, greatest for N's, and intermediate for T's, 4's, and B's. For the bandlimited 4's and B's, monkeys demonstrated the greatest relative amplitude response at the highest frequencies. Predictive mechanisms are classified as short- and long-term, depending on how much past target motion information is employed. The T's and a modification of this stimulus pattern involving a random perturbation were used to test the hypothesis that prediction is based exclusively on short-term signal processing related to target position and its derivatives. The existence of long-term predictive mechanisms in monkey smooth pursuit was unequivocally demonstrated with the use of the latter stimulus.