The smooth pursuit system interacts with the vestibular system to maintain the accuracy of eye movements in space. To understand neural mechanisms of short-term modifications of the vestibulo-ocular reflex (VOR) induced by pursuit-vestibular interactions, we used a cross-axis procedure in trained monkeys. We showed earlier that pursuit training in the plane orthogonal to the rotation plane induces adaptive cross-axis VOR in complete darkness. To further study the properties of adaptive responses, we examined here the initial eye movements during tracking of a target while being rotated with a trapezoidal waveform (peak velocity 30 or 40 degrees/s). Subjects were head-stabilized Japanese monkeys that were rewarded for accurate pursuit. Whole body rotation was applied either in the yaw or pitch plane while presenting a target moving in-phase with the chair with the same trajectory but in the orthogonal plane. Eye movements induced by equivalent chair rotation with or without the target were examined before and after training. Before training, chair rotation alone resulted only in the collinear VOR, and smooth eye movement-tracking of orthogonal target motion during rotation had a normal smooth pursuit latency (ca 100 ms). With training, the latency of orthogonal smooth tracking eye movements shortened, and the mean latency after 1 h of training was 42 ms with a mean gain, at 100 ms after stimulus onset, of 0.4. The cross-axis VOR induced by chair rotation in complete darkness had identical latencies with the orthogonal smooth tracking eye movements, but its gains were <0.2. After cross-axis pursuit training, target movement alone without chair rotation induced smooth pursuit eye movements with latencies ca 100 ms. Pursuit training alone for 1 h using the same trajectory but without chair rotation did not result in any clear change in pursuit latency (ca 100 ms) or initial eye velocity. When a new target velocity was presented during identical chair rotation after training, eye velocity was correspondingly modulated by just 80 ms after rotation onset, which was shorter than the expected latency of pursuit (ca 100 ms). These results indicate that adaptive changes were induced in the smooth pursuit system by pursuit-vestibular interaction training. We suggest that this training facilitates the response of pursuit-related neurons in the cortical smooth pursuit pathways to vestibular inputs in the orthogonal plane, thus enabling smooth eye movements to be executed with shorter latencies and larger eye velocities than in normal smooth pursuit driven only by visual feedback.