1. When a visual target is moved by the subject's hand (self-moved target tracking), smooth pursuit (SP) characteristics differ from eye-alone tracking: SP latency is shorter and maximal eye velocity is higher in self-moved target tracking than in eye-alone tracking. The aim of this study was to determine which signals (motor command and/or proprioception) generated during arm motion are responsible for the decreased time interval between arm and eye motion onsets in self-moved target tracking. 2. Six control subjects tracked a visual target whose motion was generated by active or passive movements of the observer's arm in order to determine the role played by arm proprioception in the arm-eye coordination. In a second experiment, the participation of two subjects suffering complete loss of proprioception allowed us to assess the contribution of arm motor command signals. 3. In control subjects, passive movement of the arm led to eye latencies significantly longer (130 ms) than when the arm was actively self-moved (-5 ms:negative values meaning that the eyes actually started to move before the target) but slightly shorter than in eye-alone tracking (150 ms). These observations indicate that active movement of the arm is necessary to trigger short-latency SP of self-moved targets. 4. Despite the lack of proprioceptive information about arm motion, the two deafferented subjects produced early SP (-8 ms on average) when they actively moved their arms. In this respect they did not differ from control subjects. Active control of the arm is thus sufficient to trigger short-latency SP. However, in contrast with control subjects, in deafferented subjects SP gain declined with increasing target motion frequency more rapidly in self-moved target tracking than in eye-alone tracking. 5. The deafferented subjects also tracked a self-moved target while the relationship between arm and target motions was altered either by introducing a delay between arm motion and target motion or by reversing target motion relative to arm motion. As with control subjects, delayed target motion did not affect SP latency. Furthermore, the deafferented subjects adapted to the reversed arm-target relationship faster than control subjects. 6. The results suggest that arm motor command is necessary for the eye-to-arm motion onset synchronization, because eye tracking of the passively moved arm was performed by control subjects with a latency comparable with that of eye-alone tracking of an external target. On the other hand, as evidenced by the data from the deafferented subjects, afferent information does not appear to be necessary for reducing the time between arm motion and SP onsets. However, afferent information appears to contribute to the parametric adjustment between arm motor command and visual information about arm motion.