For rehabilitative devices to restore functional movement to paralyzed individuals, user intent must be determined from signals that remain under voluntary control. Tracking eye movements is a natural way to learn about an intended reach target and, when combined with just a small set of electromyograms (EMGs) in a probabilistic mixture model, can reliably generate accurate trajectories even when the target information is uncertain. To experimentally assess the effectiveness of our algorithm in closed-loop control, we developed a robotic system to simulate a reaching neuroprosthetic. Incorporating target information by tracking subjects' gaze greatly improved performance when the set of EMGs was most limited. In addition we found that online performance was better than predicted by the offline accuracy of the training data. By enhancing the trajectory model with target information the decoder relied less on neural control signals, reducing the burden on the user.