(A) Colored arrows show the target position trajectory in two dimensions during learning trials. Different colors show learning trials with instruction speeds of 5, 10, 15, 20, 25, or 30 deg/s. Black arrow shows the subsequent probe trial (learning speed 0 deg/s). Pursuit and instruction directions were randomized from trial-to-trial (see Materials and methods). (B) Average eye velocity in the direction of the previous instruction (learning direction) as a function of time during probe trials averaged across experiments from Monkeys RE (left) and YO (right). Different colors indicate average responses for the different error magnitudes using the same color scheme as in (A). The vertical dashed line shows the time of the instruction in the preceding learning trial and the gray shaded region shows the interval used for quantitative analysis, 225 to 275 ms after the onset of pursuit target motion. (C) The amplitude of single-trial learning as a function of the error magnitude created by the previous instruction. Dashed lines show best-fit sigmoid for each monkey. (D) Example ‘error-clamp’ trial used to minimize the retinal image motion in the learning direction, allowing characterization of motor memory retention in the absence of error. Red and blue traces show eye velocity and the black dashed traces show target velocity. (E) Trial-course of forgetting after 20 learning trials, probed under error-clamp conditions for Monkeys RE (left) and YO (right). Data are normalized to the magnitude of behavioral learning in the last learning trial (trial 0). Black curves show best fitting exponentials. Error bands and error bars show ± SEM across experiments for each monkey.