Fig 1.
Acceleration and deceleration period commands exhibit some of the properties of the fast and slow system.
A. Experiment 1 trial structure. Following a vertical primary saccade, subjects experienced a rightward horizontal endpoint error. B. To quantify adaptation, we measured the horizontal displacement of the adapted saccade with respect to baseline saccades. We integrated horizontal eye velocity over the acceleration and deceleration periods to determine the displacement during each period. C. Training blocks. Each block consisted of four periods: initial adaptation, set break (brown line), re-adaptation, error clamp. D. Horizontal displacement during the acceleration period was smaller than during the deceleration period, but exhibited little or no forgetting at set breaks. E. Trial-to-trial rate of change in the horizontal displacement produced by the acceleration and deceleration period commands during the perturbation blocks. Rate of change was faster in the deceleration period. F. Effect of set break on the horizontal displacement in the perturbation blocks and error clamp blocks. At left we show adaptation aligned to set breaks and averaged across all periods. Bin size is 4 trials. Decay was much greater in deceleration period commands. At right, we calculated the percent loss during set breaks in the perturbation and error clamp periods (depicted at left). G. Decay in the acceleration and deceleration period commands during the error clamp period. Error bars are between subject SEM. Bin size is 4 trials.
Fig 2.
In Experiment 1, adapted saccades arrive at an endpoint with a trajectory that is quite different from control saccades that arrive at the same endpoint.
A. Horizontal velocity of control saccades and adapted saccades. In control trials, targets are presented at (+1o,15o), (+2o,15o), and (+3o,15o). In adapted saccades, the same endpoint is achieved but through learning. Adapted saccades exhibit a smaller horizontal displacement during the acceleration period, but a larger displacement during the deceleration period (with respect to baseline). B. Control saccade have roughly equal horizontal displacement for the acceleration and deceleration period commands. In comparison, an adapted saccade has most of its horizontal displacement due to the deceleration period commands. Error bars are between subject SEM.
Fig 3.
Both acceleration and deceleration period commands exhibit spontaneous recovery. A. Experimental design. Perturbation block was followed by an extinction block, and then a block of error clamp trials. Half of the subjects experienced horizontal errors following vertical saccades (left panel), while the other half experienced vertical errors following horizontal saccades (right panel). B. Measured data. C. A standard two-state model was fit to the measured data (trace labeled total). The fast and slow states that are predicted by the model are shown. D. Eye velocity in the direction of error during the trials at end of adaptation (t1), at end of extinction (t2), and during error clamp (t3). E. Spontaneous recovery, defined as [H(t3)−H(t2)]/H(t1).
Fig 4.
A. Model 1 represented a single controller that learned from error via a fast state xf and a slow state xs. During a saccade, a fraction of the sum of the two states was expressed during acceleration, and the remainder was expressed during deceleration. Arrow indicates the effect of set break on the model’s acceleration period commands. B. Model 2 relied on two controllers, one that aimed the saccade with state xaim, and another that independently affected the commands during the acceleration and deceleration periods. C. States of Model 2 during Exps. 1 and 2. Model 2 could account for the observation that set breaks produced a loss in the deceleration period but not the acceleration period commands. In addition, Model 2 accounted for the observation that decay following a set break differed in perturbation vs. error clamp trials: note the difference in xDec at set breaks during perturbation and error clamp blocks of Exp. 1). D. Measured set break decay data. Adaptation is shown aligned to set breaks during perturbation periods in Exps. 1 and 2 (left) and error clamp periods in Exp. 1 (right). E. Comparison of the set break effects in the models and the measured data. The models differed in their ability to account for acceleration period set break decay. Error bars for data are SEM. Participants from Exps. 1 and 2 were combined when analyzing the perturbation block set breaks. For error clamp block set breaks, only participants from Experiment 1 were used.
Table 1.
Parameter values and confidence intervals for Models 1 and 2.
Fig 5.
Following experience of a single error, correction of the subsequent saccade was primarily during deceleration. A. Subjects made vertical saccades and occasionally experienced a horizontal error. B. Average horizontal and vertical velocity of the saccade for the target along the vertical axis. C. Trial-to-trial change in vertical and horizontal velocity following experience of a positive or negative horizontal endpoint error. Vertical line indicates time of peak speed, thus separating the acceleration and deceleration periods. D. Trial-to-trial change in saccade trajectory following experience of a positive or negative horizontal endpoint error. The endpoint correction was primarily due to commands in the deceleration period. Note the large difference in the scales of the x and y axes. E. Onset time of correction in the horizontal velocity following experience of a horizontal endpoint error. Mean onset time is about 3 ms following start of the deceleration period. F. Data from Experiments 1 and 2, measuring the response following the experience of error on the very first trial (velocity on trial 2 minus trial 1, in the direction of the perturbation). Shaded error regions are between subject SEM.