Abstract
We assessed the dynamics of changes in the metrics of saccades in response to time-varying disturbances of visual feedback. Observers made naturally paced saccades in adapting blocks of 600 target steps with fixed amplitude of 8 degrees either in horizontal or unconstrained directions. During each saccade, the intra-saccadic shift (ISS)—following a sinusoidal variation as a function of the trial number—displaced the target along its vector by -25% (inward) to +25% (outward) of the presaccadic target eccentricity. Formal comparison of a range of descriptive models fitted to the oculomotor response averaged across participants revealed two clearly distinct additive components: (1) a sinusoidal variation with the same frequency, lagging the ISS by about 20 trials and scaled down in amplitude, and (2) a persistent drift in the baseline, systematically increasing hypometria despite the absence of a consistent negative visual error. We captured these dynamics using a generative model consisting of a simple state equation that predicts the size of the upcoming saccade with two contributions: a learning term in which a feedback coefficient (k) weights the visual error experienced on the previous trial, and a persistence term in which a 'retention' coefficient (c) defines what portion of the predicted size of the last executed movement is retained in the next. In this framework, the sinusoidal response, while resulting from a trial-by-trial error correction process well described in the literature on sensorimotor learning, may inform how the retention and feedback coefficients get tuned to match or mitigate the stimulus features. The drift would then be an outcome of this process that does not determine nor hamper the learning of the stimulus. Following these ideas, we discuss additional effects of a reduced retention coefficient. Our results show how periodic disturbances in the experienced saccadic error, provides new insights into learning in the oculomotor system.
Meeting abstract presented at VSS 2016