Abstract
Ocular following responses (OFR) are reflexive eye movements triggered in response to a brief coherent motion of a large area within the visual field. Initial phase of OFR reflects many properties of low-level motion processing. Herein, we probed speed estimation computations by presenting stimuli either from a set of 15 luminance noise stimuli spanning a range of spatiotemporal frequencies (Motion Clouds, MC) or a set of nine superimposed triplets of these components, defining Compound Motion Clouds (CMCs). Our rationale was to use the responses from MCs to predict OFR to MCMs, exploring a range of alternative combination rules. We recorded eye movements to MC/CMCs in 6 participants using a 1000 EyeLink video eyetracker. Volunteers were required to make a 10° centering saccade over a grey background. Large motion patterns (20° diameter) were presented for 400ms at the end of the saccade. Response latency was about 90ms, independent upon MC/CMC properties. We measured likelihoods of eye velocities at different points in time. Early eye velocities ([90–150ms] after motion onset) were best predicted by a linear superposition of responses to MCs (i.e. vector averaging). Beyond 200ms after stimulus onset, responses to CMCs were best predicted by an interaction model, where the contribution of different MCs unveils a pattern of inhibitory and excitatory interactions between the different channels. Similar to perception (Gekas et al Curr Biol 2017), we explored the shape of these spatiotemporal interactions, particularly looking at how information is combined across the orthogonal speed and scale axes. Inhibition of faster speeds was consistently seen, acting as a slow speed prior. Along the scale axis, there was a broad central excitatory pooling beyond which an inhibition whose pattern was more subtle, showing individual differences. We discuss how dynamic implementation of such inhibition influences OFR strength and variability.
Acknowledgement: Agence Nationale de la Recherche (France), Project SPEED