Abstract
Saccadic and smooth pursuit eye movements are used to bring objects onto our foveae and to keep them there. Both have been intensively studied, but mainly in isolation and with highly controlled stimuli. Our goal was to compare eye movements to reduced stimuli and natural videos. We investigated how the relative angle between sequential saccade and pursuit eye movements affects saccadic position error and pursuit gain. We calculated both measurements in different experiments with varying levels of experimental control and stimulus complexity. We used a large set of natural videos, the GazeCom dataset, for which we additionally hand-labeled pursuit targets to extract their trajectories. In experiment 1, we presented these trajectories with an isolated Gaussian blob, lacking any context information. In experiment 2, moving objects followed the same trajectories in natural scenes extracted from the videos. In both experiments an initial fixation cross was presented at a fixed distance to the start of the movement. It was collinear with the direction of the movement or shifted by 30° up- or downwards. In experiment 3 participants just watched the whole set of natural videos without any constraints and we extracted saccade-pursuit combinations comparable to experiment 1 and 2. For all experiments we found a ~10 % reduction in saccadic position error and a ~10% increase in pursuit gain for collinear saccade-pursuit combinations. While the benefit for pursuit eye movements could possibly be explained by simple muscular constraints, like the inertia of the eye, the benefit for saccades indicates a direct influence of the upcoming pursuit on the saccadic eye movement. Despite some small absolute differences in the magnitude of some measurements like the saccade position error (~1 deg with isolated blobs vs. ~1.5 deg in natural videos), we observed comparable results for highly-controlled lab stimuli and free viewing of natural scenes.
Acknowledgement: DFG TRR 135 IRTG 1901