Target frequency influences antisaccade endpoint bias: evidence for perceptual averaging.

Perceptual judgments related to stimulus-sets are represented computationally different than individual items. In particular, the perceptual averaging hypothesis contends that the visual system represents target properties (e.g., eccentricity) via a statistical summary of the individual targets included within a stimulus-set. Here we sought to determine whether perceptual averaging governs the visual information mediating an oculomotor task requiring top-down control (i.e., antisaccade). To that end, participants completed antisaccades (i.e., saccade mirror-symmetrical to a target) – and complementary prosaccades (i.e., saccade to veridical target location) – to different target eccentricities (10.5°, 15.5° and 20.5°) located left and right of a common fixation. Importantly, trials were completed in blocks wherein eccentricities were presented with equal frequency (i.e., control condition) and when the ‘proximal’ (10.5°: i.e., proximal-weighting condition) and ‘distal’ (20.5°: i.e., distal-weighting condition) targets were respectively presented five times as often as the other eccentricities. If antisaccades are governed by a statistical summary then amplitudes should be biased in the direction of the most frequently presented target within a block. As expected, pro- and antisaccade across each target eccentricity were associated with an undershooting bias and prosaccades were refractory to the manipulation of target frequency. Most notably, antisaccades in the proximal-weighting condition had a larger undershooting bias than the control condition, whereas the converse was true for the distal-weighing condition; that is, antisaccades were biased in the direction of the most frequently presented target. Thus, we propose that perceptual averaging extends to motor tasks requiring top-down cognitive control.