Alejandro G. Garcia^1,2 (), Chi T. K. Dao^1,2, Pascal Mamassian³, Peter J. Kohler^1,2; ¹Department of Psychology, York University, ²Centre for Vision Research, York University, ³CNRS & École Normale Supérieure

When four dot pairs rotate about the vertices of an implied square, they are initially perceived as four independently rotating pairs: the “local percept”. Soon, however, observers begin to perceptually group the dots into two overlapping illusory squares translating along a circular trajectory: the “global percept” (Anstis, 2003). Previous research has demonstrated that the global percept appears to move slower than the local percept (Kohler et al., 2009). This so-called “global slowdown” cannot be explained by differences in perceived size, emergent rotation, or number of perceived moving elements between the two percepts (Kohler et al., 2014). Here, we test the possibility that the effect arises from systematic differences in sensory uncertainty of speed estimates between the two percepts. If the global percept is associated with higher uncertainty, a Bayesian prior toward slow motion (Weiss et al., 2002) would lead to a reduction in perceived speed. As in previous work, we used a 2AFC task where participants compared the speed of two configurations presented on either side of fixation. The reference configuration always moved at a fixed speed, while the test configuration took on a range of different predetermined speeds across trials. We measured the slowdown effect by comparing a global-biased reference configuration to a local-biased test configuration and computing the Point of Subjective Equality. To assess uncertainty, we measured the slopes of the psychometric functions obtained from control conditions in which both reference and test configurations were either local-biased or global-biased. We found no significant difference in the slopes between the two percepts, indicating comparable levels of uncertainty. We replicated the global slowdown effect but found no systematic relationship between size of the effect and local-global differences in uncertainty. This suggests that a Bayesian prior for slow motion is not responsible for the global slowdown effect.

Acknowledgements: This research was undertaken thanks in part to funding from the Vision: Science to Applications, supported by the Canada First Research Excellence Fund. Additional support was provided by a Natural Sciences and Engineering Research Council of Canada Discovery Grant awarded to PJK.