Bayesian analysis of motion priors and speed discrimination in the peripheryA Nguyen, M S Landy, E P Simoncelli and K Bonnen.Published in Annual Meeting, Vision Sciences Society, vol.23 May 2023. |
Participants viewed stimuli on a Propixx projector. They performed a two-interval forced choice speed-discrimination task. Motion stimuli were comprised of drifting 1/f noise patches, with a circular raised cosine window. Stimuli were either presented near the fovea (4 deg) or more peripherally (20 deg). The stimulus diameter and standard speed were scaled with eccentricity: 4 deg and 20 deg, 2 deg/s and 10 deg/s, respectively. To estimate both sensitivity and bias, participants performed these speed discriminations in three uncertainty conditions: high/high, both stimuli had 75% contrast; low/low, both stimuli had 10% contrast; high/low, the standard had high and the comparison had low contrast.
Overall sensitivity was lower for both low-contrast and peripheral stimuli. For the perifoveal condition (4 deg), there was a bias in the high/low condition. On average, the speed of the matching low-contrast comparison was 24% faster than the high-contrast standard (p<.05). In the periphery, that bias was greatly reduced (or reversed in the case of one participant). These results are consistent with a Bayesian model that combines effects of reduced reliability (i.e., lower speed sensitivity for peripheral stimuli) and even greater broadening of the speed prior (reflecting the wider range of motion experienced in the periphery).