Liubov Ardasheva¹ (), Anna Montagnini¹; ¹Institut de Neurosciences de la Timone, Aix-Marseille University, CNRS

In the face of sensory uncertainty several factors can impact our perceptual choices. This study aims at exploring the influence upon human motion direction estimation, of the repeated presentation of specific visual motion features, leading to the formation of experience-based priors. We test the hypothesis that these learnt priors should influence subsequent perception and sensorimotor behavior in interaction with longer lasting, built-in beliefs. Recent studies suggest that learnt priors exert a sort of "magnetic force," drawing perception toward or pushing it away from the expected direction. The focal point of our research is to gain insight into this "repulsion" phenomenon. In our experiment, participants had to estimate the global motion direction of a Random Dot Kinematogram (RDK), with 16 possible directions of movement and with three coherence levels: 0.4 (easy), 0.15 (medium), and 0.05 (hard). Additionally, we conducted three sessions: one unbiased, fair for each angle, and two biased sessions with 60% of trials with a given direction (Right, 0°, or Up-Left, 135°) potentially inducing a non-uniform direction prior. Our results do not validate a repulsion effect of direction estimate, neither in the unbiased session nor in biased sessions. Instead, they reveal a persistent prevalence of preferences for cardinal directions in erroneous trials across all sessions (43%) and especially evident for the low coherence RDK, with a distinct choice-bias for horizontal right, with the notable exception for the session with the over-represented rightward motion, and the upper hemifield. In contrast, oculomotor anticipation displays adaptability to the experimentally-induced bias in two sessions, offering minor but noteworthy evidence supporting a dissociation between anticipatory eye-movements and perceptual decisions. Overall our results suggest that our internal beliefs carry more weight than situational biases, emphasising the importance of dissecting the different sources and the specific dynamics of internal priors in motion estimation and visuomotor tracking.

Acknowledgements: This work is partly funded by ANR21-CE37-0018-01, Vision-3E Grant