What kind of empirical evidence is needed for probabilistic mental representations? An example from visual perception

Abstract

Recent accounts of perception and cognition propose that the brain represents information probabilistically. While this assumption is common, empirical support for such probabilistic representations has recently been criticized. It has been argued that due to methodological limitations of perceptual experiments, probabilistic theories only provide a model for perceptual decisions, but not perception itself (Block 2018). Moreover, results showing that the visual system represents sensory uncertainty can be explained by non-probabilistic or summary statistical accounts that rely on correlations between stimulus uncertainty and inner visual states (Rahnev, 2017; Block 2018). Here, we argue that a recently proposed psychophysical method, Feature Distribution Learning (FDL), provides promising evidence for probabilistic representations by sidestepping these recent criticisms. The method uses priming and role-reversal effects in visual search. Observers’ search times reveal the structure of perceptual representations in which the probability distribution of distractors is encoded. This method avoids the criticism mentioned and its results cannot be accounted for by non- probabilistic representations, since they suggest that the structure of inner perceptual states follows stimulus uncertainty. We explain how FDL results provide evidence for a stronger sense of the notion of representation that relies on structural isomorphism between stimulus uncertainty and perceptual representations, rather than a co-variation between the two. We also discuss further research directions in the light of the results from the FDL method. Our main goal is to demonstrate that FDL studies provide a novel perspective on these issues and provide the most promising evidence for probabilistic mental representations to date