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 in perception has recently been criticized. Here, we evaluate these criticisms and present an account based on a recently developed psychophysical methodology, Feature Distribution Learning (FDL), which provides promising evidence for probabilistic representations by avoiding these 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 distractor features is encoded.We explain how FDL results provide evidence for a stronger notion of representation that relies on structural correspondence between stimulus uncertainty and perceptual representations, rather than a mere co-variation between the two. Moreover, such an account allows us to demonstrate what kind of empirical evidence is needed to support probabilistic representations as posited in current probabilistic Bayesian theories of perception.

Probabilistic representations as building blocks for higher-level vision

“…Taken together, our results show that observers can not only encode probabilities of features from heterogeneous stimuli in detail but also integrate them with both locations and other features that have different distributions. These results arguably represent the strongest support yet for the longstanding idea that the brain builds probabilistic models of the world (Chetverikov et al, 2017a;Fiser et al, 2010;Knill & Pouget, 2004;Orhan & Ma, 2015;Rao et al, 2002;Sahani & Dayan, 2003;Tanrıkulu et al, 2021) and show that probabilistic representations can serve as building blocks for object and scene processing. Notably, such representations are not simply limited to summary statistics (e.g., a combination of mean and variance (Cohen et al, 2016)).…”

“…One captivating idea is that the brain builds statistical models that describe probability distributions of visual features in the environment (Rao et al, 2002;Pouget et al, 2000;Zemel et al, 1998;R. D. Lange et al, 2020;Fiser et al, 2010;Knill & Pouget, 2004;Tanrıkulu et al, 2021). By combining information about different features and their locations, the brain can then form representations of objects and scenes.…”

Perception is rich and probabilistic

“…1 c). To recap, from our original design to investigate working memory 16 we made three important changes overall to adapt it to study perception, first was the switch to a single shape stimulus per display, not an array 13 , 14 , since having multiple items introduces problems with confusion among the items 18 – 21 . Second was the short stimulus timing, and third was the noise added to the stimuli.…”

“…Other studies reveal conflicting results, but since they use an array of target stimuli per display, e.g. orientations that have some spread 13 , 14 , they do not meet the ideal criteria for isolating perception. For example, Rahnev and Block 15 argue that perception is only likely to be discrete in instances in which an object is briefly presented in isolation.…”

Priming of probabilistic attentional templates

“…It is interesting to note that this equation predicts that the learning of distribution means should be easier than learning their variance and lower variance should be easier to learn than higher variance (Tanrikulu et al, 2021b). With regard to Bayesian accounts: the more uncertain the data, the more the prior influences perceptual interpretation (Series & Seitz, 2013;Tanrikulu et al, 2021a) and there is indeed evidence that uncertainty regarding target identity increases the strength of priming (Meeter & Olivers, 2006).…”