Learning What to See in a Changing World.

Katharina Schmack,Veith Weilnhammer,Klaas E Stephan,Jakob Heinzle,Philipp Sterzer

doi:10.3389/fnhum.2016.00263

Abstract

Visual perception is strongly shaped by expectations, but it is poorly understood how such perceptual expectations are learned in our dynamic sensory environment. Here, we applied a Bayesian framework to investigate whether perceptual expectations are continuously updated from different aspects of ongoing experience. In two experiments, human observers performed an associative learning task in which rapidly changing expectations about the appearance of ambiguous stimuli were induced. We found that perception of ambiguous stimuli was biased by both learned associations and previous perceptual outcomes. Computational modeling revealed that perception was best explained by a model that continuously updated priors from associative learning and perceptual history and combined these priors with the current sensory information in a probabilistic manner. Our findings suggest that the construction of visual perception is a highly dynamic process that incorporates rapidly changing expectations from different sources in a manner consistent with Bayesian learning and inference.

Highlights

Sensory signals are inherently noisy and ambiguous
Our results so far suggest that perception of the ambiguous sphere stimuli was biased by the changing contingencies, indicating that perceptual expectations are rapidly updated by associative learning and are used in perceptual inference
These results indicate a role of associative learning in addition to perceptual history in the continuous adaptation of perceptual expectations, suggesting that different aspects of prior experience are used to resolve perceptual ambiguity in a synergistic manner consistent with Bayesian learning and inference

Summary

INTRODUCTION

Sensory signals are inherently noisy and ambiguous. To make sense of such information, our perception strongly relies on expectations. When a photograph of a striped dress is consistent with two different color interpretations, the visual system resolves this ambiguity by using the expectation about how colors appear under common illumination conditions (Gegenfurtner et al, 2015; Lafer-Sousa et al, 2015) Such perceptual expectations about the appearance of a stimulus can be modified by repeated experience of the stimulus (Orbach et al, 1963; Maier et al, 2003; Chalk et al, 2010; Fischer and Whitney, 2014) or of associated cues (Sinha and Poggio, 1996; Ernst et al, 2000; Adams et al, 2004; Haijiang et al, 2006; Flanagan et al, 2008; Sterzer et al, 2008; Di Luca et al, 2010; Schmack et al, 2013). We hypothesized to find evidence for a continuous integration of different sources of experience in the moment-to-moment updating of perceptual expectations, resulting in the flexible adaptation of perceptual inference to dynamic changes in the environment

Participants

General Method

RESULTS

DISCUSSION