The Role of Real-World Statistical Regularities in Visual Perception

Abstract

Multiple models of vision propose that perception involves a process of prediction and verification. Here we argue that real-world statistical regularities—representations that, on average, more quickly make contact with meaning—serve as the basis of these predictions. We show that statistically regular images—those, we argue, that more closely match perceptual predictions—are more readily perceived and more efficiently processed than statistically irregular images.