Dissociating implicit and explicit ensemble representations reveals the limits of visual perception and the richness of behavior

Sabrina Hansmann-Roth,Andrey Chetverikov,Árni Kristjánsson,David Whitney

doi:10.1038/s41598-021-83358-y

Abstract

Our senses provide us with a rich experience of a detailed visual world, yet the empirical results seem to suggest severe limitations on our ability to perceive and remember. In recent attempts to reconcile the contradiction between what is experienced and what can be reported, it has been argued that the visual world is condensed to a set of summary statistics, explaining both the rich experience and the sparse reports. Here, we show that explicit reports of summary statistics underestimate the richness of ensemble perception. Our observers searched for an odd-one-out target among heterogeneous distractors and their representation of distractor characteristics was tested explicitly or implicitly. Observers could explicitly distinguish distractor sets with different mean and variance, but not differently-shaped probability distributions. In contrast, the implicit assessment revealed that the visual system encodes the mean, the variance, and even the shape of feature distributions. Furthermore, explicit measures had common noise sources that distinguished them from implicit measures. This suggests that explicit judgments of stimulus ensembles underestimate the richness of visual representations. We conclude that feature distributions are encoded in rich detail and can guide behavior implicitly, even when the information available for explicit summary judgments is coarse and limited.

Highlights

Our senses provide us with a rich experience of a detailed visual world, yet the empirical results seem to suggest severe limitations on our ability to perceive and remember
Several highly influential studies have suggested that the visual system exploits these redundancies in the environment to encode our surroundings as summary statistics of ensembles to save resources and bypass the bottlenecks of attention and working memory
Claims have been made that summary statistics determine the richness and limitations of conscious experience[15]. Are such simple summary statistics all that is encoded? Or is detailed information about feature distributions retained in some way? Information can be reduced to the mean and variance of a distribution and perceptual tasks like outlier detection or categorization can be performed through knowledge of these statistics[35,36,37]

Summary

Introduction

Our senses provide us with a rich experience of a detailed visual world, yet the empirical results seem to suggest severe limitations on our ability to perceive and remember. Reducing sets of items to ensemble summaries requires similar capacity and attentional effort as representing individual items[10,13] Such summary statistical representations provide a high-level description of visual features and can compress visual information and reduce noise[14]. Optimal behavior requires more than summary statistics because observers need a correct model of the environment: a “generative model” in terms of ideal observer approaches[16] Such models require knowledge of the shape of a probability distribution (for example, a uniform distribution would lead to a different inference than a Gaussian even if their variances are equal). Another possibility is that explicit tests of ensemble perception fail to reveal all the knowledge that observers have

Methods

Results

Conclusion