Date: Thursday, 04.11.21 9:00-10:40 CET
Signup: If you would like to attend the talks please register here to get a Zoom link.
Abstract:
The behaviour of any animal is determined by the sensory information it extracts from the environment. All sensory systems, however, can only encode a finite amount of information. Information processing within the human visual system, for example, is constrained by multiple bottlenecks. These bottlenecks can degrade visual resolution. They limit how much information we can remember from one moment to the next. They stop us from noticing objects and events that appear right in front of our faces. And yet we are capable of making sense of complex information within a fraction of a second. How might we explain this great divide between the known limitations of the visual system and our subjective experience of a richly detailed and ordered visual world?
In a series of studies I test one proposed answer: the visual system encodes summary representations of visual scenes. For example, the visual system may compute and encode statistical properties such as the mean or variance of a set of features. We may therefore experience a large amount of information, but “compress" various details as summaries. I will show that such summary statistics are encoded in some circumstances in which working memory demands are relatively high. I will further show that summary statistics of natural images influence observers’ judgments of whether or not small portions of an image appear “normal”. However, I will also present data that questions the role of summary statistics in other naturalistic visual tasks, such as when observers are identifying natural objects.
Summary statistics are thought to play an important role in many aspects of perception and cognition by improving the efficiency of encoding: fewer resources can be devoted to the encoding of each feature because a single summary statistic describes many features simultaneously. My research demonstrates that we are far from fully understanding in which contexts the visual system exploits summary statistics, and in which contexts they are discarded in yet another visual processing bottleneck.
