Statistical Summary Perception in Vision

Abstract

In the last 15 years, significant efforts have been made to
investigate statistical processing of object information. This includes
computing properties such as mean or variance of features of multiple
objects present in a visual display. Unlike visual search performance for
individual objects, which is typically dependent on set size, mean estimation
is usually not dependent on set size. The paper reviews studies
on the nature of representations used in statistical processing and consolidation
of the relevant information in working memory. The paper also
discusses the different attributes such as orientation, size and emotions
that have been studied in the context of estimating the mean of those
attributes. One prominent question is the role of attention in statistical
processing. While some argue that attention is not needed for statistical
processing, others argue that attention or in some cases distributed
attention is necessary for statistical processing. The paper critically evaluates
the opposing views and also presents possible issues that need to
be resolved in future.