Stimulus-driven mechanisms underlying visual search asymmetry revealed by classification image analyses

Abstract

Search asymmetry is a robust phenomenon with various stimuli and is important for understanding determinants of efficiency in visual search. However, its underlying mechanism remains unknown due to the lack of a method for estimating visual features used by human observers. This study used a classification image technique to solve this problem. Standard classification image analyses with an experiment of visual search asymmetry between Q and O revealed that observers used the same features in both search tasks, rejecting a hypothesis incorporating top-down feature selection. More quantitative data analysis and an additional experiment with a singleton search task also rejected target-dependent selective tuning of the common feature. Further model-based analyses revealed that a standard signal detection model with nonlinear signal transduction and multiplicative internal noise is sufficient to account for the classification image data. Contrary to intuitively appealing accounts based on attention and spatial uncertainty, these findings suggest that search asymmetry is a characteristic of elementary visual processing of multiple items by a nonlinear system. The classification image technique is a valuable tool for investigating search behavior beyond mere visualization of visual features.