Feature specificity in visual statistical summary processing

Abstract

People initially process object sets by averaging their features into statistical summary representations. Previous studies have shown that there is independence between summarizing of low and high-level visual information. There is also evidence showing that, processing capacity of multiple statistical summaries is limited for simultaneous averaging of same kind of visual information, but not for different kinds of visual information. The current thesis investigates whether statistical summary processing relies on a feature-specific or a feature-general mechanism, and whether there are capacity limitations to simultaneous averaging of different visual features. We asked participants to average on of the features in a set of lines that varied in size and orientation. The relevant feature was either the same throughout a block or mixed within the trials of a blocks. Even though first two experiments showed a positive relation between viewers’ size and orientation averaging performances for mixed averaging conditions, with more controlled displays we repeatedly found that performances on two tasks were unrelated both for single and mixed conditions. Viewers’ errors for size averaging were higher in mixed than single averaging conditions, however this difference disappeared with reduced task difficulty. Orientation averaging performances were similar in single and mixed conditions. Finally, viewers’ performances on size and orientation averaging tasks were similar across 50, 100 and 200 milliseconds of encoding durations. Overall, results of this thesis suggested that there are independent feature-specific statistical summary mechanisms for size and orientation features.