The Information Theory of Vision: Evidence from Eye-Tracking.

Abstract

Recently, our lab demonstrated a logarithmic relationship between reaction time and the number of items in a search display (e.g., Lleras & Buetti, VSS 2014; Cronin, Buetti, and Lleras, VSS 2014). Until now, the Reaction Time X Set-Size function was thought to be linear. Lleras and colleagues' displays differed from typical search displays in that there were various item-types differing in their similarity to the target. To account for their findings, the authors presented a new theory of visual search, the Information Theory of Vision (ITV). ITV proposes a two-stage model of visual search. The first stage is an unlimited capacity, but resolution-limited parallel processor. The second stage is a limited capacity, but resolution-unlimited processor. ITV uniquely proposes that the first stage of search proceeds logarithmically as unlikely targets (lures) are sequentially rejected. Items that are very dissimilar to the target are rejected rapidly, while items that are more similar to the target are rejected more slowly. The items most similar to the target (candidates) are passed on to the second stage for more detailed processing. The present study attempts to provide more evidence for ITV by replicating the results of two of Lleras and colleagues' behavioral experiments while monitoring for eye movements. Using eye-tracking allowed us to evaluate several unique claims made by ITV, in particular, the predicted rates of false alarm eye movements to items of different similarities to the target. This new analysis allows us to better differentiate ITV from other models of search, including Guided Search (Wolfe, 1994), Similarity Theory (Duncan & Humphreys, 1989), and the Target Acquisition Model (Zielinsky, 2008). Meeting abstract presented at VSS 2015.

DOI: 10.1167/15.12.959

Cite this paper

@article{Cronin2015TheIT, title={The Information Theory of Vision: Evidence from Eye-Tracking.}, author={Deborah Cronin and Alejandro Lleras and Simona Buetti}, journal={Journal of vision}, year={2015}, volume={15 12}, pages={959} }