Foraging for targets with saccades

Abstract

Experienced players exploit their knowledge to scan chess games more efficiently with eye movements than novice observers (Reingold et al. 2001). In contrast, finding a friend's face in the crowd or searching for a dropped ring on a pebbled road may be frustrating for novice and experienced searchers alike. For such crowded scenes, the brain should attempt to optimize search in a statistical sense, because the peripheral visual information is too limited to guide search. We report here saccadic amplitude distributions for one-dimensional and two-dimensional search tasks. In two-dimensional crowded displays strings of similarly directed saccades appear. Those strings break down the search into piecewise one-dimensional domains. The distribution of lengths of the one-dimensional domains follows an inverse power law (Levy distribution) with an exponent in the range between −1 and −2 (Viswanathan et al, 1999). Foraging animals that can detect food sources only in their immediate vicinity optimize search by such a distribution for their flight lengths. Humans appear to use a ‘saccadic foraging’ strategy to optimize search in crowded displays.

Footnotes

 This work was supported by NWO-ALW grant#809.37.003