Abstract
<p>Individuals explore options with their eyes before selection, in recognition decisions and beyond. We analyzed eye movement behavior in a 3-alternative forced-choice face recognition task, and characterized the results in terms of eight empirical benchmarks. We then evaluated the ability of a suite of models to recover those benchmark behaviors. At the heart of all the competing models is a random walk for each examined face, with one boundary leading to a potential endorsement of that face and the other leading towards examination of another face. The most successful model had three novel characteristics: starting bias becomes more liberal over successive fixations; an early response deferral mechanism allows possible transitions to other faces even when an endorsement boundary is reached, especially on early fixations; and a late response advancement mechanism allows selection of a previously viewed face when a transitioning boundary was reached, especially on later fixations. All of these novel mechanisms represent a shift towards more liberal responding on later fixations. This model provides a starting point for understanding how individuals strategically deploy visual attention over complex displays in service of memory decisions.</p>