Abstract
Finding a target object in a complex visual scene requires search across the visual field. Although early visual processing is characteristically parallel, difficult visual search tasks with significant set size effects on search time are often associated with time-limited serial operations of the attention system. However, the typical measures of the set size effects (increased average response time and decreased error rates for larger displays) could be consistent with either a serial or a parallel architecture. A full time-course analysis of visual search is necessary. We measured target detection accuracy at search times from 0.1 to 2.3 s for several difficult and error prone searches for different display sizes. Using computational models with probabilistic responses (Dosher, Han, & Lu, 2004) corresponding with independent information accumulation across objects within a single epoch of information acquisition, we distinguish unlimited capacity parallel versus serial search mechanisms. Rather than invoking serial, time-limited deployment of covert attention, a range of difficult searches, including search asymmetries, heterogeneous distractor effects, and some conjunction searches, instead exhibited nearly identical search dynamics regardless of the size of the display [although accuracies differ] — information-limited, not time-limited processes. Slightly slowed dynamics in large displays reflects the impact of classification errors on the early dynamics of probabilistic parallel search. These observations are related to but go beyond unlimited capacity statistical models of asymptotic accuracy of search (e.g., Palmer, Verghese, Pavel, 2000; Eckstein, Thomas, Palmer, Shimozaki, 2001) and relate directly to results in multi-target visual search (Thornton & Gilden, 2007). These time-course dynamics also support the assumptions of recent models of parallel information acquisition across the field in Bayesian prediction and analysis of eye movement selection (Najeminik & Geisler, 2005). The visual and item-comparison systems operate as an unlimited-capacity parallel analysis over the visual field within a single eye fixation.
Funded by AFOSR, NSF, NEI.