Item

Abstract

Bringing a visual object at a peripheral visual location to fovea by a saccade helps recognize this object. Human observers can integrate pre- and post-saccadic visual inputs for the recognition. To our knowledge, this integration has only been studied by an instructed saccade to a target at a prescribed and known location. Furthermore, the target is typically the only meaningful object in visual display, and the post-saccadic viewing duration of the target is often fixed by experimental design. For the first time, we study trans-saccadic integration in visual exploratory behavior, when observers decided themselves when and to which locations to make their saccades. We ask whether the pre-saccadic perception may be too ambiguous to contribute to the trans-saccadic integration when the pre-saccadic target is too peripheral from fovea in a crowded scene. To answer this question, we study the trans-saccadic integration in visual search behavior to find and report as soon as possible a target among 404 non-targets. Distributed locations of the search items, each 1.32° × 0.6° in visual angle, spanned 57.3° × 33.8° in visual angle. We measured (1) the pre-saccadic retinal eccentricity, e, of the target, and (2) the post- saccadic foveal viewing duration, T, which ended when the target was reported. We found that T increased with increasing e and eventually saturated at e around 10°–20°. We model our data by a decision-making process that integrates the pre-saccadic peripheral visual inputs with post-saccadic central visual inputs. Due to crowding, the pre-saccadic contribution to this integration diminishes with e. Meanwhile, the post-saccadic contribution increases with T. To reach a threshold for decision-making, our model predicts the experimental observation that trans-saccadic integration peters out with increasingly peripheral pre-saccadic object. This observation should apply to general visual exploratory behavior, at least when object perception is highly vulnerable to crowding.