Abstract
In ensemble perception, the visual system extracts summary statistical information of multiple items. Ensemble perception occurs for low-level features such as orientation, as well as higher level features such as facial expressions. The different contributions of foveal and parafoveal vision to the extraction of average facial expressions are still under debate. In two experiments, groups of faces with varying emotions (11 levels of morphed expressions from happy to disgusted) were presented for 100ms at three different eccentricities (0°, 3°, 8°). Observers reported the perceived emotion (responding “positive” or “negative”) of either only the central face (central task) or the average of all faces (average task). In both tasks, stimuli consisted of a central face flanked by 8 faces. Additional stimuli consisted of a single face (central task) and a set without central face (average task). In Experiment 1, flanker emotions were uniform, in Experiment 2 they were varied. As expected, performance in the central task declined with increasing eccentricity in all conditions. Reporting the average emotion, however, was superior when stimuli were presented at 3° compared to 0° and 8°. Importantly, at 0°, the perceived average emotion was strongly biased by the central face. When the central face was absent, performance was best at 0° (with flankers surrounding the fixated region). The central face bias was also evident when comparing congruent trials, where central and flanker emotions were both either positive or negative, with incongruent trials. Performance was better when central face and flankers were congruent compared to incongruent. Our results showed how foveal input determined the perceived emotion of ensembles of faces, suggesting that ensemble perception may fail when salient target information is available in central vision.