Abstract
Ensemble coding is a visual processing mechanism that allows for the rapid extraction of statistical summary information from groups of stimuli. It has been shown that ensemble coding operates for low-level visual features such as line orientation (e.g, Dakin & Watt, 1997; Parkes, Lund, Angelucci, Solomon, & Morgan, 2001), as well as higher-level attributes such as facial emotion (e.g, Haberman & Whitney, 2007). Recent work suggests that semantic information is extracted automatically from arrays of digits and influences summary representations of subsequent digit displays (Van Opstal, de Lange, & Dehaene, 2011). Here, we investigated whether numerical information can be extracted in parallel across digits within a display. Participants reported the average (< 5, >5) of digit ensembles with varying numbers of items within the displays. We compared reaction time (RT) patterns across increasing display sizes for the averaging task to RTs from the same participants on visual search tasks using identical displays as in the averaging task, but designed to yield RT slope characteristics of serial (search for a digit < 5 or >5; target is the same color as distractors) and parallel (search for a digit < 5 or >5; target is always red) processing. In a second experiment we controlled for brightness, a low-level cue that could have contributed to performance in experiment 1. Across experiments, search tasks yielded RT slopes typical of serial and parallel search tasks. Unlike in either search task, however, RTs for the ensemble (averaging) tasks decreased with increasing display size. These results indicate that semantic information from multiple digits may be extracted by a parallel processing mechanism. Furthermore, decreasing RTs with increasing display size may reflect a potential benefit of larger displays, as has been previously reported for ensemble perception (Sweeny & Whitney, 2014).
Meeting abstract presented at VSS 2018