Abstract
Approximate number of items in a group is an ensemble feature. Extracting number information from a group takes time and involves both perceptual and memory processes. It is unclear how limited cognitive resource and capacity is allocated among multiple groups to extract their approximate numbers. Is there a tradeoff between the quality and the quantity of encoded number representation? That is, is encoding number information from one group faster and more accurate than multiple groups? Here we show the surprising result that there is little or no cost for enumerating two groups in parallel. On each trial, one, two, or four sets of dots (number of dots per set range: 5–25) in different colors were briefly presented and followed by masking sets of dots with variable stimulus onset asynchrony (SOA, 33, 67, 100, 133, or 217 ms). One of these sets was cued with a circle. Participants estimated the number of dots in the cued set by clicking on a continuous number scale. We found that resolution of number representation increased over time, especially for shorter SOAs, suggesting a gradual accumulation process. The accumulation of two sets was as accurate and fast as the accumulation of one set. The accumulation slowed down for four sets of dots. Our findings support a parallel processing of numerosity of multiple sets in perception, in sharp contrast to a serial consolidation process in visual working memory at a later stage.