Abstract
Adults can enumerate up to three simultaneously- or sequentially-presented sets in parallel (Feigenson, 2008; Halberda et al., 2006). Critically, the stimulus items in previous studies contrasted in category membership such that each item belonged to only one set—hence observers could store information about three independent sets.
We asked whether observers can also store a single, unified representation for the purposes of later re-parsing, and what the limits of this ability might be. In Experiment 1, observers verbally shadowed while watching a stream of red triangles and circles and blue triangles and circles placed into two buckets in unpredictable, temporally intermixed order. Hence any given object simultaneously belonged to two sets (e.g., the set of red objects, and the set of triangles). Observers successfully judged the relative numerosity of all possible set parsings (i.e, successfully indicated which bucket had more red objects, blue objects, triangles, circles). Because of the 3-set limit demonstrated in earlier work, observers could not have succeeded by tracking all 4 overlapping sets, but instead must have flexibly parsed a representation from memory.
We next probed the limits of this ability. In Experiment 2, observers saw three orthogonal dimensions (colour, shape, topology) and were found to make accurate numerosity judgments for sets defined by all three dimensions (i.e., 6 different sets). In Experiment 3, observers saw four dimensions (colour, shape, topology, size), and failed to make accurate judgments along all four dimensions (8 different sets), thus replicating the 3-set limit of WM.
These findings suggest adults can simultaneously track multiple overlapping sets and repeatedly parse the array while it remains stored in WM. That the previously demonstrated 3-item limit of WM appears to converge with a 3-dimension limit suggests that the units of WM may be objects, sets, or featural dimensions such as those explored here.