Perceptual grouping, not covert attention, drives the connectedness effect in the ANS.

A robust finding in numerical cognition is that connecting items within an array leads to systematic underestimation of numerosity. This provides evidence that approximate numerosity perception relies on discrete objects rather than on continuous variables (e.g., total area, density, convex hull). While this connectedness effect is often attributed to perceptual grouping, an alternative interpretation is that connected items may capture covert attention, thereby biasing the sampling of visual information. We tested these competing accounts by combining a numerosity estimation task with a target-detection task modeled after Posner’s cueing paradigm. On each trial, participants viewed dot arrays (14–20 items) that included two red lines, either connecting a pair of dots or terminating near unconnected dots. A target diamond could appear either near (congruent) or far (incongruent) from the quadrant containing the red lines. Participants first performed a go/no-go detection task, then estimated the array numerosity. Replicating prior work, connected arrays were consistently underestimated relative to unconnected ones. Crucially, detection performance showed no evidence of attentional capture: reaction times and accuracy did not differ as a function of connectedness or target position. These findings demonstrate that the underestimation effect cannot be attributed to covert attentional allocation. Instead, they support the view that perceptual grouping—rather than attentional biases—drives the connectedness effect in the Approximate Number System. More broadly, our results strengthen the case for segmentation-based mechanisms as a critical foundation of visual number perception.