Numerosity adaptation resists filtering: Insights from an illusory contour paradigm

The mechanisms underlying numerosity perception remain debated, with some theories proposing a dedicated system for segmented items and others suggesting reliance on low-level features like spatial-frequency or texture-density. Numerosity adaptation—where exposure to one array alters the perceived numerosity of a subsequent one—has been interpreted as evidence for a numerosity-specific mechanism. However, recent accounts argue that this effect may result from filtering previously processed information. To clarify the underlying mechanisms, we employed a novel adaptation paradigm using Ehrenstein-based illusory-dots as adaptors and real dots as test stimuli. This design allowed us to dissociate numerosity adaptation from low-level features or filtering, as the number of illusory-dots is negatively correlated with spatial-frequency content and, crucially, adaptors and test stimuli contain completely different items. Contrary to predictions from filtering or texture-based accounts, we found a significant increase in the PSE after adaptation, indicating a genuine numerosity-driven effect. Crucially, the point of maximal RTs uncertainty shifted in the same direction, suggesting sensory rather than decisional effects.

To corroborate these findings, in a second experiment, participants estimated the numerosity of grids containing either real or Ehrenstein-based illusory dots, presented in connected or unconnected configurations, keeping constant low-level cues between connectedness conditions. Connected items were consistently underestimated—regardless of stimulus type—confirming that Ehrenstein-based illusory dots are perceived as discrete units.

Together, these results provide strong evidence that numerosity adaptation is not driven by low-level visual features or filtering mechanisms, but by the perceived number of discrete items, supporting the existence of a numerosity-selective system.