No consistent bidirectional aftereffect from numerosity adaptation

Numerosity has been proposed as a primary visual attribute, akin to motion or colour, and has been reported to show bidirectional adaptation consistent with population coding by tuned mechanisms: after exposure to sparse or dense arrays, the perceived numerosity of an array with a moderate number of dots shifts away from the adaptor. However, recent studies argue against bidirectionality, claiming that adaptation to sparse arrays failed to increase perceived numerosity of a denser array. This suggests that previous accounts may reflect decisional rather than perceptual biases. To test this, we employed a comparison-of-comparisons paradigm using a four-quadrant display. Observers adapted to two spatially distinct dot clouds in upper quadrants of the visual field. Following adaptation, dot clouds were shown in all four quadrants. Observers judged which of the two bottom clouds appeared more different to the reference numerosity contained in each upper quadrant. Simple decisional biases such as a preference for selecting a previously adapted or unadapted stimulus cannot masquerade as perceptual bias in this paradigm. Across six observers, three adaptor numerosities (0, 25, 100) and three randomly interleaved reference numerosities (34, 50, 74), we found clear evidence for unidirectional aftereffects consistent with monotonic neuronal responses. Adapting to dense clouds (100 dots) produced the largest downward bias in perceived numerosity, with smaller reductions following 25-dot adaptors. No condition produced upward shifts, suggesting that previous reports of bidirectionality may reflect non-perceptual biases.