Gabriel C. L. Waterhouse¹, Sami R. Yousif²; ¹University of North Carolina at Chapel Hill, ²The Ohio State University

Imagine looking out at a forest. What might you see? Visual masses of green, the number of trees, the various sizes of the leaves. Is one of these dimensions more salient than the others? According to a popular view, numerical information is uniquely prioritized in perception, being extracted and utilized in preference to other dimensions. This “number bias” is more than a perceptual preference: It has been argued that this proclivity for numerical information is directly responsible for the cultural evolution of mathematical knowledge. Here, we question this orthodoxy. In eleven experiments and six unique paradigms, we demonstrate that multiple other visual ensembles are prioritized over number by both adults and children. We first used a categorization task to contrast number and average size. Participants categorized displays using abstract symbols, learning each symbol corresponded to either “less” or “more” in both dimensions. When the dimensions were later pitted against each other, participants overwhelmingly categorized based on average size rather than number. In another set of experiments, participants were asked to place a line to divide an 8x8 stimulus space (each dimension corresponding to a different ensemble) into two distinct groups. Across three versions of this task, with various dimensions contrasted, participants decisively failed to use number, instead dividing by average size or cumulative area. We then put this “size bias” to the test, using several unique paradigms to measure the relative salience of average size and number. Across all experiments, we found that both adults and children preferentially matched, distinguished, and rapidly identified stimuli based on average size rather than number. Collectively, these results undermine claims about a universal number bias. While numerical knowledge is indeed a crucial cultural artifact, this knowledge may not be rooted in perceptual biases.