Generation of Numerical Arrays Unveils Statistical Knowledge of Shape Distributions
Yoel Shilat, Omer Gabay, Naama Katzin, Moti Salti, Avishai Henik · 2024
In natural settings, the shape of item arrays (sets), defined by the smallest convex polygon encompassing all objects, predicts the total number of items. Building on our previous work showing array shape's influence on observers’ numerical perception, we introduce a novel generation task. This approach addresses limitations in traditional perceptual paradigms used in numerical cognition. Participants exhibited both implicit and explicit statistical knowledge of shape distributions, demonstrating a fit to randomly scattered item distributions. Intriguingly, as predicted by our model numerical array shape representations were archetypical, with each quantity associated with a specific set of shapes. Our findings offer new insights into the way individuals statistically learn. Moreover, our use of generation paradigms offers a valuable tool for future research into other domains investigating visual representations across cognitive science and neuroscience.