Response to “Crows recognize geometric regularity”

Schmidbauer et al. (2025) claim that crows attend to geometric regularities in the same way as humans. They base their conclusions on an intruder task experiment with a modified subset of the quadrilateral stimuli previous used by Sablé-Meyer et al. (2021) in humans and baboons. We show that, in this stimulus set, geometric features (e.g., sensitivity to parallels, right angles, or symmetry) can be conflated with low‑level visual features. To better understand the crows’ behavior, we re-analyzed Schmidbauer et al.’s data using established models designed to disentangle geometric and visual features and previously validated on human and baboon data. Results show that the perceptual model, captured by a convolutional neural network (CNN), suffices to explain the crows’ performance, and we find no evidence to support the claim of a geometric regularity effect in crows