Proceedings of the Annual Meeting of the Cognitive Science Society

Global motion detection performance is often described via ___ (i.e., equal-variance Gaussian Signal Detection Theory), de- spite evidence that global motion detection requires a threshold representation which is incompatible with ___. Both ___ and threshold representation share the core assumption of random- scale representation (RSR), which posits a latent evidence scale without committing to specific distributional forms. We apply critical tests to assess whether global motion detection requires a RSR and whether a threshold representation is admissible. Participants completed an __-alternative forced-choice random dot motion task with __ = {2 . . . 8}. Accuracy across __ sat- isfied RSR constraints, providing strong evidence that global motion detection relies on RSR. We derived a hazard func- tion from the accuracy and tested the monotonicity constraint implied by threshold representation. The hazard function satis- fied monotonicity, consistent with a threshold account. Recon- structed ROC curves were piece-wise linear, inconsistent with equal-variance Gaussian SDT, while generally consistent with threshold models.