Noisy variability does not benefit face learning.

Previous research provides abundant evidence that variability plays a key role in the process by which a newly encountered face becomes familiar. Theoretical models propose that exposure to variable images improves recognition because such variability is diagnostic (i.e., predicts idiosyncratic within-person variability in appearance). We sought disconfirmatory evidence for such models by examining whether noisy variability also benefits learning. The present study provides compelling evidence that two sources of noisy variability-contextual variability and augmentation (visual noise)-do not improve generalization of learning. Across five preregistered experiments and a total sample of over 1,000 participants, Bayesian statistics revealed no benefit of noisy variability, even when the study phase required active retrieval and even among child participants (i.e., participants with less expertise). These findings support current models of face learning and pave a path toward a deeper understanding of how variability operates. These findings also contribute to recent calls to identify core principles of variability across domains. (PsycInfo Database Record (c) 2026 APA, all rights reserved).