Proceedings of the Annual Meeting of the Cognitive Science Society

The verbal overshadowing effect, a phenomenon where verbal descriptions of an encoded face hinder subsequent recognition, has been linked to the similarity in facial image sets used in recognition tasks. However, the specific aspects of similarity that influence this effect remained underexplored. This study, therefore, employed the Stable Diffusion image-generation model to create image sets that are similar either verbally or nonverbally. Experimental results using these sets revealed the presence of the verbal overshadowing effect in the verbally-similar set, but it was not evident in the nonverbally-similar set. These findings align with existing explanations of the verbal overshadowing effect and contribute to enhancing its predictability.