Jérémy Lamontagne¹, Vicki Ledrou-Paquet, Anthony Proulx, Laurianne Côté, Isabelle Charbonneau, Caroline Blais, Justin Duncan, Daniel Fiset; ¹Université du Québec en Outaouais, ²Université d'Ottawa

Although studies using the Bubbles technique (Gosselin & Schyns, 2001) typically require large numbers of trials, they have not directly examined how information-use strategies change after very extensive exposure to the same faces. To quantify such changes, we asked four participants to complete 12,000 trials of a face-recognition task over 8–11 sessions. Previous work from our laboratory has shown that better face recognizers rely more heavily on the eye region than poorer recognizers (Royer et al., 2018; Tardif et al., 2019), suggesting that information use may become increasingly optimized during learning and with increasing contact with a given identity. Here, 3D Bubbles were used to investigate how the facial features and spatial frequencies supporting recognition evolved as participants were repeatedly exposed to the same initially unfamiliar identities. Participants were also required to study the faces every day to maintain a high level of contact with those identities and to mimic the kind of repeated exposure that occurs in everyday life. Across all sessions, the eyes remained the most useful information for recognition in all spatial frequencies, but further analyses revealed significant quantitative changes in the use of facial features over the course of the experiment. In particular, participants’ accuracy correlated significantly more with the use of both eyes, in the last session compared to the first, specifically in higher spatial scales, suggesting that the eyes become more important as a face becomes more familiar and that learning sharpens an already eye-centered strategy rather than shifting it to other features. Overall, this experiment shows that changes in the information used for face recognition during learning are quantitative and reflect a progressively more efficient use of the eyes. Extending these results to more naturally varying stimuli would help demonstrate the robustness and ecological validity of this finding.