Anita Paparelli¹ (), Lisa Stacchi¹, Roberto Caldara¹; ¹Eye and Brain Mapping Lab, Department of Psychology, University of Fribourg, Fribourg, Switzerland

Over the past decade, individual differences in visual sampling strategies during facial identity processing have gained considerable interest. Studies have shown that these idiosyncrasies are reliable over time, generalize to real-world behavior, and are directly related to neural responses. However, it is still unclear whether observers simply fixate different face regions while extracting the same diagnostic information, or whether these differences instead reflect genuine individual variation in the facial information used to process identity. To address this question, we first asked a large sample of young healthy adult Western observers to learn eight facial identities presented with a neutral expression. These identities were then presented with different emotional expressions to ensure genuine identity-level processing instead of image matching, in two complementary paradigms. In the first paradignm, observers recognized these identities under free-viewing conditions while their eye-movements were recorded, mapping each observer idiosyncratic visual sampling pattern. Subsequently, they completed the same task using a Bubbles reverse-correlation technique. By computing classification images from correct trials, we identified the diagnostic facial information each observer relied on. We then quantified the pixelwise correspondence between each observer’s fixation pattern and their diagnostic information map to assess whether the information they foveated aligned with the information they used. Our results revealed individual differences in both information sampling and information usage. Crucially, within-observer comparisons showed a strong correspondence between these two components: observers who fixated the eyes, performed correctly when Bubbles masks revealed eye information, whereas those who fixated the mouth required mouth information to succeed. Our findings demonstrate that idiosyncratic sampling strategies reflect a genuine functional difference in the visual information used for effective identification. Our data provide the first direct mechanistic explanation of individual differences in facial identity recognition, further questioning the universality of this critical biological ability in humans.

Acknowledgements: This study was supported by the grant no 10001C_201145 from the Swiss National Science Foundation awarded to RC and ARR.