Abstract
Face recognition is a form of perceptual expertise that develops into adulthood. Daily visual exposure to faces, described as the 'face-diet,' fundamentally shapes facial recognition abilities. For example, in racially homogenous societies, the lack of exposure to racially diverse faces contributes to the 'other-race-effect.' Furthermore, an enriched face-diet with many unique identities is necessary to support and refine face processing competence. A previous study that examined the adult face-diet characterized exposure on several variables including duration, viewing distance, and pose but did not address exposure statistics for unique facial identities (Oruc et al. 2018). Here, we developed a novel deep learning-based tool to quantify the number of unique facial identities in first-person perspective footage for 30 adults in Metro Vancouver, British Columbia, Canada. We show that an adult observer encounters an average of 40 unique faces during a typical day, with familiar faces accounting for approximately half. Our findings showed that 77% of the total face exposure was familiar, and 70% was accounted for by the top five most frequently encountered identities. These results may reflect the longer encounters typical of social interactions with familiar individuals. For all but two participants, the most frequently occurring face was a familiar identity that accounted for an average of 33% of total face exposure. For the majority of participants (63%), the most frequently occurring familiar face was of the opposite sex. Contrary to the predicted outcomes of studying observers in a region known for its ethnically diverse population, the racial distribution of identities diverged from Metro Vancouver's census data (2016) of Statistics Canada and was biased towards own-race faces for both familiar and unfamiliar identities. This last finding emphasizes the prevalence of societal and cultural barriers that may prevent co-mingling between racially and ethnically diverse communities despite favourable demographics.