People are remarkable at processing human faces. In a split second, one can recognize a person’s gender, race, or emotions. However, this expertise only works when the faces belong to one’s own racial group. Known as the own-race advantage (ORA), this psycholgical limitation affects people of all colors. People are notoriously bad at distinguishing between members of unfamiliar racial groups. Scientists have reliably demonstrated the ORA effect across different ethnic groups and with various experimental paradigms. These data show that ORA is not because some races are more homogenous than others. Instead, it results from the absence of interracial contact experience. From an ontogenetical point of view, it is of great interest to investigate when and how the own-race advantage develops.

Past research using a procedure designed for studying non-verbal infants called the “preferential looking method” has found that 3-month-old Caucasian infants looked longer at Caucasian faces compared to faces from other racial groups . Likewise, 3-month-old Chinese infants looked longer at Chinese faces. However, Ethiopian infants who were frequently exposed to both Ethiopian and Israeli adults showed similar looking preferences for both African and Caucasian faces. These findings suggest that 3-month-old babies are sensitive to race, and pay more attention to faces of familiar race in their environment.

Growing up, we naturally encounter more own-race faces than other-race faces. How does this asymmetrical exposure shape the long-term development of the own-race advantage? There are two complementary views on the ontogeny of ORA. The perceptual narrowing view states that appropriate visual experience is necessary for maintaining the neural representations for faces from different racial groups. Evidence for this view comes from studies in which 3-month-old Caucasian infants exhibited a broad discriminability for faces of own and other races. Whereas 6-month-olds retained discriminability for Caucasian and Chinese faces, by nine months, infants were unable to distinguish among faces belonging to any racial groups other than their own. A follow-up study with 3, 6, and 9-month-old Chinese infants showed the same pattern of development consistent with the perceptual narrowing hypothesis. These observations echo a classic phenomenon in the early development of phonetic perception—Japanese babies can distinguish between all phonemes, including those not present in their native language (e.g., “R” and “L”), but this ability is lost as the infants grow older.

In our research we explored the second viewpoint, called the perceptual learning view, which states that frequent exposure to particular stimuli improves sensitivity to those stimuli. Our starting point comes from a contrasting finding in the existing literature: several studies reported that 3- and 4-month-old infants, instead of showing a broad discriminability, exhibited a better discriminability for own-race faces. Although perceptual narrowing happens in infant speech perception, it is inaccurate to compare the stimulus variability in human languages with those in human faces. Languages in the world are vastly diversified, but faces of all races share many similarities in structure. In two studies, my colleagues and I explored whether frequent exposure to own-race faces improves the ability to discriminate among own-race faces without hurting the discriminability of other-race faces in later development.

First, we showed 4-, 6-, and 9-month old Taiwanese infants Taiwanese, Caucasian, and Filipino faces. To examine whether infants show differential improvements in distinguishing among faces of own versus other races, we created a set of discrimination tasks with increasing levels of difficulty of the face stimuli such as “change identity (change to a different person),” “change eyes (replace the original eyes with someone else’s eyes),” and “widen eye spacing (increasing the distance between the eyes).” In support of the perceptual learning view, 4-month-old infants showed significant discriminability only in the Taiwanese “change identity” condition (and not any others). At six months, their discriminability for Taiwanese faces progressed to the “change eyes” condition. At nine months, they begin to be able to distinguish among Caucasian and Filipino faces but only at the “change identity” level. In other words, young infants started out with selective discriminability for own-race faces, and their discriminability extends to other-race faces a few months later.

In the second study, we examined how own-race experience affects the development of perceptual discriminability in 5- to 12-year-old Taiwanese children. We used a morphing face paradigm, taking an Asian female face and a Caucasian female face to create a continuum of morphed female faces that ranged from more Asian/less Caucasian to less Asian/more Caucasian. This method gave us precise control over the physical similarity between two selected faces. In a series of trials, participants judged whether the “target” face (an Asian face in the Asian condition and a Caucasian face in the Caucasian condition) looked the same as one of the morphed faces.

We then estimated each participant’s discrimination threshold, the minimum physical difference needed to tell two faces apart, for both the Asian and the Caucasian conditions. We found that 11-12-year-old children exhibited a smaller discrimination threshold in the Asian condition than in the Caucasian condition, indicating an own-race advantage. Children aged 5 to 12 displayed a gradual improvement in lowering the threshold in the Asian condition, however, their discrmination thresholds in the Caucasian condition remained relatively unchanged. Hence, continued exposure to own-race faces does facilitate the perceptual discriminability for own-race faces significantly, while the facilitation for other-race faces is not as prominent.

In conclusion, our studies suggest that the perceptual learning view appears to best describe the developmet of children’s face perception. The own-race advantage seen in adulthood reflects a natural outcome of prolonged learning specific to the faces most commonly seen in one’s visual environment.


Dr. Sarina Hui-Lin Chien is Associate Professor at the Graduate Institute of Neural & Cognitive Sciences and the Graduate Institute of Biomedical Sciences at China Medical University in Taichung, Taiwan