Neural correlates of face learning: The N250 distinguishes newly-learnt from similar-looking novel faces
Authors: Alice Nevard, Soazig Casteau, Ulrik Beierholm, Holger Wiese 

Learning a new face can be described as gradual formation of a robust representation that supports recognition of an individual in novel contexts. Previous event-related potential (ERP) research has demonstrated that more negative amplitudes from approximately 200 ms onwards are associated with newly learnt faces compared to novel ones, suggesting a robust neural learning effect. The current study examines how representations of newly learnt identities are formed following a brief real-life interaction, compared to viewing novel matched (i.e. similar-looking) and unmatched (i.e. dissimilar) identities. Participants had a 10-minute conversation with one of four individuals they had not met previously. Participants then saw multiple ambient images with high within-person variability, of the “learnt”, an “unmatched” and a “matched” face while their EEG was recorded. N250 amplitudes (200-300 ms, occipito-temporal channels) were more negative for the learnt face compared to both the unmatched and matched identities, indicating the emergence of a distinct neural representation for the newly learnt face. Additionally, no differences were found between the matched and unmatched novel identities, indicating that both novel faces were similarly differentiated from the learnt identity. Furthermore, this ERP learning effect was carried over into later time windows showing sustained familiarity effects, and learning effects were generally stronger over the right hemisphere after 300ms. Together these findings indicate that brief real-world interactions allow to establish robust neural representations that support the recognition of newly learnt faces – even in comparison to similar-looking novel identities. 

Have we met before? Face learning over multiple interactions
Authors: Alice Nevard, Soazig Casteau, Ulrik Beierholm, Holger Wiese 

Learning a new face can be described as forming a robust representation over time that allows for recognition in novel situations. Event-related potential research has linked this process with more negative amplitudes for learnt compared to novel faces at occipito-temporal channels from ~200 ms onwards. While this N250 effect represents robust learning, its amplitude is typically small - reflecting limited exposure to novel faces, which are typically learnt in a single session. The current research examines how new face representations are established over multiple sessions by comparing the N250 after repeatedly meeting a new person. In pre-training sessions, participants saw highly variable ambient images of the “to be learnt face”, an “unmatched face” and a “matched” identity, the latter resembling the first with respect to age, gender, hair colour and ethnicity. In subsequent sessions across four days, participants first had a ten-minute in-person interaction with the same learnt identity, followed by a test phase in which participants saw ambient images of the three identities. While new images were used in all sessions, the learnt and unmatched identities were repeated. However, a new matched identity was presented in each session. There were no differences in the N250 in the pre-training session. However, after one interaction, N250 responses were larger for the learnt face compared to the unmatched identity. Critically, this effect was substantially enhanced after four interactions. Moreover, learning effects in comparison to the matched faces were generally small. Our findings show how representations of novel faces build up with repeated exposure. Importantly, while in-person interactions provide sufficient information for building such representations over multiple sessions, repeatedly viewing ambient images of the same person does not.