A robust neural index of automatic generalization across variable natural views of familiar face identities

Abstract

ABSTRACT The ability to automatically and rapidly recognize a familiar face identity across different views is well documented but generally measured with explicit behavioural tasks involving many other processes. Here we provide a sensitive implicit neural measure of this ability using electroencephalographic (EEG) recordings coupled with fast periodic visual stimulation (FPVS). Images of two famous face identities alternated for 15 seconds at a frequency rate of 6 Hz, preceded by a 15 seconds adaptation period to images of one of the face identities. Beyond previous studies, numerous (N = 60) non-edited variable natural face images of each identity were presented, with different images appearing at adaptation (N = 30) and test. At the group-level (N = 16), 3 stimulation sequences only were sufficient to elicit highly significant EEG peaks at odd harmonics (3 and 9 Hz), reflecting face identity discrimination across variable novel views. This response was maximally localized over the right occipito-temporal region and statistically significant in 75% of individuals tested. Importantly, the face identity discrimination response at odd harmonics was significantly larger than when the exact same stimulation sequences were preceded by images of a face identity not present in the alternating sequence (irrelevant adaptation condition). Overall, this study provides a highly sensitive and objective technique to implicitly measure an individual’s ability to generalize identity across different views of a familiar face identity.