Decoding representations of personally familiar others in the default mode network

Abstract The human brain’s ability to form and maintain rich internal representations of personally familiar others, even in the absence of direct sensory input, is a prominent feature of social cognition. We tested whether the identity of familiar individuals can be decoded from neural activity patterns elicited during internally generated reflection. Using a round-robin fMRI design ( N = 113; 20 groups of 5–6 participants), subjects completed a trait-judgment task requiring yes/no decisions about whether adjectives described themselves or specific members of their social group. Multivariate searchlight classification revealed significant above-chance decoding of target identity throughout the default mode network. This decoding was observed in the absence of faces or voices, suggesting that the discriminative signal reflects abstract person knowledge rather than perceptual identity cues. Decoding performance was robust across endorsement conditions, with similar results observed for trials involving endorsed and nonendorsed traits. Together, these findings indicate that the default mode network supports stable, internally generated representations of personally familiar others that remain detectable independent of trait endorsement.