Shaofeng Liu¹, Eric Nemrodov², Kinkini Monaragala¹, Ilya Nudnou³, Dan Nemrodov¹, Matthias Niemeier¹, Adrian Nestor¹; ¹University of Toronto, ²University of Waterloo, ³North Dakota State University

Visual imagery is thought to recruit some of the same neural machinery as perception, but how identity-level representations unfold over time during imagery, and how they relate to perceptual processing, remains unclear. Here we used EEG and multivariate decoding to compare the spatiotemporal profiles of face perception and imagery. Seventeen participants viewed 25 famous and 25 unfamiliar female faces shown upright or inverted, and later generated mental images of a subset of five highly familiar identities cued by their names. We used temporally cumulative and time-resolved analyses to decode face identity, along with multivariate channel selection, representational similarity analysis, and cross-temporal generalization to characterize the underlying code. Identity could be reliably decoded from neural signals elicited by both perceived and imagined faces. Perceptual decoding emerged early and was modulated by both face inversion and familiarity, which contributed largely independently to performance. Imagery-based decoding showed a markedly later onset and more extended time course, but relied on scalp topographies that overlapped with those supporting familiar upright face perception. Across the five imagery targets, representational dissimilarity matrices for perception and imagery were significantly correlated, indicating a partially shared identity space despite differences in temporal dynamics. Together, these findings demonstrate that EEG carries fine-grained information about individual face representations not only during perception but also during voluntary imagery, and they highlight both common and distinct organizational principles for perceptual and mnemonic face codes.