Abstract
Individual differences in human vision are ubiquitous, but poorly understood. Here, we aimed to understand the neural dynamics and representational code underlying outstanding visual processing abilities. We recruited 15 “super-recognizers” (SRs; individuals in the top 2% of face-recognition ability spectrum) and their matched-controls. Participants completed two visual tasks totalling >80,000 trials per group while we measured their brain activity with high-density EEG. We performed multivariate analyses on the time-resolved brain patterns of both groups while they identified newly learned faces. Specifically, we produced the time-course of task-related representational distances between face identities. This analysis revealed more distinct identity representations in SRs after the first feedforward sweep of the visual system (200-500ms after face-onset), accompanied by stronger face-identification performance of experimentally learned identities. These results indicate that the real-life diagnostic advantage for faces in SRs is associated with richer brain representations for face stimuli. Next, we ask whether this superior visual processing extends beyond faces, i.e. whether it is domain-general. We assessed this in a second EEG experiment during which Representational Similarity Analysis was used to characterise the representational code behind their processing of a wide set of visual stimuli (including objects, animals, scenes). This showed increased distinction between face vs non-face category representations in SRs, but also revealed more distinct representations within non-face categories, indicating an enhancement that extends beyond face stimuli. Considering the time-course of all these findings (>200ms), we propose that super-recognition is underlied by visual representations that are enriched by a recurrent, domain-general mechanism. This study provides the first evidence for a direct link between real-life visual abilities and the richness of stimulus representations in the human brain. In addition, our findings suggest that the general quality of a person’s representations, even of simple objects, predicts their ability to recognize faces of different individuals around them.