Temporal Dynamics of Ensemble Face Representations Under Magnocellular- and Parvocellular-Biased Visual Inputs: A Time-Resolved EEG Decoding Study

Abstract

Highlights What are the main findings? Gender information from M-biased face ensembles was reliably decodable early in processing and showed significant early temporal generalization. Under M-biased stimulation, ensemble gender information generalized to single-face judgments during late processing. What are the implications of the main findings? Ensemble face-gender processing shows a multistage temporal organization, with gender information becoming decodable early and remaining available at later stages. Shared task-related information between ensemble and single-face judgments during late processing suggests a potential link between ensemble representations and item-level gender information. Abstract Background/Objectives: Ensemble perception enables the visual system to rapidly extract summary statistics from multiple objects. However, the temporal dynamics of ensemble average representations and their dependence on magnocellular-biased (M-biased) and parvocellular-biased (P-biased) visual information remain unclear. This study examined when ensemble face-gender information could be decoded under these two stimulus conditions. Methods: Twenty-four healthy adults performed a two-alternative forced-choice gender judgment task with neutral face ensembles and single faces. M-biased stimuli were low-luminance-contrast grayscale faces, whereas P-biased stimuli were isoluminant red–green faces. Time-resolved multivariate pattern analysis, temporal generalization, and cross-condition generalization characterized gender-related EEG dynamics. Results: Behaviorally, responses were faster for P-biased than M-biased stimuli and for single faces than ensembles, with no significant accuracy differences. Balanced Integration Scores were significantly higher for P-biased than M-biased stimuli in both ensemble and single-face conditions. M-biased ensemble gender information was reliably decodable during early (95–300 ms) and late processing (775–930 ms), with significant early temporal generalization. P-biased face ensembles showed no significant cluster-corrected time-resolved decoding but exhibited significant temporal generalization during middle-to-late processing. Under M-biased stimulation, ensemble gender information generalized to single-face judgments from 690–995 ms. Conclusions: Ensemble face-gender information showed a multistage temporal organization. M-biased ensemble information became decodable early. Late cross-condition generalization further indicated that ensemble and single-face judgments shared gender-discriminative task information. These findings characterize the temporal evolution of ensemble gender information under pathway-biased visual stimulation and provide a temporal perspective on ensemble perception.