Ting-Yun Chang¹ (), Oakyoon Cha², Yanchao Bi^3,4,5, Isabel Gauthier⁶; ¹State Key Laboratory of Cognitive Neuroscience and Learning & IDG/McGovern Institute for Brain Research at Beijing Normal University, China, ²Department of Psychology, Sogang University, South Korea, ³School of Psychological and Cognitive Sciences and Beijing Key Laboratory of Behavior and Mental Health, Peking University, China, ⁴IDG/McGovern Institute for Brain Research, Peking University, China, ⁵Institute for Artificial Intelligence, Peking University, China, ⁶Department of Psychology, Vanderbilt University, USA

Ensemble Perception (EP) is the ability to extract summary statistics from groups of objects. While pairwise correlations have suggested that EP for faces and simple features may rely on distinct mechanisms, recent multivariate work revealed a domain-general EP ability spanning low- and high-level nonfacial features, one that correlates with constructs associated with general intelligence (g). Two questions remain: Does this domain-general ability extend to facial features, and is it separable from g? We administered seven EP tasks—three facial (identity, expression, age), two low-level (orientation, aspect ratio), and two high-level nonface (robots, birds)—alongside standard measures of processing speed, reasoning, and crystallized intelligence. Confirmatory factor analyses showed that low- and high-level nonface EP tasks loaded onto a single factor, and the face and object EP factors were perfectly correlated, suggesting a common EP mechanism. Although CFA models including g produced improper solutions, likely due to strong collinearity, results indicated that EP and g share substantial variance. To examine the relationship between EP and g, we conducted regression analyses controlling for age and gender. For object EP, performance on each task (simple or complex) was uniquely predicted by both g and the aggregate performance of tasks at the other level, indicating that object EP tasks were related above and beyond the g variance. For face EP, g uniquely predicted only the facial expression task. Each face task was significantly predicted by the aggregate across the four object tasks, but not by the other face tasks, suggesting no face-specific variance once object EP variance was accounted for. A small portion (1－7%) of EP performance was explained by an ability distinct from g. These findings support a domain-general EP ability that spans facial and nonfacial tasks and indicate that this ability, while largely explained by g, may include a specialized component specific to ensemble processing.

Acknowledgements: The STI2030-Major Project (2021ZD0204100) awarded to Yanchao Bi; the David K. Wilson Chair Research Fund at Vanderbilt University and NSF Award 2316474 awarded to Isabel Gauthier; and the NSFC Grant 32500938 and the Postdoctoral Fellowship Program of China Postdoctoral Science Foundation (GZC20240129 and 2025T180944) awarded to Ting-Yun Chang.