Perceiving a single face in a crowd: Insights from image reconstruction

Poster Session F - Tuesday, April 16, 2024, 8:00 – 10:00 am EDT, Sheraton Hall ABC

Marco Sama¹ (marco.sama@mail.utoronto.ca), Moaz Shoura¹, Adrian Nestor¹, Jonathan Cant¹; ¹University of Toronto Scarborough

Face ensemble processing – or crowd perception – involves synthesizing a summary representation (e.g., an average expression) across an array of faces which circumvents visual working memory limitations. However, little is known about how single face targets are processed within an ensemble. To address this, in our study participants viewed face ensembles either without a central face (n = 32), or with a central face while attention was focused centrally (n = 38) or distributed across the entire ensemble (n = 38). Critically, the emotional expression of central faces was either consistent or inconsistent (i.e., having either the same or different valence for central versus surround faces, respectively). Participants completed an expression similarity-rating task between ensembles and single-face probes. Similarity data were then used to generate an expression-based face space and to yield image reconstructions of summary ensemble representations. Reconstructions were compared against the central face, the surrounding exemplars, and the pixelwise surround average. We found that focused attention enhances target face representation without altering that of the surround. We also found that the ensemble summary dominates ensemble percepts, as revealed by reconstruction, but not during focused attention to target faces within inconsistent ensembles (e.g., happy target face surrounded by faces with negative valence). These findings not only provide a more robust understanding of crowd perception, but offer insights into the interaction between single face targets and surrounding ensembles. Critically, our work is among the first to show how single faces are represented within the context of ensemble processing.

Topic Area: PERCEPTION & ACTION: Vision