Sanika Paranjape¹, Sarah Shomstein¹, Dwight Kravitz¹; ¹The George Washington University

Visual working memory (VWM) is a cognitive process by which visual information is temporarily maintained and manipulated, and is traditionally studied using basic low-level features (e.g., color, shape). The maintenance of simple features has been shown to involve the same neural circuits as perception (e.g., visual area V1 for orientation), and therefore affects ongoing feature perceptual processing (Teng & Kravitz, 2019; Kiyonaga & Egner, 2013). However, few studies have investigated these interactions for complex naturalistic stimuli, such as faces, and their associated complex decisions (e.g., trustworthiness). If VWM recruits the perceptual face network, there should be interactions between maintained faces and ongoing high-level judgments of other faces that scale with their physical similarity. Here, using an orthogonal dual-task paradigm (Teng & Kravitz, 2019) on Amazon Mechanical Turk, we test whether maintaining faces of darker skin tones engenders an implicit race bias that carries over to trustworthiness judgments of other physically similar faces. Preliminary results show a difficulty effect when faces of darker skin tones are held in working memory, causing an attraction bias towards faces with a similar structure to the memory stimulus. However, the observed bias is larger when the question has a negative bias (e.g., “Which person is the criminal?”) as opposed to positive bias (e.g., “Which person would you invest with?”), providing evidence of implicit bias influencing ongoing perception. Demonstrating that implicit bias in VWM affects ongoing perception provides strong support for the recruitment of the perceptual face network by VWM.

Acknowledgements: Research supported by NSF 2022572, NSF BCS-1921415, and BCS-2022572 to SS