Abstract
Stimuli in natural vision differ in various physical attributes such as size, location, and surface feature, posing a unique challenge for visual cognition. However, previous research in visual working memory (VWM) mostly used simple and well-controlled laboratory stimuli (colors with different hues but matched saturation), making it difficult to assess how diverse stimulus features affect VWM. To examine this issue, this study developed a two-dimensional face space with continuous variations in the circular dimension for face identities and in the radial dimension for feature saturation. The center of this face space represents the average (norm) face. The radial distance of a face represents its strength/intensity, similar to color saturation defined as the eccentricity of a color in standard color space. In Experiment 1, participants remembered either 1 or 2 faces randomly sampled from the face space across different radial and circular locations and recalled the face on both dimensions after a short delay. Modeling of recall errors on the circular axis showed that VWM precision was reduced by set size and increased by face saturation. However, the number of retained VWM items was only significantly influenced by set size but not by face saturation. Furthermore, participants’ recall on the radial axis showed more centripetal errors at a larger set size, suggesting a general bias toward the norm. To examine how face saturation affects neural activities underlying VWM maintenance, in Experiment 2, we recorded electrocorticography signals along the ventral processing pathway in epileptic patients during the face VWM task. Representational similarity analyses showed that face saturation strengthened the association between delay-period neural activities and retained VWM representations estimated by behavioral recall performance. Together, these preliminary findings reveal novel neurocognitive effects of feature saturation on VWM maintenance, suggesting that saturation may be a critical aspect of VWM representations beyond capacity and precision.
Acknowledgement: NINDS/NIH Visiting Fellows Program