Shruti Nishith^1,2, Nicholas B. Turk-Browne^1,2; ¹Department of Psychology, Yale University, ²Wu-Tsai Institute, Yale University

Selective attention has well-known effects on perception and working memory, but its influence on long-term memory remains underexplored. Here, we characterize how feature-based attention modulates the encoding of competing visual features of an object into long-term memory. Objects were created by sampling shape and color values from circular, perceptually uniform feature spaces. Each color-shape object was paired with a unique face image to form an associative memory. Participants completed two sessions in which they were instructed to attend to the color or shape of the objects, respectively. Each session started with six rounds of learning, with each round consisting of three phases: (1) study phase to memorize associations between the attended feature and the face, (2) feature test phase to probe memory for the attended feature using a shape/color wheel, and (3) face test phase to probe memory for the associate face. Across learning rounds, participants significantly improved their memory performance in the feature test and the face test. The session ended with a final test phase in which participants were cued with a face and had to recreate both the color and the shape of the associated object from long-term memory (regardless of which feature they had attended to during learning). Participants showed lower mean error for color memory when color had been attended versus unattended during learning and lower mean error for shape memory when shape was attended versus unattended during learning, resulting in an interaction between attended feature and the feature memory. There was also a main effect of feature type, with greater memory accuracy for color than shape. Objects varied in the similarity of their features to those of other objects, allowing for additional analyses of how feature-based attention modulates memory competition and interference. These findings suggest that task-relevant features are prioritized in visual long-term memory.