Search Papers | Poster Sessions | All Posters
Poster C86 in Poster Session C - Friday, August 9, 2024, 11:15 am – 1:15 pm, Johnson Ice Rink
Pattern separation using compressed and semantic representations of memory
Dale Zhou1, Sharon Noh1, Michael Yassa1, Aaron Bornstein1; 1University of California, Irvine
Recognizing whether information is novel is crucial for many behaviors. For example, decision-makers need to discriminate between perceptually or mnemonically aliased states when planning through uncertain contingencies. In the mnemonic similarity task, images are judged as novel, similar, or old when the presented image was never seen, already seen, or slightly altered as an image "lure." Diminished lure discrimination has been diagnostic of poorer memory in clinical settings, predictive of imprecision in decision-making, and indicates altered pattern separation computations in the hippocampus. However, it is unknown what properties of stimuli drive difficulties in discrimination and whether their influence varies across individuals. We hypothesize that the lure discrimination is related to the image properties of lossy compression, semantic similarity, and intrinsic memorability. Consistent with our hypothesis, we find that participants (n=366) perform better when original and lure images have lossier compression, greater semantic distance, and larger differences in intrinsic memorability. Sensitivity to these image properties tends to worsen with age. Perceptual, semantic, and mnemonic differences may construct distinct memory representations to support pattern separation.
Keywords: memory efficient coding lossy compression