Schedule of Events | Search Abstracts | Invited Symposia | Symposia | Poster Sessions | Data Blitz

High-Level Visual Representations Drive Memory Errors in Older Adults

Poster Session B - Sunday, March 8, 2026, 8:00 – 10:00 am PDT, Fairview/Kitsilano Ballroom

Ricardo Morales-Torres¹ (), Loris Naspi², Simon W. Davis³, Roberto Cabeza¹; ¹Duke University, ²Humboldt University of Berlin, ³Indiana University

Aging tends to be associated with the decline of different cognitive domains, with episodic memory being the prime example. In contrast, domains such as semantic knowledge tend to be preserved. This latter pattern has inspired several functional magnetic resonance imaging (fMRI) studies that have shown that older adults exhibit stronger representations of the high-level/semantic features of stimuli. Moreover, the neural representations of these features are predictive of better memory performance. However, whether the same features that aid encoding might drive memory errors remains an open question for fMRI research. In the present fMRI study, we aimed to test whether visual memory errors in older adults are driven by high-level/semantic representations. In the current experiment, young and older adults encoded several images of daily objects (e.g., a dog). During retrieval, each object was paired with a lure (i.e., a different dog), and subjects had to choose which image was seen during encoding. Employing a recurrent convolutional neural network, we modeled the similarity of each target and its lure, from their low-level to high-level visual features. Our behavioural results indicate that visual recognition errors in older adults were predicted only by high-level visual similarity. Convergent results were obtained by employing representational similarity analysis during retrieval. Stronger representations of high-level visual features, in the ventral visual stream, predicted memory errors. None of these effects were observed in younger adults. Taken together, our results indicate that visual memory errors in older adults are strongly driven by semantic/high-level visual representations.

Topic Area: LONG-TERM MEMORY: Development & aging