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

Sketchpad Series 

Priority-Driven Transformation of Visual Working Memory Content

Poster Session C - Sunday, March 8, 2026, 5:00 – 7:00 pm PDT, Fairview/Kitsilano Ballroom

Jung Woo Hur¹ (), Brad Postle¹; ¹University of Wisconsin-Madison

The control of priority in visual working memory is a hallmark of the flexible control of behavior. To study this mechanism, we trained recurrent neural networks (RNNs) to perform a double-serial retrocuing (DSR) task in which two stimulus items (item A and item B) are presented serially, followed by a first cue (100% valid) indicating which will be tested at the first recall epoch, then a second cue (100% valid), indicating which will be tested at the second recall epoch. In order to visualize the priority-based transformations of the representational geometry, we applied demixed principal component analysis (dPCA) to the hidden layer during post-stimulus epochs and derived mnemonic subspaces. Hidden layer activity data were then projected to these subspaces to track the representational transformations of the stimulus identities. Following the results from previous research in our lab, which provided evidence that the unprioritized memory item (UMI) adopts an “opposite” representational format to that of the prioritized item (PMI), we expect the representational geometry of the UMI to be ‘flipped’ (i.e., reflected) relative to the PMI following the priority signal (i.e., the retrocue). Our research will provide important insights into our understanding of the encoding of prioritization in visual working memory.

Topic Area: EXECUTIVE PROCESSES: Working memory