Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster D in Poster Session D: Wednesday, August 5, 2:00 – 3:45 pm, Kimmel Center, Shorin & Rosenthal Rooms

Two-state dynamics structure working memory in the human medial temporal lobe

Zelin Yu¹, Nuttida Rungratsameetaweemana¹; ¹Columbia University

Presenter: Zelin Yu

Working memory (WM) is traditionally thought to rely on persistent neuronal firing during maintenance. Recent primate work has begun to reveal additional temporal structure, suggesting that information may be intermittently expressed during maintenance. Whether such dynamics exist in humans remains unclear. We analyzed single-neuron activity from the human medial temporal lobe (MTL) during verbal and visual WM tasks. Using time-resolved decoding with linear classifiers, we found that load information was not continuously represented during maintenance, but instead fluctuated between periods of high and near-chance decodability. These fluctuations indicate that WM representations are temporally structured and intermittently expressed at the population level. Modeling these dynamics with hidden Markov models (HMMs), corresponding to an "on" state with high decodability and an "off" state with near-chance decodability, better captured the data than a one-state model. Together, these results suggest that human WM is not supported by sustained activity alone, but instead involves dynamic transitions between active expression and quiescence in population coding, potentially balancing accessibility and efficiency.

Topic Area: Memory, Learning & Knowledge Structures