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

Theta Oscillations Shape Load-Dependent Working Memory via Inhibitory Synchronization

Renee Tung¹, Robert Kim², Nuttida Rungratsameetaweemana¹; ¹Columbia University, ²Cedars-Sinai Medical Center

Presenter: Renee Tung

Working memory (WM) enables short-term maintenance of task-relevant information to support goal-directed behavior. While advances in human intracranial recordings have begun to elucidate the circuit computations underlying WM, the mechanisms by which they adapt to or fail under increased cognitive demand remain poorly understood. In addition, the specific contributions of excitatory and inhibitory populations and the role of local field potential (LFP) dynamics in complementing single-neuron activity have not been characterized. To address these gaps, we analyzed a dataset of human intracranial recordings as neurosurgical patients performed a verbal WM task with varying load, and found that LFP theta power was higher during memory maintenance under increased load. We then developed biophysically realistic spiking recurrent neural networks (RNNs) with excitatory and inhibitory units that received either theta-frequency LFP-like input alongside task input or task input alone. Models trained with theta oscillatory input during maintenance outperformed those trained without it on a longer-maintenance WM task. Furthermore, when inhibitory units did not receive oscillatory input, model performance dropped to chance levels. These findings suggest that theta synchronization of inhibitory units is critical for sustaining WM under increased cognitive demand.

Topic Area: Memory, Learning & Knowledge Structures