Contributed Talk Sessions | Poster Sessions | All Posters | Search Papers

Poster Session A: Tuesday, August 12, 1:30 – 4:30 pm, de Brug & E‑Hall

Functional Role Division in Working Memory Emerges from Intrinsic Neural Timescale Diversity

Tomoki Kurikawa¹; ¹Future University Hakodate

Presenter: Tomoki Kurikawa

Neural activity spans a wide range of timescales, both across and within cortical areas. However, the functional significance of this temporal diversity remains poorly understood. In this study, we investigate how heterogeneity in neural timescales within the frontal cortex—a hub for higher cognitive functions—supports performance in a context-dependent working memory task. Specifically, we model a delayed match-to-sample task with context cues, which requires the maintenance of information over time and flexible adaptation of behavior based on contextual input. We construct a recurrent neural network (RNN) composed of units with varying intrinsic time constants to perform this task. Our analysis reveals that task performance is optimized when neural timescales are appropriately balanced, consistent with experimental observations. Notably, neurons with slower dynamics play a causal role in improving task performance despite not showing stronger selectivity to task-relevant signals compared to faster neurons. In contrast, fast neurons encode task-relevant information more precisely but only transiently. These findings suggest that diversity in neural timescales supports a functional division of labor, enabling stable memory maintenance and flexible signal encoding. This work provides a mechanistic account of how temporal heterogeneity in neural populations can support complex cognitive computations.

Topic Area: Memory, Spatial Cognition & Skill Learning

Extended Abstract: Full Text PDF