Neural dynamics of flexible representations in working memory

Working memory (WM) involves forming temporary mental representations that persist beyond external sensory stimuli. These representations are not static reproductions of initial inputs, but can be shaped by both past information and future requirements: by modulation from priors, and anticipation of reconfiguration for new contexts. Flexible WM representations appear to possess characteristics that are well-aligned to, and hence, be supported by, the dynamical principles broadly observed in neural activity, such as persistence without inputs, history-dependence, and state-switching. However, the link between dynamical features of neural activity, and flexible WM representations, remains unknown. I propose to study the neural dynamics of WM, under three Aims: (1) developing novel WM tasks that manipulate the influence of priors, and contexts, (2) in task-trained rats, record and model neural activity under dynamical systems theory, which provides a mathematical framework for understanding dynamical features in recorded activity, (3) using patterned optogenetic perturbations for fine control of neural dynamics in causal tests of their links to WM representations. This project will reveal how neural dynamics support WM, but also advance principles and methodology relevant to neural dynamics in other flexible behaviors, with implications for controlling neural dynamics in applications such as neuro-prosthetics and brain-machine interfaces.