THE EFFECT OF SYSTEMIC KETAMINE ON WORKING MEMORY HISTORY DEPENDENCIES

Working memory (WM) responses in primates are attracted to recent stimuli (serial dependence) and to the mean of the long-term distribution of stimuli (long-term history dependence). In humans, attractive serial dependence is reduced in schizophrenia and autoimmune anti-NMDAR encephalitis, conditions linked to NMDAR hypofunction, and in autism, whereas dyslexia is associated with reduced long-term history dependence. These distinct patterns highlight the importance of identifying the mechanisms underlying WM history-dependent biases.

Here, we study the mechanisms underlying WM history dependence in four macaque monkeys performing a biased visuospatial oculomotor delayed response task. In each session, stimuli followed a bimodal Gaussian distribution, with two diametrically opposed mode locations, which varied by session. In some sessions, monkeys were administered ketamine, an NMDAR antagonist, systemically.

Using linear models, we quantified serial and long-term history dependence in behavioral responses. Surprisingly, monkeys exhibited mostly repulsive serial and long-term history biases. Ketamine reduced repulsive serial dependence but increased repulsive long-term history bias, suggesting distinct mechanisms for the two processes.

To investigate candidate circuit mechanisms, we trained excitatory–inhibitory recurrent neural network models to perform the same task under identical stimulus statistics. The models incorporated biophysically motivated mechanisms, including neuronal adaptation, synaptic plasticity, and NMDA receptor–mediated currents. To mimic ketamine, we perturbed trained networks by removing NMDA currents and reducing adaptation. Simulations reproduced the opposing effects of ketamine on serial and long-term history dependence, providing a mechanistic hypothesis of how NMDAR hypofunction differentially alters these two WM biases.