NMDAR BLOCKADE REDUCES ENGAGEMENT BY PROMOTING CHOICE REPETITION IN MICE PERFORMING AUDITORY DECISION-MAKING

During perceptual decision-making, relevant stimulus evidence is temporally integrated in circuits representing each choice until commitment. Previous modelling work has proposed that N-methyl-D-aspartate receptors (NMDARs) play a key role in the circuit dynamics that integrate evidence leading to choice. In addition, NMDARs have been suggested to underlie the dependence on past choices via an action reinforcement mechanism. Nevertheless, it remains unclear whether and how NMDARs impact mouse choices in a psychophysical task. To test that, we developed a two-alternative forced-choice (2AFC) task where head-fixed mice had to discriminate the loudness of two amplitude-modulated sounds presented on each side of the head (left or right). Psychometric analysis revealed that all mice exhibited a significant fraction of lapses, defined as responses non-guided by the stimulus. We modeled mice's behavior using a Hidden Markov Model (HMM) with two states and three covariates each: the stimulus, the action trace (a weighted sum of previous choices), and a fixed bias. After fitting the model, an engaged and a disengaged state emerged, relying mostly on the stimulus and the action trace, respectively. Under systemic NMDAR blockade using MK-801, the HMM revealed that, although the time spent in each state was similar between the drug and saline conditions, the disengaged state showed reduced stimulus and increased action trace weights. Our results point to NMDAR hypofunction generating task deficits through an enhancement of repetitive behavior, expressed as task disengagement, rather than a disruption in the stimulus representation or processing.