Within-session repeated response or rule switching in rodents and marmosets

Cognitive flexibility enables the suppression of a current choice or strategy in favor of a more adaptive alternative when context shifts. Its impairment is a transdiagnostic symptom across multiple neuropsychiatric disorders. To investigate the neural basis of flexibility, we recorded large neuronal ensembles from the medial frontal cortex of mice and rats during rule or response switching, revealing dynamic activity patterns aligned with these switches. In the primate brain, how the cortical networks support flexibility is less well-understood.

We addressed this gap by training marmosets on a novel feedback-based, intra- and extra-dimensional target switching task. After pretraining on color patches and black shapes separately, on their first exposure to colorful shapes, all marmosets learned to choose the compound stimulus containing the target feature and then switched to a new target feature at least once. With practice, they were able to switch both intra- and extra-dimensionally among the features up to 9 times within session and readily transferred this performance to new feature sets.

To better understand their task performance, we tested 3 different Feature Reinforcement Learning models. The marmosets’ estimated learning rate, choice determinism and dimension-related parameters fell within the range observed in healthy humans, highlighting the translational value of our task and paving way to in-depth neural mechanistic analysis.

Hybrid Zoom link: https://utoronto.zoom.us/j/86328669091