Mouse Study Finds Prefrontal Circuits Can Reuse Working Memory

A study in mice has identified groups of prefrontal-cortex neurons that can be repurposed to hold different kinds of information in working memory. The finding offers evidence for a long-standing theory that the brain reuses computational circuits rather than assigning a dedicated neural group to every task.

The research, led by scientists at MIT and Princeton University, was published August 17, 2026, in Nature Neuroscience. The team trained mice to compare two tones—one higher or lower in pitch—and respond based on whether the sounds matched. While the animals performed the task, the researchers recorded electrical activity from thousands of neurons in the prefrontal and parietal cortices.

The scientists examined two intervals in the task. During the first, between the initial and second tones, the mice had to retain a memory of the first sound. During the second, after the second tone, they had to maintain a decision and the action they planned to take.

Neurons in the parietal cortex appeared to remain focused on storing the tone memory. In the prefrontal cortex, however, the researchers found a cluster that changed roles: it represented the first tone during the earlier interval and the planned action during the later one. Computational analyses were used to identify groups of neurons associated with these different types of information.

The researchers describe these groups as flexible modules. Rather than constructing an entirely new circuit whenever an animal learns a different task, the same neural population may perform a similar memory-related computation with different content. In this view, neural circuits function as reusable components that can be combined in different ways to support varied behavior.

The result comes from a mouse experiment, so it does not establish that the same flexible modules operate in humans. It also does not show that the identified neurons are necessary for the behavior. The researchers said they plan to test that question by inhibiting the modules at different points in the task and observing whether the animals’ performance changes.

The study builds on earlier work from Timothy Buschman’s laboratory showing that animals can assemble neural circuits to perform different parts of categorization tasks. Together, the work examines how a finite number of neurons might represent a wide range of information and support cognitive flexibility.