A disk just 2 millimeters in radius held the virtual brain cells in a University of Tokyo experiment that found a surprising route toward a neural tipping point: inhibitory connections that reach farther can make critical activity more likely.
The brain criticality hypothesis says neural activity may sit near a boundary between two patterns. In a subcritical state, activity fades quickly. In a supercritical state, it can spread and grow. At the boundary, called criticality, a small burst may stay local or ripple across the network in an avalanche of activity. Researchers are interested in this balance because networks near criticality can have strong information-sharing and processing abilities. How the brain reaches that state remains an open question.
A team led by associate professor Zenas C. Chao at the International Research Center for Neurointelligence at the University of Tokyo built a computer model that included physical space. Earlier models often assumed neurons were randomly linked or fully connected, Chao said, while real neurons occupy space and signals take time to travel. The team also included University of Tokyo researchers Felix Benjamin Kern and Takahisa Date. Their study was published in PLOS Computational Biology on Oct. 1, 2026.
The model contained 100 neurons: 80 excitatory cells, which promote activity, and 20 inhibitory cells, which restrain it. They were scattered across a two-dimensional disk with a 2-millimeter radius. The researchers kept the excitatory connections fixed while changing the maximum reach and probability of inhibitory connections. This produced 25 combinations of inhibitory connection patterns. They tested each combination in 10 separately generated network prototypes, starting each simulation with synaptic weights at zero and running it for 15 hours.
The team examined the final two hours, once the networks had settled. Supercritical activity was the most common result across most settings. Criticality appeared less often, but was seen most frequently when inhibitory connections had intermediate-to-high density. At the highest densities, the networks were mostly subcritical, with activity tending to die out.
The researchers also found that the physical reach of inhibitory connections mattered: longer-reaching inhibition could push the simulated networks toward criticality. The result adds distance to a picture that has often focused on the balance between excitation and inhibition, or on how many connections a network has. It suggests that where signals can travel, not just how strongly neurons excite or restrain one another, may help shape the brain-like state.
This was a computer study, not a demonstration that changing connections in a living brain would produce the same effect. Still, the model offers a way to test how spatial wiring and changes in synapses work together. The next challenge is to understand whether these patterns help explain how real neural circuits develop and maintain a balance where activity can both settle down and spread.