Decoding How Brain Circuits Control Behavior Through Spiking Neural Networks
Everyday perception is relentlessly anticipatory. When a driver spots a yellow traffic light, an approaching siren sounds in the distance, or footsteps echo in a hallway, the human brain instantly constructs a multifaceted forecast. It predicts what sensory event is coming, exactly when it will occur, and the statistical likelihood of its arrival.
Despite how effortlessly our minds calculate these variables, computational neuroscience and artificial intelligence have long struggled to replicate them biologically. Traditional machine learning models typically segregate these dimensions into separate, specialized modules or rely on backpropagation—a mathematically powerful training algorithm that requires globally coordinated feedback signals difficult to reconcile with the physical wiring and metabolic constraints of biological synapses.
Now, a groundbreaking study led by Associate Professor Zenas C. Chao and Academic Specialist Yohei Yamada at the International Research Center for Neurointelligence (WPI-IRCN) at the University of Tokyo has introduced a unified, biologically grounded alternative. Published in Communications Biology, their research demonstrates that a single recurrent population of spiking neurons can simultaneously learn and update event identity, timing, and probability using strictly local learning rules.
What is Neuroscience? – Introduction to Neuroscience
To understand the significance of this discovery, it is essential to establish what neuroscience is and how it investigates biological information processing. Neuroscience is the multidisciplinary scientific study of the nervous system, encompassing the brain, spinal cord, and peripheral neural networks. It bridges molecular biology, anatomy, physiology, psychology, and computation to answer one of humanity's ultimate questions: How does human intelligence arise?
Within neuroscience, systems neuroscience and computational neuroscience focus specifically on how neural circuits process information, represent variables, and control behavior. While neuroanatomists map the physical wiring of the brain, computational neuroscientists construct mathematical and biophysical models of neurons and synapses. By simulating how populations of neurons fire, adapt, and communicate, researchers can test hypotheses about cognitive functions such as learning, memory, perception, and decision-making—bridging the gap between microscopic cellular activity and macroscopic behavioral control.
Decoding How Brain Circuits Control Behavior: The Multi-Event Expectation Task
The brain predicts not only what will occur, but also when it will occur and with what probability. Researchers at the University of Tokyo refer to this joint representation of identity, timing, and frequency-based probability as a complete prediction object.
To examine how neural circuits manage these complex expectations, the Tokyo team constructed a recurrent spiking neural network (RSNN) consisting of 1,000 interconnected neurons. They designed a rigorous "Multi-Event Expectation Task" in which the network was presented with a brief sensory cue predicting one of two subsequent target events. Crucially, each event possessed its own discrete delay window, and the probability of each event occurring was independently modulated. Learning was confined strictly to the synaptic weights of the network’s readout connections.
When trained across 100-trial blocks, the network integrated all three predictive dimensions without interference:
- Magnitude Reflects Likelihood: When the probability of an event increased, the internal firing activity representing that expectation scaled up proportionally.
- Temporal Shifting: When the anticipated delay changed, the network dynamically adjusted the timing of its anticipatory burst to match the new interval.
- Factorized Within One Circuit: Instead of assigning "what" and "when" to separate clusters of cells, the network formed factorized, overlapping patterns within the exact same neural population.
Remarkably, the network achieved the bulk of its predictive precision within the first 50 trials of a training block, illustrating rapid, sample-efficient learning.
Local Learning vs. Global Backpropagation in Brain Circuits
A central challenge in neuro-inspired computing is reconciling the brain's physical architecture with learning algorithms. Artificial neural networks typically rely on backpropagation through time (BPTT), which requires global error signals and precise credit assignment across deep network layers. Biological brains, however, lack centralized clocks and global error broadcast mechanisms.
Instead, biological synapses rely on local plasticity rules—such as spike-timing-dependent plasticity (STDP) and neuromodulator-gated Hebbian learning—where synaptic updates depend solely on the pre- and post-synaptic activity of directly connected neurons, modulated by local chemical signals like dopamine or acetylcholine representing prediction errors.
In their Communications Biology study (DOI: 10.1038/s42003-026-10836-2), Yamada and Chao employed an error-modulated, attention-gated Hebbian rule within the recurrent spiking network. This biologically plausible mechanism allowed the network to acquire and flexibly maintain complete prediction objects without ever invoking global backpropagation. The model demonstrates that local synaptic plasticity is fully sufficient to coordinate complex temporal and identity predictions across a shared neural pool.
Factorized Subspaces and Dynamic Environmental Volatility
Real-world environments are characterized by constant volatility: a traffic delay might lengthen, or a previously reliable outcome can suddenly become rare. To test cognitive flexibility, the investigators abruptly scrambled timing intervals and outcome probabilities during simulations.
Following these sudden statistical shifts, prediction errors spiked immediately, but the network rapidly adapted its internal model, re-stabilizing its anticipatory representations within a small number of trials. When benchmarked against alternative computational frameworks—including global least-squares optimization and models lacking online temporal updates—the local learning network displayed superior stability and faster recovery times.
Further analyses revealed that probability and timing self-organized into factorized subspaces within the shared population of 1,000 neurons. Rather than requiring distinct anatomical modules for "what" and "when," the recurrent circuitry multiplexes these variables into overlapping, mixed-selective population codes—mirroring how cortical circuits in biological brains achieve high-dimensional cognitive flexibility.
Implications for Neuromorphic Computing and Brain-Inspired AI
While the authors emphasize that the model serves as a computational proof of concept rather than direct physiological proof of how biological brains operate, the discovery holds profound implications for artificial intelligence and neuromorphic engineering.
Traditional AI architectures demand immense computational power and energy-hungry data centers for training via backpropagation. In contrast, neuromorphic hardware designed around local error-modulated plasticity can operate on low-power edge devices, enabling autonomous robots, prosthetic interfaces, and IoT sensors to learn, anticipate, and adapt to unpredictable real-world data streams in real time.
By connecting everyday cognitive anticipation with biologically plausible local learning rules, University of Tokyo researchers have established a robust computational foundation. This framework opens new pathways for investigating how biological brain circuits control behavior and how next-generation artificial intelligence can achieve true adaptive autonomy.
decoding how brain circuits control behavior through spiking