Neurobiologists have concluded that waves of electrical activity moving through the brain carry out the work of vision. These spreading brain waves, the authors assert, shape the world as perceived by a person. The findings were published in the journal Neuron.
This gives a long-overlooked signal a central role in perception. It also helps explain why an object can be in plain sight and yet go unnoticed, depending on where exactly the wave happens to be.
For decades, these waves seemed like a quirk of an anesthetized brain. Researchers largely assumed they disappeared once an animal woke up.
That assumption fell apart in 2020. That year, John Reynolds from the Salk Institute for Biological Studies led a team that disproved it. His group recorded waves traveling through the visual cortex of awake, active animals. This is a layer of tissue at the back of the brain responsible for sight.
These were marmosets, small monkeys trained to detect faint targets on a screen. The waves arose on their own, without any objects on the screen to trigger them. They spread at 0.6 meters per second, matching the speed of a signal traveling along a bare nerve fiber.
The position of the wave when a target appeared predicted whether the animal would report seeing it. That study turned waves from a random feature of recordings into something with a real impact on behavior.
Reynolds said, “When a traveling wave brings a region of the visual cortex into a more excitable state, neurons in that area respond more strongly to a weak visual object, and that object is much more likely to be seen.”
A trickier question remained. No one could say why the brain would create these waves in the first place.
The answer begins with anatomy. Only about 5% of the connections feeding a neuron in the visual cortex carry signals from the eyes. Roughly 80% of connections are formed through horizontal links. These fibers run sideways, connecting a neuron to other neurons in the same tissue layer. Most of them come from cells located in a small nearby patch.
These lateral fiber cables work slowly. A signal takes longer to reach a distant neighbor than a close one. Local wiring and this delay are enough to create a wave. Thus, the visual cortex is less of a channel passing images upward. Most of its neural connections interact with itself, and these signals reflect what that internal dialogue looks like from the outside.
An earlier review by several of the same authors described similar patterns extending far beyond the visual system. The new approach brings together biology and computer modeling into a unified whole. No one has assembled these pieces into a single framework before. It assigns four tasks to each of these directions.
Two of them are tied to the present moment. Waves adjust perception from moment to moment and embed recent information into the internal picture of what is happening. The other two types of waves stretch across time. Waves generate short-term predictions about what will happen, and they store and replay sequences that stand in for memories of events.
This claim about prediction has already been tested in a simulation. In an earlier study, a neural network was built like the cortex, with local connections and delays that increased with distance. A flash of light at the input triggered a wave.
Software that analyzed network activity after the flash disappeared could pinpoint exactly where and when it had appeared. Removing the lateral connections caused performance to drop to chance levels. Randomizing the connections preserved the timing but lost the location.
After training on a short video clip, the neural network generated the rest of the image using its own settings. It produced up to 100 future images without any additional input. Shuffling the connections and delays made both the waves and the prediction vanish simultaneously.
“The new idea is that these waves themselves perform computations,” Reynolds explained.
Waves are only half the story. The strength of the connections that generate them changes with use, and the authors argue that this is where information from the outside world is recorded. Every sight and every movement slightly adjusts these connections. Over months and years, this network absorbs everything that happens in the surrounding environment.
Objects keep their shape in three dimensions. Images slide across the retina with every movement. Physics behaves the same as yesterday. None of this needs to be recalculated from scratch if the neural connections already assume it.
“You never experience the world directly. Your brain builds an internal model of the world based on experience, expectations, and sensory data,” Reynolds explained.
He draws a parallel with large language models. These systems learn statistical patterns in text and then produce text that matches them. A generative model built from brain tissue would do the same for vision. It would manage the world rather than simply labeling static images.
The practical side appears in everything that happens quickly. A major league pitch covers 18.3 meters in about 400 milliseconds. Suppose the brain takes 150 milliseconds to see the ball and figure out where it is going. By then, the ball has already traveled more than 6.7 meters.
A batter swinging at the ball where it is would miss every time. Waves traveling ahead of the incoming signal offer one possible way to close that gap. Scientists have described these waves for decades. Yet their function remained debated for most of that time. This review provides a detailed account of what they compute and how the underlying circuit does it.
This gives model developers something concrete to test. In vision models, brain cells have long been treated as fixed detectors, each waiting for its own specific feature to appear. In wave dynamics, lateral connections take center stage. Moreover, such models contain testable predictions about wave speed and spatial spread.
The authors focused on vision because it is the most studied sense. Similar neural connections run across the rest of the brain’s outer shell.
“One of the major takeaways from the review is that traveling wave dynamics are not restricted to any single brain region or function,” Reynolds explained.
Similar activity appears in areas responsible for movement, planning, and memory. The next test will be direct. Researchers can now try to disrupt specific wave patterns in awake animals and humans and observe what happens to perception.
About the Author
