HOUSTON – (Sept. 18, 2025) – The brain is famously plastic: Neurons' ability to change their behavior in response to new stimuli is what makes learning possible. And even neurons' response to the same stimuli changes over time ⎯ a phenomenon known as representational drift. Yet our day-to-day perception of the world is relatively stable. How so?
Resolving such puzzles matters for future brain-computer interfaces, sensory prostheses and therapies for neurological disease. On a quest for an answer, Rice University scientists have built ultraflexible probes thousands of times thinner than a human hair and used them to track neurons in the visual cortex of mice for 15 consecutive days as the animals viewed thousands of images ⎯ from line patterns to pictures of the natural world. The devices, called nanoelectronic threads, or NETs, embed seamlessly with brain tissue, allowing for high-fidelity chronic recordings of brain activity.
According to a study in Nature Communications, this custom, large-scale neurorecording array revealed that the stability of visual representations is better captured by neurons' millisecond rhythms ⎯ the temporal code ⎯ than by counting how many times a neuron fires over secondslong intervals ⎯ a measure known as firing rate code or "volume."
"Prior studies mostly measured volume because their probes read brain activity too slowly to catch fine timing," said Hanlin Zhu, a Rice postdoctoral associate who is the first author on the paper. "This can make the brain's message look more 'drifty' than it really is. Our fast electrical recordings let us read the rhythm directly, and we found that rhythm beats volume when it comes to explaining how the brain maintains a stable picture of the world from day to day."
Years in the making, NETs not only made this experiment possible but are also being applied to other frontiers ⎯ from mapping spinal cord circuits to developing more precise brain stimulation therapies. This technology illustrates why sustained investment in neuroscience matters. In Texas, voters this fall will consider Proposition 14 , a $3 billion measure to support brain research. If approved, it would create the Dementia Prevention and Research Institute of Texas, positioning the state as a national leader in advancing research in dementia and related disorders.
In the experiment, mice implanted with NETs viewed nearly 12,000 images a day ⎯ moving stripes, still stripes, tiny dotlike patches used to map where each cell "looks" and natural scenes. The researchers tracked neuronal tuning across these stimuli and evaluated performance at the level of single cells and whole populations.
When judged by firing rate, many neurons looked unreliable.
"But looking at temporal code showed that each cell's preferences ⎯ which picture it 'likes' ⎯ is actually stable across days, especially for cells that looked unreliable if you judged them by volume alone," Zhu said.
NETs made it possible to not only track hundreds of individual neurons but also to look at what Zhu called "their friend network" ⎯ the cells that tend to fire at nearly the same time. This revealed that the stability of visual representations is a group effort ⎯ a population-level effect resilient to individual neurons' instability.
"We tracked the same between-neuron relationships day after day at millisecond precision and across all four stimulus sets ⎯ who tends to talk to whom and with what lag ⎯ something that's been very hard to do at scale," Zhu said. "To our knowledge, this is the first day-to-day tracking of the same interneuron functional connectivity in mouse visual cortex at this temporal resolution and scale. That network view neatly explains why timing ⎯ not loudness ⎯ anchors our sense of the familiar."
The temporal code-level data recorded by the NETs enabled computer models to analyze it to identify which stimulus the mouse was seeing, even days after the model was initially trained, without needing to be readjusted. It also helped reduce the "drift," or fading accuracy, of these predictions over time.
Chong Xie , professor of electrical and computer engineering, and Lan Luan , associate professor of electrical and computer engineering, led the study. Both are members of the Rice Neuroengineering Initiative and have spent years refining the NETs.
"This work shows how advanced recording tools can reveal organizing principles of the brain that were not visible before," Xie and Luan said in a joint statement. "These insights are the foundation for building practical technologies, from brain-computer interfaces to new therapies. They also underscore why initiatives like the Dementia Prevention and Research Institute of Texas are so important. Long-term investment gives researchers the ability to push the limits of technology and discovery in ways that can ultimately change lives."
The research was supported by the National Institutes of Health (R01NS102917, U01NS115588, R01NS109361, UF1NS126566, R01EY036094), Rice and the Intelligence Advanced Research Projects Activity via Department of Interior/Interior Business Center (D16PC00003). The content in this press release is solely the responsibility of the authors and does not necessarily represent the official views of funding organizations and institutions.
