Meet Neuropixels: how probes that track hundreds of brain cells at once are revolutionizing neuroscience

In a bright surgical theatre in central London, neurosurgeon Ahmed Toma makes an incision on the side of a woman’s head. Then, he bores a small hole into her skull, revealing a glimpse of the pink, pulsating brain beneath.

Even the unconscious brain can learn — and predict what you’ll say next

Toma is preparing to implant a shunt to drain excess cerebrospinal fluid — a routine treatment for a condition called normal pressure hydrocephalus, in which fluid builds up inside the skull. His patient at the National Hospital for Neurology and Neurosurgery (NHNN) has also agreed to take part in an experiment: one that involves briefly placing a centimetre-long probe, thinner than a human hair, into her brain. Called Neuropixels, this probe is designed to record the electrical activity of hundreds of neurons at once.

As Toma and a team of researchers carefully insert the probe into the patient’s brain, Mihaly Kollo, a research scientist at the Francis Crick Institute in London, stands nearby, monitoring a multicoloured array of outputs from the probe’s recording sites in real time. After controlling for electrical noise from the operating table and other surgical equipment, Kollo plays an audio sequence of repeating terms — ‘blind spot’, ‘parking spot, ‘piggy bank’ and so on — to examine how the brain represents the transitions between stimuli, and how they are influenced by frequency and familiarity. As the words ring through the theatre, some trigger brain perturbations that sweep across the waveforms representing the probe’s output — a clear sign that it is picking up activity.

Scientists have long conducted experiments during neurosurgery, which offers a rare opportunity to access a living (and often awake) human brain. But Neuropixels probes might provide researchers with the sharpest view of the organ yet.

Since their debut in 2017, Neuropixels have transformed how neuroscientists do electrophysiology experiments in animals, allowing them to record electrical activity from hundreds, sometimes even tens of thousands, of neurons at once. Over the past year, at the NHNN, which is part of the University College London Hospitals, Kollo and his colleagues have been collaborating with neurosurgeons such as Toma to take Neuropixels recordings in people.

Medicine Nobel awarded for brain ‘switch’ that controls neurons with light

There are other ways of surveying single-cell activity in large populations of neurons, such as calcium imaging, which tracks the movement of calcium ions that occurs each time a neuron fires. But these techniques can be too slow to capture the action potentials in certain neurons, or require genetically encoded sensors, limiting their use in some organisms and precluding it in humans.

For people, researchers have relied mostly on devices that sit on the surface of the brain and measure population-level activity, such as high-density electrocorticography. There have also long been tools that can make a relatively small number of recordings at the single-cell level, such as the Utah array — a tiny device studded with miniature needles that penetrate about a millimetre into the brain and that has 96 recording channels.

Pierre Bourdillon, a neurosurgeon at the Adolphe de Rothschild Foundation Hospital in Paris, likens the technological leap for human neuroscience enabled by Neuropixels to the difference between trying to decipher conversations from a single microphone at a busy intersection and listening in on those same discussions using individual microphones attached to the speakers’ collars.

“We’re just starting to get to the level of microphones on individual people,” he says. “For non-human animals, this revolution began about 10–15 years ago. Now we’re just starting to see the first teams using this kind of technology in humans.”

Recording en masse

A short walk from the NHNN, at the hospital-turned-research institute known as the Cruciform Building at University College London, neuroscientist Matteo Carandini is applying the most cutting-edge version of Neuropixels technology in organisms with much smaller brains.

His team has been using Neuropixels to record tens of thousands of neurons in mouse brains, sometimes for months at a time. On a sunny Monday afternoon in September, he shows me the set-up for Neuropixels Opto¹. This version of the technology, which he has helped to develop, combines electrical recording with optogenetics, a method that allows researchers to switch neurons on and off with light.

Carandini was one of the first neuroscientists to use Neuropixels, a device that was designed for research in rodents. The idea began with Tim Harris, a biomedical engineer who arrived at the Howard Hughes Medical Institute’s Janelia Research Campus in Ashburn, Virginia, in 2008 with the task of identifying and developing tools that researchers needed. When he asked his colleagues: ‘What do you want to measure that you can’t?’, he discovered that they were missing probes for conducting long-term, large-scale electrophysiology experiments. In living animals, these involve directly measuring the electrical properties of neurons.

At the time, researchers mostly probed the brain with arrays of twisted four-wire electrodes, which could be combined to record from a few dozen channels at once, or silicon probes, which typically had around 32 recording sites. But they wanted a more comprehensive view of how the cells in the brain work together to perform the organ’s crucial functions — and for that, they would need to collect recordings from many more neurons.

After attempting a few designs on his own, Harris contacted the Interuniversity Microelectronics Centre (Imec), a research institute in Leuven, Belgium, that helped to pioneer some of the world’s most advanced microchips. To Harris’s surprise, Barun Dutta, chief scientist of Imec’s semiconductor technology and systems centre, agreed to take on the task.

How our brains decode speech: special neurons process certain sounds

Together, they developed Neuropixels 1.0. The silicon probe comprises 960 recording sites along the hair-thin shank. Each of the recording channels was small enough to capture the action potentials of individual neurons — and 384 of those could be read out in parallel. The shank was attached to a chip the size of a nano SIM card, which holds the electronics required to read, amplify and digitize signals. The outputs could be read with just two wires; this was a major advance at a time when most other devices required each recording channel to have its own wire.

The team reported² its design in 2017. In that article, Carandini and around three dozen neuroscientists across several laboratories who had tested the device showed that it was possible to insert two of these probes — each weighing about 250 milligrams — into a mouse, allowing them to record around 700 neurons across 5 brain regions at the same time.

Neuropixels 2.0, a smaller, lighter-weight, four-shank probe with 5,120 recording sites (1,536 of which could read out at once), arrived four years later³. (The 1.0 and 2.0 probes are available from Imec for about US$1,700 and $1,800 each, respectively.)

Inside the mind of an animal

Neuropixels probes have transformed the way neuroscience can be done, says Carandini, who is using the probes to study how the brain processes and uses sensory signals. They enabled the tracking of individual action potentials from thousands of neurons at once and made it possible to record across multiple brain regions in the mouse brain, which is only about one centimetre long.

Using Neuropixels, researchers have found that signals associated with certain sensory, cognitive and motor functions are more widely distributed in the brain than previously thought⁴. These probes have also enabled scientists to study neural manifolds, the unique patterns of activity that form in populations of brain cells.

Several collaborative efforts known as brain observatories, in which multiple independent teams carry out the same experiments in a standardized manner, have collected vast amounts of Neuropixels data. In one 2025 study⁵, for example, a dozen labs in the United States and Europe, known as the International Brain Laboratory, came together to collect recordings from more than 620,000 neurons across 279 regions from 139 mice. The result was a brain-wide map of neural activity for both cognitive and motor functions, such as vision, movement and decision-making.

From rodents to humans

Around 1,700 labs around the world have used Neuropixels in more than two dozen species, including zebra finches, bearded dragons and octopuses. “I was completely taken aback at the variety of species that researchers used these for,” Harris says.

Scientists studying non-human primates were particularly keen: their existing tools either had too few recording channels or could access only the superficial layers of the brain. Harris, Dutta and their colleagues worked with the research community to create Neuropixels 1.0 NHP⁶, which featured a thicker shank to penetrate the tough membrane around the brain of non-human primates and, at 4 cm in length, could record from deeper structures. (These cost about $4,200 each, according to Dutta.)

Tirin Moore, a neuroscientist at Stanford University in California, who was involved in that effort, says that before Neuropixels, researchers would collect recordings from around two dozen neurons of non-human primates on a good day; now, they routinely record from at least 100. Without these tools, primate neuroscience “would still be in the dark ages”, he says. Moore and his colleagues plan to use the NHP probes to create a high-resolution neural activity map of the visual system of macaque monkeys as part of the Enigma Project, an initiative based at Stanford to record and decode the brain.