// Contents
“I’m so surprised the voices have gone for now. My mind is really quiet.” A person with schizophrenia said this after a series of sessions with no injections and no implants. A headset on their head sent ultrasound through the skull to a spot about the size of a pea, roughly two and a half inches deep.
The Focused Ultrasound Foundation shared those words on October 30, 2026. The foundation is funding a new trial at the University of California, San Francisco (UCSF). These are studies in people, but tiny ones: the first had three patients, and the new one will have up to twelve. In the first study the voices faded only after real sessions, even though a sham with the same energy but no focus fooled all three patients. The catch: the voices came back. Within a week, the quiet was gone.
Voices that pills don’t reach
Schizophrenia affects about one person in 300, or some 27 million people, according to the World Health Organization. More than 70% of them hear voices. Medication helps many, but in about 40% the voices persist despite antipsychotics.
The symptom is hard to picture from the outside. The trial’s second participant put it this way: “Hours can pass, and I feel like I lose track of time when I hear the voices and they seem to originate from outside and insert thoughts in my head.”
Karuna Subramaniam, the UCSF psychiatry professor who leads the study, isn’t targeting the whole brain, just one hub: the caudate. It sits deep beneath the cortex and is part of the striatum, a cluster of structures that runs on dopamine. This is exactly where antipsychotics act, damping dopamine’s effect.
In people who hear voices, the striatum is wired too tightly to the auditory cortex, the part of the brain that makes sense of sound. The researchers suspect this makes the brain misread the return address: your own inner speech arrives in the auditory cortex tagged as “outside.” It’s as if a thought called from an unknown number.
Waves that meet at a point
The striatum is hard to reach. Deep brain stimulation electrodes can get there, but that takes surgery, with its risks and lasting side effects. Magnetic stimulation, already used for depression, skips the surgery but works on the surface of the cortex and can’t reach the deep nuclei.
Ultrasound can. The headset, built by Attune Neurosciences, a California company, carries 128 emitters, 64 on each side of the head. Each sends out a weak wave, and a computer times them so the waves add up at a single point. Through skin and bone the energy stays spread thin. Only at the focus does it gather. The frequency is 500 kHz, twenty-five times higher than the highest pitch a human ear can catch.
At high power, the same physics is already in clinics. In 2016 the US Food and Drug Administration approved Exablate Neuro, which uses focused ultrasound to burn a tiny spot deep in the brain and calm tremor. The UCSF headset runs at far lower power and destroys no tissue: its weak pulses only change how the neurons at the focus behave, and only for a while. By the researchers’ calculations, tissue at the target warmed by no more than 0.6 °C (about 1 °F).
How a session worked
Device: ATTN201, frequency 500 kHz. Pulses of 3 milliseconds, twenty times a second, in cycles of 40 seconds on and 20 seconds off. Thirty minutes per hemisphere, one hour in total. The target was a sphere with a 5 mm radius: the head of the caudate, about 65 mm deep, or the nucleus accumbens, about 70 mm deep. The team planned the spot on each patient’s MRI scan. In the sham session the same device delivered the same energy without a focus, so the wave spread through the head.
Three people and a sham session
Subramaniam’s team posted the first study as a preprint in January 2026, and the journal Brain Stimulation later published it. It enrolled three people with schizophrenia whose voices medication didn’t stop. Each came in for three sessions, a week apart: ultrasound to the caudate, ultrasound to a neighboring hub of the striatum called the nucleus accumbens, and a sham session, in random order.
The sham was clever. Its wave carried the same energy but no focus, so at any point in the brain it stayed tens to hundreds of times weaker than at the focus of a real session. Asked afterward, all three said the sessions had felt alike.
The measure was simple. Every day for a week before each session and a week after, patients rated on a 0-to-100 scale how often they heard voices and how much the voices bothered them. After a real session, both scores fell by more than half in all three.
By the end of the week, everything was back to baseline.
In MRI scans of two patients (the third couldn’t be scheduled), the extra coupling between the striatum and the auditory cortex weakened after ultrasound. That is the very link that lets a thought “call from an unknown number,” and it loosened, at least for a while.
The side effects: faint clicks the patients could hear, and a tight feeling in the scalp where the headset pressed. A treatment for voices that makes a little noise of its own drew almost no complaints.
New York aimed somewhere else
NYU Langone Health in New York ran a similar experiment. Psychiatrists Wei Qi and Donald Goff enrolled ten people with schizophrenia or schizoaffective disorder (psychosis combined with mood episodes) whose symptoms persisted on medication. They chose a different target: the internal globus pallidus, another deep hub next to the striatum. Each session lasted 10 minutes. Everyone got one real and one sham session, and neither the patients nor the clinicians rating symptoms knew which was which.
The results, published on October 20, 2026, in npj Acoustics, look more modest. Here the main yardstick was delusions, not voices: several times over the week, patients answered questions about their delusional thoughts. After a real session the score fell by an average of 1.2 points; after a sham it rose by 0.4. The shift is small, but real and sham sessions pushed in opposite directions, and the gap held through the week of follow-up. The voices, however, eased no more after a real session than after the sham. Fatigue, headache and brief flare-ups of symptoms showed up equally often after both kinds of session.
In NYU Langone’s press release, Goff chose his words carefully: further work is needed to establish whether targeted ultrasound “could one day supplement or even replace standard medicines for schizophrenia.” Even “supplement” would mean a lot to people whose medication only partly works.
Five sessions instead of one
The new UCSF trial tests what the first study couldn’t show: whether more sessions make the quiet last. Each participant (there will be up to twelve) gets five daily sessions aimed at the caudate and five sham sessions, with a break between the two series. The team will track the voices for eight weeks, according to the trial registry. MRI will show whether the link between the striatum and the auditory cortex weakens, and for how long. The foundation doesn’t say how long the quiet has lasted for the new participants; eight weeks of follow-up will answer that. The quotes it has published so far are first impressions, not results. Every participant goes through both series, and the foundation doesn’t say which one came before these words.
Still, those impressions capture where things stand today. For the second participant, the one who used to lose hours to the voices, the voices grew softer after the sessions, moved “more like in the distant background,” and seemed to be fading. “I feel it’s very peaceful,” the participant said. But one of those voices still got a word in: “I’m still here.”
