Credit: LightField Studios on Shutterstock
In A Nutshell
- Researchers used focused sound waves to temporarily quiet the amygdala in healthy adults, reaching a deep brain structure without surgery.
- When the amygdala was dampened, participants were slower to learn fear associations and quicker to stop responding to threat cues once danger was gone.
- A control experiment targeting a different brain region produced no such changes, confirming the effect was amygdala-specific.
- Findings raise the possibility of noninvasive approaches for treating PTSD and other anxiety disorders, though clinical testing is still needed.
Deep inside the human brain sits a small, almond-shaped structure called the amygdala. Neuroscientists have long suspected it drives one of the brain’s most stubborn patterns: threatening memories form quickly and can be slow to fade. For example, anyone living with PTSD all carry that burden. Researchers suspected the amygdala was the engine, but proving it directly in living, healthy humans without surgery had never been done.
A team at Radboud University in the Netherlands has now changed that. Using focused sound waves to temporarily quiet the amygdala during fear learning, they produced rare direct, cause-and-effect evidence in humans that this structure drives what they call an “emotional learning state,” one in which the brain learns threats rapidly and sheds them reluctantly. Published in Science Advances, their findings could reshape how clinicians approach trauma-related disorders.
How Scientists Reached the Amygdala Without Surgery
To reach the amygdala without cutting into the skull, the team used transcranial ultrasound stimulation, a technique that sends precisely focused, low-intensity sound waves through the skull to temporarily modulate activity in a targeted brain region. Each of 25 participants viewed images of snakes while occasionally receiving a mild electric shock. Some snake images were paired with a shock half the time, turning them into fear cues. Others were never paired with a shock. In a within-participant design, one set of snake cues was always paired with active ultrasound aimed at both sides of the amygdala, while a second set was always paired with a fake, sham version, letting researchers compare how the same person learned under both conditions.
Fear responses were tracked by measuring tiny changes in skin moisture, an involuntary reaction to perceived threat. By watching these signals trial by trial, researchers could observe fear learning as it happened.
Amygdala Stimulation Slowed Fear Learning and Accelerated Extinction
Under sham conditions, participants learned the threat association within a handful of trials. Skin responses spiked for dangerous snake images and stayed low for safe ones. When shocks stopped, those fear responses lingered before gradually subsiding. Under amygdala ultrasound, the picture shifted. Threat learning slowed significantly during the earliest trials, precisely the window when prior brain imaging shows the amygdala is most active. Once the shocks ended, participants who had received amygdala stimulation showed reduced fear responses several trials faster than those in the sham condition.
Computational models confirmed what the raw data suggested. Amygdala stimulation reduced how quickly participants updated their threat expectations during learning and increased how quickly they updated them once danger was gone. Crucially, overall learning ability was unaffected. Disrupting the amygdala did not make people worse learners in general. It specifically affected how the brain learned and updated threat information.
One finding added an unexpected wrinkle. After the task, participants who received amygdala stimulation overestimated how often threat cues had actually been paired with shocks, producing less accurate threat memories than those in the sham group. Normal amygdala function, it appears, helps calibrate the precision of fear memories, not just their strength.
Why the Hippocampus Test Mattered
To rule out the possibility that any brain stimulation would produce the same result, the researchers ran a second experiment with 25 new participants, this time aiming ultrasound at the hippocampus, a brain region involved in memory but not in this type of simple cue-based fear learning. Both groups experienced real ultrasound with identical physical sensations. If the results from the first experiment were simply a side effect of sound waves hitting the head, they should have shown up here too. They did not. Hippocampus stimulation produced no changes in how people learned or forgot fear, confirming the effect was specific to the amygdala.
What This Could Mean for Fear Memory and PTSD Treatment
PTSD, phobias, and other anxiety disorders are, at their core, disorders of threat memory. Treatments like exposure therapy work by building new safety memories to compete with old fear ones, but the original fear trace often survives and resurfaces under stress. This raises the possibility that targeting the amygdala during treatment could make those fear responses easier to extinguish from the start.
Researchers note that their study involved stimulating the amygdala during the initial learning of a threat, not after a memory had already formed. For clinical use, the logical next step is testing whether ultrasound can influence existing memories during the brief window when a recalled memory is temporarily open to change. Surgical deep brain stimulation of the amygdala has already shown early promise in severe, treatment-resistant PTSD, and ultrasound could offer a noninvasive way to explore similar approaches without the operating room.
What this work establishes is that the amygdala plays a central role in how the brain learns about danger, not just registering threats but actively shaping how fast fear takes hold and how long it persists. For the first time, scientists can reach that system from outside the skull and begin to shift the balance.
Disclaimer: This article is based on a single study involving healthy adult volunteers in a controlled laboratory setting. Sample sizes were small, and findings have not been tested in clinical populations or with real-world trauma. Nothing in this article should be taken as medical advice. Consult a qualified healthcare professional with any mental health concerns.
Paper Notes
Limitations
Results may not transfer to different ultrasound settings or procedures, as the direction and magnitude of stimulation effects likely depend on specific technical parameters and the brain’s state at the time of application. Because the study did not combine ultrasound with brain imaging during the task, there is no direct measure confirming that amygdala activity was changing as intended; conclusions rest on tight experimental controls rather than direct neural readouts. Stimulation was applied only during initial threat learning, not after memories had already formed, which limits immediate clinical application. Future studies should assess whether participants could distinguish active from sham stimulation on a trial-by-trial basis, and should incorporate in-skull exposure measurements to account for individual differences in skull shape and sound transmission.
Funding and Disclosures
Funding was provided by the European Research Council Consolidator grant DARE2APPROACH (ERC_CoG_772337 to K. Roelofs), the European Innovation Council Pathfinder Project CITRUS (grant 101071008; L. Verhagen, co-applicant), the ERC MediCoDe project (L. Verhagen, co-applicant), and the Dutch Research Council NWO VIDI Fellowship 18919 (to L. Verhagen). Co-author Lennart Verhagen is a board member of ITRUSST and the Brainbox Initiative, has received nonfinancial support from Image Guided Therapy SA, Sonic Concepts LLC, and Brainbox Ltd., and has received consulting fees from Nudge LLC. All other authors declare no competing interests. Participants provided written informed consent; the study was approved by ethics committee CMO2021-8238 and preregistered on the Open Science Framework.
Publication Details
Title: The human amygdala in threat learning and extinction | Authors: Sjoerd W. Meijer (corresponding author), Eleonora Carpino, Benjamin R. Kop, Jesse Lam, Lycia D. de Voogd, Karin Roelofs, and Lennart Verhagen. Roelofs and Verhagen contributed equally as senior authors. All authors are affiliated with the Donders Institute for Brain, Cognition and Behavior at Radboud University, Nijmegen, Netherlands, with additional affiliations at the University of Oxford, University of Plymouth, and Leiden University. | Journal: Science Advances, Volume 12, eaea8233 | Publication Date: March 25, 2026 | DOI: 10.1126/sciadv.aea8233 | License: Creative Commons Attribution License 4.0 (CC BY)







