Summary:
Using direct recordings from electrodes implanted in the human orbitofrontal cortex, researchers have identified two neighboring neural circuits that engage in a millisecond-by-millisecond “tug-of-war” between approach and avoidance. Distinct electrical patterns predict whether a person will take a risk or play it safe roughly half a second before the action, suggesting precise neural targets for treatments of obsessive-compulsive disorder (OCD), depression, addiction, and other conditions involving maladaptive approach–avoidance behavior.
Key Facts:
- Direct millisecond readout: Intracranial electrodes in patients undergoing epilepsy evaluation allowed real-time monitoring of orbitofrontal cortex activity, overcoming the temporal and spatial limits of fMRI for this region.
- Opposing circuitry within millimeters: Activity in a medial patch (medial orbital sulcus) rises when participants choose to approach risk, while a lateral patch about two centimeters away activates when they decide to avoid threat. The two regions alternate rapidly, behaving like a switch rather than a slowly turning dial.
- Clinical and therapeutic potential: The neural signature anticipates choices about 500 milliseconds before movement, offering an objective biomarker to guide targeted neuromodulation—potentially enabling millimeter-level stimulation strategies for disorders marked by excessive avoidance or compulsive approach.
Source: UCSF
Every day the human brain runs countless split-second assessments that weigh reward against risk: should you sprint through a yellow light to avoid being late, or slow down to reduce danger? Should you place another bet or walk away? An interdisciplinary team at UC San Francisco and UC Berkeley has now localized where those high-stakes choices are resolved in real time.
Published in Nature Neuroscience, the study shows that two adjacent patches of the orbitofrontal cortex engage in rapid, alternating activity that predicts a participant’s choice about half a second before action. These findings provide new insight into the neural computations of approach–avoidance decision-making and clarify how neighboring subregions within the orbitofrontal cortex exert opposite influences on behavior.
“We’ve suspected this area plays a central role in weighing reward against risk, but until now we couldn’t observe those computations as they unfolded in the living human brain,” said Edward Chang, MD, chair of Neurological Surgery at UCSF and co-senior author of the study.
Moving Beyond fMRI with Immersive Gaming
Most human decision-making studies have relied on functional MRI, but the orbitofrontal cortex is notoriously difficult to image accurately due to magnetic field distortion from air-filled sinuses. Static tasks in the scanner can also fail to engage participants, reducing the ecological validity of results.
To capture genuine motivation and realistic risk assessment, lead author Clara Starkweather, MD, PhD, designed an immersive video game. Players navigate winding corridors lined with explosives to reach glowing treasure chests. Each trial presents a single choice: run through a bomb-strewn hallway to collect rubies or avoid the corridor to preserve the current score. The task generates meaningful, variable approach–avoidance decisions that mimic everyday trade-offs.
Six patients with temporary intracranial electrodes implanted for epilepsy evaluation volunteered to play the game while researchers recorded neural signals at millisecond resolution. This intracranial approach provided a direct window into the orbitofrontal cortex’s rapid dynamics during naturalistic decision-making.
A Biological Tug-of-War
The direct recordings revealed a clear anatomical and functional separation within the orbitofrontal cortex:
- Medial orbital sulcus: Activity in this medial region increases when participants decide to approach a reward despite potential danger.
- Lateral orbitofrontal patch: Activity in this lateral area rises when participants choose to retreat or avoid risk.
These two clusters alternate in precise inverse patterns: when one surges, the other wanes. Rather than a smooth accumulation of evidence, the neural activity resembled a rapidly flipping switch. On clear-cut decisions, the “go” or “stop” signal dominated almost immediately; on ambiguous trials, the two patches oscillated back and forth until one prevailed.
Computational models developed by the team described this dynamic as alternating discrete pro-approach and pro-avoidant states, rather than continuous integration. Crucially, monitoring which circuit prevailed allowed accurate prediction of a participant’s choice about 500 milliseconds before they acted.
Personalizing Psychiatric Neuromodulation
These findings have immediate implications for neuropsychiatry. Disorders such as major depressive disorder, severe anxiety, and OCD often feature excessive avoidance, while addiction and pathological gambling reflect excessive approach. Current neuromodulation treatments—like deep brain stimulation (DBS) or noninvasive stimulation—have shown mixed results when the orbitofrontal cortex is targeted as a single area.
Differentiating a localized “stop” zone from a neighboring “go” zone could enable clinicians to steer stimulation with far greater precision, targeting the specific circuit that is overactive or underactive in a given patient. Objective neural signatures of approach–avoidance balance could complement clinical assessments and improve personalization of therapy.
“Psychiatry has relied heavily on self-report. By identifying measurable brain signals that reflect how someone weights risk and reward, we can develop interventions that directly address the dysfunctional circuit,” said Starkweather.
Editorial Notes:
- Edited by a Neuroscience News editor.
- Journal paper reviewed in full by staff.
- Additional context provided by the editorial team.
About this Neuroscience Research:
- Media Contact: Laura Kurtzman
- Source: UCSF
- Image Credit: Image generated for Neuroscience News
- Original Research (Open Access): Nature Neuroscience (September 15, 2026). “Intracranial recordings in humans reveal differential contributions of medial and lateral orbitofrontal cortex to approach–avoidance decision-making.” Authors: Clara Kwon Starkweather, Ethan H. Willbrand, Kristin Sellers, Patrick W. Hullett, Andrew D. Krystal, A. Moses Lee, Kevin S. Weiner, Jon T. Willie, Peter Brunner, Ming Hsu, Edward F. Chang & Robert T. Knight.
- DOI: 10.1038/s41593-026-02444-4
Abstract
Intracranial recordings in humans reveal differential contributions of medial and lateral orbitofrontal cortex to approach–avoidance decision-making
Approach–avoidance describes the conflict that arises when pursuing a reward carries potential punishment. The orbitofrontal cortex (OFC) has been proposed to arbitrate these decisions, but the real-time computations within human OFC have been largely unknown. We recorded intracranial activity from the OFC of human participants implanted with stereotactic electroencephalography electrodes while they performed a gamified approach–avoidance task.
We identified two anatomically distinct signals prior to approach: an increase in activity in a medial compartment and a concurrent decrease in a lateral compartment. Before decisions were made, these signals alternated rapidly between discrete pro-approach and pro-avoidant states, revealing a new functional architecture within human OFC during decision-making.