Every difficult decision can feel like an internal argument.
Take the job with the bigger salary but move away from everyone you know. Invest in the opportunity that could pay off or keep the money somewhere safe. Walk into the unfamiliar situation or avoid it.
New research suggests that this familiar feeling of conflict may have a remarkably literal counterpart inside the brain.
Scientists recording electrical activity directly from the human orbitofrontal cortex found two neighbouring brain regions behaving almost like opposing teams while people weighed reward against risk.
One region became more active before participants chose to move towards a potential reward. The other became relatively more active when the brain leaned towards avoiding the threat.
When the choice became difficult, activity rapidly alternated between these competing states before settling on a decision.
The study, published in Nature Neuroscience on 15 September 2026, provides an unusually direct view of what happens in the human brain during approach–avoidance conflict.
Rather than inferring decision-making from brain scans taken over seconds, the researchers were able to observe electrical activity changing on the scale of milliseconds.
The result suggests that some risky decisions may literally involve a neural tug-of-war before one option wins.
The researchers had a rare opportunity to record directly from the human brain
Studying the orbitofrontal cortex, or OFC, is difficult.
The region sits low in the frontal lobe, directly above the eye sockets and close to the sinuses. Its location makes it harder to measure accurately using many standard brain-imaging techniques.
Functional MRI can show changes in blood flow associated with neural activity, but those changes are relatively slow compared with the electrical signals neurons themselves produce.
The new study was possible because the researchers worked with six patients who already had depth electrodes implanted for clinical reasons.
Five participants were treated at the University of California, San Francisco, while one was treated at Washington University in St. Louis.
The electrodes had been placed to assist with the treatment or investigation of neurological and psychiatric conditions, not specifically for the research.
Because some of those electrodes passed through the orbitofrontal cortex, the scientists could record brain activity directly while the participants performed a specially designed decision-making task.
The task involved treasure chests and bombs
The experiment was built around a simple video-game-style environment.
Participants repeatedly viewed corridors containing different numbers of treasure chests and bombs.
Treasure represented a potential reward.
Bombs represented the possibility of losing points.
On each trial, participants had up to six seconds to decide whether to enter the corridor or avoid it.
Some decisions were easy.
A corridor containing many treasure chests and no bombs provided little reason to hesitate.
Other trials created a much stronger conflict.
A corridor might offer several possible rewards while also containing enough bombs to make the decision genuinely uncomfortable.
The researchers could therefore compare brain activity during relatively obvious decisions with activity during moments when reward and danger were more closely balanced.
Two parts of the orbitofrontal cortex behaved in opposite directions
The most striking result came from two anatomically distinct regions within the OFC.
Before participants chose to approach the corridor, activity increased in a medial portion of the orbitofrontal cortex.
At roughly the same time, activity in a more lateral portion decreased.
The pattern reversed when the system moved towards avoidance.
These signals were strongly anti-correlated.
When one side of the decision system became more active, the other tended to become less active.
This opposition appeared to represent more than a simple response to the amount of reward or danger shown on the screen.
Instead, the activity tracked the evolving choice itself.
Hard decisions produced rapid switching between states
The most revealing moments occurred when participants faced an ambiguous choice.
During these trials, the neural state did not simply move smoothly from uncertainty towards a final answer.
It flickered.
The researchers observed rapid alternation between a pro-approach state and a pro-avoidance state during the decision period.
One interpretation is that the brain was repeatedly evaluating the competing value of the reward and the potential punishment before commitment.
The choice only became behaviour once one state ultimately dominated strongly enough.
This provides a possible neural explanation for the subjective experience of hesitation.
A difficult choice may feel unstable because the decision system itself is temporarily unstable.
The brain may not calculate risk with one single value signal
Traditional economic models often describe choice as though the brain calculates one combined value for each option and then selects whichever comes out highest.
Neuroscience has supported versions of this idea by showing that the orbitofrontal cortex represents information related to reward and economic value.
The new findings suggest that approach–avoidance decisions may involve a more distributed competition.
Rather than one signal simply representing whether an option is good or bad, separate populations within the OFC may support opposing behavioural tendencies.
One pushes towards engagement with the opportunity.
The other supports withdrawal.
The final decision may emerge through competition between them.
This could help explain why uncertainty slows us down
Most people take longer to decide when the positive and negative consequences of an option are closely matched.
The behavioural data in the study reflected this familiar pattern.
Participants hesitated more when the corridor contained a combination of reward and danger that made the correct choice less obvious.
The neural recordings provide a possible mechanism for that delay.
If competing decision states continue alternating, the system may take longer to stabilise.
An easy choice can therefore be made rapidly because one neural tendency dominates almost immediately.
A difficult choice remains unresolved because neither side initially wins.
This kind of conflict extends far beyond video games
Approach–avoidance conflict is one of the most common structures in real-world decision-making.
The opportunity and the threat often exist in the same choice.
A promising career move may require leaving a familiar city.
A financial investment may offer greater returns while creating a risk of loss.
A difficult conversation may improve a relationship while also carrying the possibility of conflict.
An individual with social anxiety may want connection while simultaneously wanting to avoid the discomfort of social interaction.
The treasure-and-bomb game is obviously much simpler than these real decisions.
But it allows researchers to isolate the same underlying problem: should the person approach something potentially rewarding when doing so also creates risk?
The findings may eventually matter for anxiety and addiction
The researchers and independent experts have pointed to several clinical conditions in which approach–avoidance decisions can become distorted.
Anxiety disorders can involve excessive avoidance even when the real danger is relatively small.
Addiction can involve unusually strong approach behaviour towards a reward despite substantial negative consequences.
Gambling disorder presents another obvious example, where the prospect of reward can outweigh increasingly serious financial risks.
Obsessive–compulsive disorder can also involve maladaptive decisions around perceived threat and avoidance.
If medial and lateral OFC circuits contribute differently to these tendencies, future research may eventually help explain why certain disorders push decision-making consistently towards one side.
That does not mean the new study has identified a treatment.
It has identified a candidate neural architecture that researchers can investigate further.
One of the strengths is the speed of the recordings
Direct intracranial recordings provide temporal precision that most non-invasive methods cannot match.
Neural activity can change within milliseconds.
A brain scanner that effectively averages activity over much longer intervals may show that the orbitofrontal cortex was involved in a decision without revealing the rapid switching taking place inside it.
By measuring electrical activity directly, the researchers could observe the competing signals unfolding during the decision itself.
This is why a study involving only six participants can still provide valuable mechanistic information.
The number of people is small, but the recordings from each participant are unusually detailed.
The sample size is also the study’s biggest limitation
Six people cannot represent the full diversity of human decision-making.
The participants were not healthy volunteers randomly selected from the population.
They had electrodes implanted because of clinical needs, including epilepsy or psychiatric treatment.
Electrode placement was also determined entirely by those clinical needs.
The researchers could only study the parts of the orbitofrontal cortex that happened to be sampled in each person.
This makes replication essential.
The same pattern will need to be observed in additional patients, across different tasks and ideally with complementary non-invasive methods before researchers can confidently describe it as a general architecture of human decision-making.
Real decisions are also more complicated than bombs and treasure
The game deliberately reduced risk and reward to clear numerical possibilities.
Real-world choices include many additional factors.
People consider social consequences, memories, emotions, uncertainty, moral values, future plans and what other people may think.
Risk is often impossible to quantify accurately.
The consequences may also unfold over months or years rather than seconds.
The experiment therefore captures one fundamental component of decision-making rather than reproducing the entire process.
Other regions of the brain almost certainly contribute to complex real-life choices.
A decision can change before we are consciously aware of it
The study also raises a broader psychological question.
People often describe making decisions as a deliberate sequence: consider the options, weigh the advantages and disadvantages, then choose.
Neural recordings reveal a far more dynamic process underneath that experience.
The brain is continuously shifting between competing representations before behaviour becomes visible.
By the time a person experiences the final decision as settled, substantial competition may already have occurred.
This does not mean conscious reasoning is an illusion.
It means conscious deliberation is built on neural systems that are changing from moment to moment.
The result fits with a broader move towards studying decisions in real time
Decision neuroscience has historically relied heavily on averaging.
Researchers might compare brain activity across all risky choices with activity across all safe choices.
That approach reveals which regions participate in the task.
But averaging can hide the internal dynamics of individual decisions.
The new work focuses instead on transitions between states.
This is increasingly important because the brain may not remain in one stable computational mode while a person thinks.
It may rapidly move between possible interpretations and behavioural plans until the system reaches a point where one action becomes dominant.
Why this matters for economics too
Risk-taking sits at the centre of behavioural economics.
Traditional economic theory assumes that people evaluate possible outcomes according to preferences and probabilities.
Behavioural economics has repeatedly shown that those preferences are not perfectly stable.
Risk tolerance can change depending on framing, recent experiences, emotional state and context.
The new neural findings suggest one possible biological reason that preferences can feel unstable.
If competing approach and avoidance systems are continuously interacting, small changes in context could shift the balance between them.
A decision that feels worthwhile today may feel too dangerous tomorrow even when the objective probabilities have barely changed.
There is no single ‘risk-taking centre’ in the brain
It is tempting to turn neuroscience findings into simple anatomical stories.
One region becomes the fear centre.
Another becomes the reward centre.
The reality is more complicated.
The OFC is interconnected with many other systems involved in emotion, memory, motivation and action.
The medial and lateral signals observed in this study should therefore not be interpreted as two isolated switches controlling all risky behaviour.
They are part of a broader network.
The study’s contribution is to show that even within one important decision-making region, neighbouring areas can carry opposing signals at the same moment.
The internal argument may be biologically useful
Hesitation is often treated as a weakness.
Quick decisions are associated with confidence, while indecision can feel inefficient.
But a system that immediately commits to every attractive reward would be dangerous.
A system that automatically avoids every threat would be equally limiting.
Competing neural signals may provide a useful safeguard.
The brain can represent the attraction of the reward and the danger of the punishment simultaneously rather than forcing one to disappear too early.
The uncomfortable feeling of being torn between two options may therefore reflect a decision system doing exactly what it was designed to do.
The next step is to understand when the balance goes wrong
The study gives researchers a clearer map of what normal approach–avoidance conflict may look like inside the human orbitofrontal cortex.
The next question is whether the same neural competition behaves differently in people who consistently take too many risks or avoid too many opportunities.
Researchers could also investigate how sleep, stress, medication, addiction or psychiatric illness affect the balance between the two states.
Future work may reveal why one side becomes dominant in some individuals and whether that balance can be modified.
For now, the study offers something rarer: a direct recording of the human brain while a difficult choice is still undecided.
The familiar feeling of weighing an opportunity against its danger may not just be a metaphorical struggle.
For a few milliseconds at a time, the brain really does appear to move back and forth between go and don’t go before one side finally wins.
Source Information
Study Title: 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 et al.
Journal: Nature Neuroscience
Published: 15 September 2026
Participants: Six patients with clinically implanted intracranial depth electrodes
Method: Intracranial stereotactic electroencephalography during a gamified approach–avoidance task
Main finding: Medial and lateral orbitofrontal regions showed opposing activity patterns that rapidly alternated before difficult approach–avoidance decisions
DOI: 10.1038/s41593-026-02444-4
