The brain gathered evidence only when a marmoset was looking at its partner.
In a Yale laboratory, two small primates named Kanga and Dodson have offered neuroscience a quiet revelation: the brain mechanism long thought to govern how we perceive the physical world may be the very same one we use to read and respond to each other. Researchers studying marmosets found that neurons in the prefrontal cortex accumulate evidence about a partner's behavior before triggering a coordinated action — suggesting that social decision-making is not a separate, higher faculty, but perhaps a variation on something ancient and shared. The finding invites us to reconsider the boundaries we have drawn between perceiving the world and understanding one another.
- The central tension is deceptively simple: can the same neural computation that tells you a light is bright enough also tell you when a friend is about to move?
- Classic cooperation tasks kept exhausting the marmosets too quickly, threatening to leave researchers with data too thin to trust — so the team redesigned the experiment to feel less like a test and more like play.
- Lightweight neural implants allowed the animals to move freely while their prefrontal cortex was recorded in real time, capturing the telltale ramp of evidence accumulation right up to the moment of synchronized action.
- A striking detail disrupted easy conclusions: the neural buildup halted whenever a marmoset looked away from its partner, tying the entire computation to active social gaze.
- The findings land as a strong correlation, not yet a cause — future experiments must stimulate or silence these neurons mid-task to determine whether they are driving decisions or merely echoing them.
Kanga watches Dodson from across the enclosure. In her dorsomedial prefrontal cortex, neurons begin firing with mounting intensity — not randomly, but in a deliberate climb toward a threshold. When the signal peaks, she pulls the lever. Dodson pulls at the same instant. Both marmosets receive their reward.
This small synchronized moment sits at the center of a larger scientific question. For decades, neuroscientists have mapped a process called evidence accumulation — the way brains gather information, wait for it to cross a threshold, and then act — in animals making choices about lights, sounds, and other features of the physical world. But those studies involved solitary animals and static stimuli. What happens when the decision depends on another living creature, one whose behavior you influence and who influences yours in return?
A Yale team led by Monika Jadi and Steve Chang designed a lever-pulling task flexible enough to let marmosets cooperate at their own pace, freely and repeatedly, while lightweight neural probes recorded activity in their prefrontal cortex. The results mapped almost perfectly onto the evidence accumulation model: weak information about a partner produced a slow neural ramp; clear signals produced a fast one. The activity peaked just before the pull.
One detail proved especially telling. The accumulation only occurred while the marmoset was actively watching its partner. When gaze shifted away, the buildup stopped — suggesting the brain manages social complexity by anchoring itself to what it can directly observe in the moment.
Outside researchers noted the significance of finding this pattern outside a controlled, single-animal setting. If the same computation governs perceptual and social decisions alike, the categories neuroscience has long kept separate may be less distinct than assumed. Still, correlation is not causation. The team's next steps include stimulating these neurons during the task to test whether they drive decisions or merely reflect them, recording across multiple brain regions simultaneously, and exploring how trust and familiarity reshape the mechanism. Jadi calls the current work an early exploration — one that has opened more questions than it has closed.
Kanga reaches for the lever. Her eyes find Dodson across the space between them, watching for the moment when he'll make his move. In her brain, in a region called the dorsomedial prefrontal cortex, neurons begin to fire with increasing intensity. They're not firing randomly—they're building toward something, accumulating information about what Dodson is about to do. When the signal reaches its peak, Kanga pulls. Dodson pulls at the same instant. Both marmosets get their reward: a sip of marshmallow fluff.
This small synchronized moment reveals something large about how brains work. For decades, neuroscientists have understood that when we make a decision about what we perceive—whether a light is bright enough, whether a sound is loud—our brains follow a predictable pattern. They gather evidence. They wait. When the evidence crosses a threshold, they act. This is called the evidence accumulation model, and it has held up across hundreds of studies in controlled labs. But those studies mostly involved single animals making choices about the physical world around them. What happens when the decision depends on another living creature? When your choice influences what they do, and their behavior shapes what you decide? That's messier. That's social.
A team at Yale University, led by Monika Jadi and Steve Chang, decided to find out whether the same brain mechanism that handles simple perceptual decisions could handle social ones. They couldn't use the classic test of cooperation—two animals pulling ropes to reach food—because that task exhausted the marmosets after just a few repetitions, leaving researchers with too little data. Instead, they designed a lever-pulling game that the marmosets could play repeatedly, whenever they wanted, at their own pace. It was controlled enough for science but loose enough to feel like play. The marmosets could see each other. They could choose when to start and stop. They could fail or succeed based on their own timing and attention.
To watch what was happening inside their brains, the researchers implanted lightweight neural probes in the marmosets' prefrontal cortex—the region involved in planning and decision-making. The probes were small enough and light enough that the animals could move freely, pulling levers without constraint. What the recordings showed matched the evidence accumulation model almost perfectly. When a marmoset had weak information about what its partner was doing, the neurons ramped up slowly. When the evidence was strong—when the partner was clearly about to act—the neurons ramped up faster. The activity peaked right before the decision to pull.
But there was a crucial detail: the brain only gathered this evidence when the marmoset was actually looking at its partner. When the marmoset looked away, the accumulation stopped. This finding, Jadi explains, suggests that social gaze simplifies the problem. Instead of trying to track everything the other animal might do, the brain focuses on what it can see in real time. It's a way of reducing complexity.
The implications ripple outward. Timothy Hanks, a neuroscientist at UC Davis who was not involved in the work, notes that evidence accumulation has been demonstrated in highly controlled settings. The fact that the same pattern emerges in a social task—messier, less constrained, dependent on another animal's behavior—suggests this is not a special-case computation. It's something more general. Cory Miller, a psychologist at UC San Diego, puts it differently: the categories we've created—social decisions, perceptual decisions, foraging decisions—may not be as different as we think. "There may not be anything acutely different about them," he says.
Still, much remains unknown. The current findings show correlation: brain activity correlates with decision-making. But correlation is not causation. Future work will need to stimulate or inhibit these neurons during the task to see whether they actually drive the decision or simply reflect it after the fact. The researchers are also designing new implants that can record from multiple brain regions simultaneously, to understand how different parts of the brain communicate during cooperation. And they want to test whether the same mechanism holds for decisions involving trust, or for marmosets working with strangers versus familiar partners. Jadi describes the current work as "an early exploration." With this paradigm, she says, there are far deeper questions still to ask.
Notable Quotes
It's a very recurrent system—making a social decision relies on the changing behavior of another animal, and the actions of the decider can influence what the other animal does.— Monika Jadi, Yale University
The social gaze is reducing the complexity of the problem.— Monika Jadi, Yale University
The Hearth Conversation Another angle on the story
Why does it matter that marmosets use the same decision-making process for social choices as they do for simple perceptual ones?
Because it suggests the brain isn't running completely different software for social versus non-social problems. If the same computational rule applies across contexts, it means we might be able to understand cooperation and coordination using the same tools we've already developed for studying perception.
But social decisions seem harder. You have to predict what another animal will do, and your own actions change what they do next.
Exactly. That's what makes this surprising. You'd expect the brain to need something more elaborate, more specialized. Instead, it seems to solve the problem by narrowing its focus—by only gathering evidence when it's actually looking at the other animal.
So the marmoset isn't trying to model the other marmoset's entire mind?
Not in this task. It's simpler than that. It's watching, gathering real-time information about what the partner is about to do, and accumulating evidence until it's confident enough to act in sync.
What happens next in this research?
They want to know if stimulating these neurons actually causes decisions, or if the activity is just a side effect. They also want to see if the same mechanism works for decisions involving trust or long-term relationships—things that require more complex social reasoning.
Could this eventually explain how humans cooperate?
That's the long-term hope. Marmosets are social animals like we are. If we can map the neural basis of their cooperation, it gives us a foundation for understanding our own. But we're still in the early stages.