How serotonin prepares us for what's next

If you’ve ever waited in line for ice cream, you’ve experienced a curious mix of hope, patience, and maybe a little frustration. Your brain knows the reward is coming, but it hasn’t arrived yet. That ability to anticipate, to look just a few seconds into the future and prepare for good or bad outcomes, is essential to how we learn and act.

Now, a team of neuroscientists from the University of Ottawa, Caltech, the Allen Institute for Neural Dynamics, and Johns Hopkins has put forward a bold new theory about how one of the brain’s most mysterious chemicals, serotonin, might help us do exactly that. Their research, published in Nature, proposes that serotonin neurons aren’t just reacting to rewards and punishments as they happen. Instead, they may be encoding a kind of “prospective value signal”, a running estimate of what the near future holds.

And that small shift in perspective could resolve a decades-long puzzle in neuroscience.

The serotonin puzzle

Serotonin is often nicknamed the “happiness chemical,” thanks to its role in mood and antidepressants. But in reality, it’s involved in much more: decision-making, learning, patience, even how we respond to both good and bad surprises.

The challenge is that serotonin neurons, especially those in a brain region called the dorsal raphe nucleus, behave in puzzling ways. In experiments with mice, these neurons fire when animals get rewards like drops of water, but also when they face punishments, like puffs of air. They seem to get especially excited by unexpected rewards, but not by unexpected punishments. Sometimes their firing ramps up slowly, sometimes it spikes, and sometimes it lingers at a steady plateau.

Over the years, scientists have floated competing theories: maybe serotonin encodes reward, or surprise, or salience (basically, “how important” a stimulus is), or uncertainty. Each idea explained part of the data, but none tied the whole picture together.

The authors of the work suggest we’ve been looking at serotonin the wrong way. Instead of treating it as a signal about the present, we should think of it as a forecast about the near future.

In the language of reinforcement learning (a field of AI that has long inspired neuroscience), this is called a value signal: the brain’s estimate of how much reward or punishment is likely coming soon.

The novelty is that serotonin doesn’t just copy this value signal raw. According to the new model, serotonin neurons apply a kind of filter, exaggerating sudden changes (like the surprise of an unexpected cookie) while compressing slow, predictable trends (like the routine of brushing your teeth). This “prospective coding” makes the signal more efficient: it highlights rare, meaningful shifts without wasting energy broadcasting the obvious.

Once you view serotonin as a prospective value code, a lot of odd behaviors suddenly make sense! Rewards and punishments both activate serotonin. That’s not a contradiction: the system is tracking changes in expected value, whether it’s going up (a reward) or bouncing back from something bad (the end of a punishment). Surprising rewards are exciting, but surprising punishments aren’t. Why? Because in the math of the model, an unexpected reward produces overlapping “peaks” in the signal, while an unexpected punishment cancels itself out. Slow ramps, fast spikes, or steady plateaus all appear. Depending on the timing of cues and the brain’s own “discount rate” (how far into the future it looks), the same underlying code can generate different-looking patterns.

In short, the messy diversity of serotonin responses seen in experiments may just be different faces of the same underlying principle.

Putting the model to the test

The researchers didn’t stop at theory. They tested their model against real recordings from mouse brains. In one set of experiments, mice heard a sound that signaled a water drop was coming a couple of seconds later. Sure enough, serotonin neurons fired in patterns that matched the prospective value predictions: a burst at the cue, a dip during consumption, and a reset afterwards. In another dataset, where rewards were given unpredictably, the model correctly predicted how some neurons tracked not just the current outcome but also the history of recent rewards. That suggests serotonin may help animals weigh whether persistence is worthwhile, should you keep waiting, or give up and try something else?

When compared head-to-head with rival theories (reward, surprise, salience), the prospective value model consistently did better at explaining both individual neurons and the population as a whole.

Serotonin is deeply tied to mental health. Drugs like SSRIs (selective serotonin reuptake inhibitors) alter serotonin levels and are widely used to treat depression and anxiety. Yet exactly what serotonin does in the brain has remained frustratingly vague. A clearer computational theory could help refine treatments or inspire new ones. There are also big-picture lessons about how the brain predicts and prepares. Just as dopamine has been linked to learning from mistakes and adjusting behavior, serotonin may be the partner system that helps us plan patiently, tolerate uncertainty, and keep going in the face of setbacks. In fact, the authors suggest serotonin may encourage “patient waiting”, by boosting the sense that rewards are still out there, even if delayed. That’s a striking idea when you think about procrastination, motivation, or resilience.

Many questions are still open. If serotonin encodes prospective value, do different groups of neurons track different time horizons, some looking seconds ahead, others minutes? How does this coding interact with dopamine, its famous counterpart? And how might this explain the diversity of serotonin’s effects across mood, cognition, and social behavior?

For now, one thing is clear: serotonin is a forecaster, a patient accountant of the near future: when you’re waiting for that ice cream, remember, part of your brain is already running the numbers, predicting the value of what comes next. And thanks to this new research, we’re one step closer to understanding how.

If you want to learn more, the original article titled "A prospective code for value in the serotonin system" on Nature at https://www.nature.com/articles/s41586-025-08731-7