Dr. Read Montague: How Dopamine and Serotonin Really Shape Learning, Motivation and Decisions

High-Level Overview

Andrew Huberman talks with Dr. Read Montague, director of the Center for Human Neuroscience Research at Virginia Tech and a pioneer in measuring dopamine and serotonin in living, conscious humans. The conversation takes apart the popular picture of dopamine as a pleasure chemical delivered in hits. In Montague's account dopamine is first of all a learning signal, one that tracks how our expectations change from moment to moment, long before any reward arrives, and that same signal sets how motivated we should be. Serotonin, he argues, works in opposition to it and carries information about negative outcomes and waiting.

From that base the two cover Parkinson's disease, ADHD and the explorer versus focus modes found even in bees, short-form video and phones, what stress and hunger do to the reward system, a striking claim about how SSRIs interact with dopamine terminals, new ways of recording neurochemistry through the nose, and how the brain's learning rule became the engine of modern AI. The tone is collegial, curious and often funny, with Montague dry and self-deprecating and Huberman supplying everyday examples. It moves between mechanism, personal stories about sport, sleep and science careers, and listener questions at the end.

The Discussion Explores:

• Dopamine as a learning signal built on successive predictions, not simply expectation versus reward
• Why the same learning rule found in bees and humans also powers game-playing AI
• Baseline versus fast dopamine changes, and Parkinson's disease as a flattened sense of value
• Explorer and exploiter modes, ADHD, short-form video, and keeping phones out of reach
• Serotonin as dopamine's opponent, and Montague's argument about what SSRIs do at dopamine terminals
• How hunger, severe stress and trauma can flip what the dopamine system treats as worth learning
• Measuring dopamine and serotonin in people, including through a probe in the nose
• Sport, failure and science as training for long-term motivation

Deep Dive into Main Ideas

Dopamine is a learning signal about changing expectations

Montague starts with what has changed. It used to be thought that dopamine equals pleasure. The newer work, much of it coming from what is now called artificial intelligence, shows that dopamine fluctuations control learning, play a role in motivation, and may also relate to how we feel, though that last link is the least understood.

The familiar textbook idea is reward prediction error: you expect something, you get an outcome, and the difference teaches you. Montague says psychologists had that worked out decades ago, in the Rescorla-Wagner rule of 1972, but that it models reality poorly, because reality usually gives long stretches of nothing. The insight he credits to Rich Sutton and Andy Barto is to learn from successive predictions. His example: on Wednesday you predict two inches of rain for Saturday; on Thursday, with new information, you predict ten. Nothing has happened yet, but the gap between the two predictions is itself an error signal, called a temporal difference error, and dopamine appears to encode it. A forager, or a player forty moves from the end of a game, learns at every step this way.

Huberman offers a dating example: a friend whose excitement about a new person rises and falls with each text message and each new piece of information, with no clear final reward in sight. Montague says that is exactly the idea. He adds that the simpler rule cannot chain learning backwards through a sequence of cues, while this one can. He recalls working on it around 1990 and finding that it could explain every figure in the published dopamine recordings of Wolfram Schultz.

The algorithm that crawled out of the brain

The complaint back then, Montague says, was that reinforcement learning of this kind could not learn anything impressive. That was true at the time. Then the same algorithm was used to build AlphaGo and AlphaGo Zero, the system trained from scratch that he says has never been beaten, and the approach contributed to the protein structure work that won Nobel Prizes. He finds the convergence remarkable: a learning rule that runs in a small set of brainstem neurons, in songbirds, sea slugs and honeybees as well as humans, has been externalized into programs that now outperform us. Neither man can think of another example like it.

Later in the conversation both say they use large language models. Huberman uses Claude to compare and contrast research positions, something he cannot do in a literature database. Montague describes asking one a deliberately odd question linking the subjunctive mood in grammar to the mathematics of quantum mechanics, and getting back an essay he says no person he knows could have written. He never expected neural network training to scale the way it has, and he expects such networks to become standard tools for analyzing human neuroscience data.

Motivation, the forward drive, and why hits is a blunt word

Where does motivation come in? Montague points to an idea from other researchers that prediction errors are ideal signals for deciding how much you should want something. Fast fluctuations track changing expectations, and a slower envelope on top of them looks like motivation, so one system can do both jobs.

Huberman asks whether social media and other businesses are built to keep updating our expectations so that we keep foraging. Montague's answer is one of the central lines of the episode. "If there was a final outcome, you wouldn't keep living." If any achieved goal were enough, the system would stop; instead, "Your nervous system keeps pushing you forward." On the popular phrase dopamine hits, he says an unexpected reward does cause a dopamine fluctuation, so the phrase is incomplete, not false: "It's not wrong, but it is blunt."

He distinguishes slower, tonic levels from the quick changes riding on them. A wonderful evening fills the well a little; a disappointment or a depleting drug lowers it, and that changes how the fast signals affect learning. In Parkinson's disease, he says, by the time symptoms bring someone to a doctor they have lost 70 to 75 percent of the dopamine neurons in the brainstem. The remaining signal is so noisy that the brain cannot read differences in value between possible actions, so it stays put. He thinks of it as an active freezing disease. On drugs of abuse and overindulgence generally, he says that on average it "resets expectations where very few, if any, natural events can exceed them".

Explorers, exploiters, and screens

Asked about ADHD and stimulant medication, Montague guesses that raising dopamine and norepinephrine stabilizes brain states and thought sequences so they do not divert. He then describes a collaborator's bees. Some bees fly straight to the nectar source they have been told about; others set off and get distracted, and in doing so discover new sources. A hive needs both, exploitation and exploration, and so does a person or a company. "You've got multiple bees inside your head." He adds that in some settings, such as combat, rapid, wide-ranging attention is required, and people train hard for it.

Huberman wonders whether constant short-form video strengthens the rapid-switching circuits at the expense of slow, goal-directed ones. Montague's reply, "you build your ADHD muscle", comes with a caveat: he does not know of evidence in people, only that artificial systems must be carefully balanced so they do not overtrain on one mode. As a father of five he calls screen time his main nemesis and a battle he is losing. He has read The Anxious Generation and respects its author, but says nobody really knows what phones are doing overall and that collective fixes are hard. He praises his children's school for banning phones during the school day.

Huberman describes his own approach: social media lives on an old phone locked in a box for most of the day, after he read that a phone in the same room lowers cognitive performance. He asks whether the dopamine system can reward resisting a behavior. Montague says yes, with anorexia as the pathological example, and recalls relishing hard training while others slept. Both suspect that effortful, slower activities such as reading produce more lasting learning, though Montague notes it is unclear whether effort or slowness is the cause. Huberman mentions one short clip that did stay with him, a guest saying the point of life is to "learn to enjoy the passage of time".

Serotonin as the opponent, and the SSRI argument

Montague says his is the only group recording dopamine and serotonin at sub-second speed in conscious humans doing tasks. The theme that emerges is opponency: when dopamine rises, serotonin falls, and the reverse. Dopamine goes with positive events or their anticipation; serotonin has inherited the negative side and active inhibition. "Serotonin tells you to get ready to wait." In economic games played by patients, he sees dopamine up and serotonin down when things are going well, and the opposite when they are not.

Then comes his claim about SSRIs, which should be read as his interpretation. These drugs block serotonin reuptake, and he acknowledges that for some people they are life changing and that responses vary widely. But he cites a 2005 rodent study showing that the extra serotonin is taken up into dopamine terminals through the dopamine transporter. If serotonin is the negative signal and it ends up being released from neurons that fire for positive things, he reasons, a person could have a hard time learning about positive things and might negatively condition on things they should pursue. He says he does not know that this has been followed up in humans. Later, answering listener questions, he goes further and calls these drugs heterogeneous in effect, often pathological, and toxic over decades, and says that with psychotropic medication 50 to 80 percent of a good outcome is placebo, "but that doesn't mean it's a fake effect". Huberman responds that he knows people who swear by SSRIs with minimal side effects; Montague agrees it would be valuable to identify in advance who will benefit. On side effects, he says the drugs bind to many of the numerous serotonin receptor types.

Hunger, stress and trauma flip the system

In rodents made very hungry, Montague says, dopamine in the amygdala comes to encode something like punishment prediction errors instead of reward prediction errors. He explains the logic: in an emergency, staying alive matters more than chasing rewards, so the system redirects attention and motivation to predicting bad events. His group is now testing patients before and after eating.

This connects to trauma. A catastrophic, unexpected event leads a nervous system to overgeneralize, which he describes as rational and as the pattern seen in PTSD. For that reason, "Training animals with really negative feedback is a really bad thing to do", and the same goes for a teacher who punishes a child for a wrong answer. He remembers a badly abused dog he rescued as a boy that never recovered; for it, basic safety had become the reward. Huberman describes a friend who took about three years in a calm relationship to stop bracing for trouble.

Recording neurochemistry in people, and up the nose

The human recordings are made during brain surgery, with consent, in patients receiving deep brain stimulation electrodes for Parkinson's disease or essential tremor, or being monitored for epilepsy. A neural network model interprets the electrical signal on the electrode as fluctuations in dopamine, serotonin, norepinephrine and other measures. Montague says the method took well over a decade to make work across sites.

The newer approach comes from a collaboration with Christina Zelano at Northwestern: a thin, flexible electrode is passed up the nose to rest against the olfactory tissue. Healthy volunteers can do this, which removes the objection that all the data come from patients. He says the nasal signals look like what would be expected from midbrain recordings. In breathing experiments, neuromodulators cycle with relaxed breathing like a metronome, and in an ultimatum game breathing, a marker of mitochondrial activity and the dopamine signal line up around moments that require learning. "It's the most amazing looking data I have ever seen." A postdoc of his is starting a company to turn the probe into something people could use themselves, which Montague imagines helping a child learn to concentrate.

Sleep, sport, and science as a contact sport

Montague, who is 66, says he has always slept little, in two blocks, and gets up around 3:30 in the morning because he needs quiet to think; he stresses that sleep need varies between people. A former decathlete and wrestler, he has made all his children do a sport. The value, for him, is learning to lose after giving everything, and to manage panic. "I can't teach a kid a lesson that good." He says a divorce taught him to be less judgmental, and that "science is a contact sport" at the leading edge, where any good result is attacked. Huberman adds that research trained him to expect big rewards only every couple of years.

In the listener questions, Montague says the biggest public misunderstanding is that dopamine equals pleasure, calls a dopamine theory of schizophrenia rational but incomplete, and says the question of grit versus sunk cost cannot yet be answered because so little is known about how the brain sets expectations.

Conclusion

The picture of dopamine that Montague offers is richer and more useful than the pleasure story. Dopamine teaches by tracking how expectations shift, it sets the level of drive, and it never settles, because a creature that was satisfied would stop moving. Serotonin pulls the other way and handles the negative side, and stress or deprivation can turn the whole arrangement toward survival. His view of SSRIs is provocative and comes with his own admission that the human follow-up has not been done, and Huberman notes that many people do well on them. The practical threads are modest and sensible: favor slower, effortful learning, keep the phone at a distance, avoid harsh punishment as a teaching tool, and give children sport so they learn to fail and try again. Montague is optimistic that the same algorithm his field once struggled to defend will now help explain the brain it came from.

This is a summary of "How Dopamine & Serotonin Shape Decisions, Motivation & Learning | Dr. Read Montague" from Huberman Lab, giving you the GIST so you can decide if you want the REST. Listen to the full episode.