Andrew HubermanEpisode 294 Ralph Adolphs
Andrew Huberman
An Ice Bath Changed His Anger Response 0:00
Andrew Huberman opens by describing a moment of being honked at while parking, noting the immediate anger this normally triggers. He explains that after taking up ice baths, this reaction became flat, meaning his automatic emotional response to a psychological stressor was downregulated, a change he believes was trained and generalized from cold exposure. He is careful to call this a personal experiment rather than a published finding, but suggests that emotional regulation can become automatic and effortless with the right kind of training, rather than something you have to consciously fight through each time.
Introducing Ralph Adolfs and the Episode Topics 2:00
Huberman introduces his guest, Ralph Adolfs, a professor of psychology, neuroscience, and biology at Caltech and a leading expert on the neurobiology of emotions. The conversation is framed around what emotions actually are, how they grow and shrink in the brain and body, why some emotions strongly drive behavior while others do not, and where emotions are mapped physically. Huberman also flags that they will address common myths about emotions being stored in the body, and will discuss why some people can observe their emotions without being overtaken by them.
Defining Emotions by What They Do 4:00
Adolfs explains that his approach, developed with David Anderson in their earlier book and expanded in his forthcoming one with Princeton University Press, treats emotions as functional states defined by what they do rather than by their exact physical wiring in a given species. He places emotions between rigid reflexes, like yanking your hand off a hot stove, and fully flexible, goal-directed planning. A charging bear or tiger is his example of a challenge too varied for a fixed reflex but too urgent for slow deliberation, which is exactly the gap emotions fill by offering a flexible but fast response system.
Why the Question Is Answerable 6:00
Adolfs pushes back on the idea, raised by philosopher Paul Griffiths in a 1997 book, that emotions might not be a coherent scientific category at all. He argues the disagreement among researchers like Joseph LeDoux, Lisa Feldman Barrett, and Paul Ekman doesn't mean the concept is empty. His method is to list clear examples, such as anger, fear, and disgust, the way you might list a Honda or a Toyota as examples of cars, and then ask what shared features make them all emotions.
Four Features That Define an Emotion 8:00
Drawing on his work with David Anderson, Adolfs outlines the defining features of emotions. Priority means an emotion can interrupt ongoing behavior, much like a reflex, a point he traces to Herbert Simon's work on interrupt mechanisms. Valence, a concept Darwin also noted, is the basic pleasant-or-unpleasant, approach-or-avoid quality that lets you compare emotions to one another. Scalability, studied extensively by David Anderson in mice, is the ability of an emotion like fear to grow from wariness to panic depending on how close the threat gets, something a binary reflex cannot do. Temporal persistence is the fourth feature, meaning an emotional state like fear or aggression lingers well beyond the triggering moment, since a bear might still be lurking nearby.
Sadness Without Any Memory of Why 13:00
Adolfs describes a study from his time at the University of Iowa with patients who had severe amnesia from hippocampal damage, unable to retain new memories for more than a minute or so. These patients and healthy controls both watched sad movie clips and reported feeling sad immediately afterward. Five minutes later, the controls still felt somewhat sad and could explain why, but the amnesic patients also still felt sad despite having no memory of ever watching a movie. Adolfs presents this as clean evidence that emotional persistence operates independently of declarative memory, unlike a reflex, which has no such lingering quality at all.
Do Animals and AI Have Emotions 18:00
Asked whether animals meet these functional criteria, Adolfs answers plainly that they do, including the states we recognize in pets like dogs and cats. He also raises the more surprising claim, made by the company Anthropic, that large language models might have functional emotions as well, since the criteria are defined by function rather than biology. The complication arises with conscious experience, which most people intuitively treat as the real proof of an emotion, since we normally judge our own sadness or fear by how it feels.
Separating Emotion From Consciousness 20:30
Adolfs argues that tying emotion to conscious feeling creates two serious scientific problems. First, it becomes nearly impossible to test whether animals have emotions, since there is no reliable way to test for conscious experience in them, even though he personally has no doubt they do. Second, if emotion is treated as identical to conscious feeling, then studying emotion actually becomes studying consciousness, a problem he considers far harder and one he would rather avoid, much as vision or memory researchers study those processes without needing to solve the puzzle of conscious experience itself. Keeping emotion separate from consciousness, he says, is what allows an integrated science of emotion across species, including humans, animals, and even machines, to move forward.
Watching Your Own Anger Rise 23:00
Huberman raises the everyday experience of feeling an emotion build without fully taking over, like standing in a slow-moving line and feeling frustration rise while still being able to observe it and choose whether to speak up. He contrasts this with the opposite extreme, viral videos of people completely losing control in public, such as on airplanes. Adolfs affirms that this human capacity to consciously monitor one's own emotions, and thereby regulate them, is likely unique or at least highly developed compared to other animals. He credits Stanford's James Gross with founding the field of emotion regulation in the 1990s, noting that it has grown so large that roughly a quarter to a third of submissions to the journal Affective Science, where Adolfs recently finished a term as co-editor, deal with this topic.
Emotions Aren't the Problem, Control Is 25:32
Most emotional pathology, Adolphs says, is not about the emotion itself but about our ability or inability to regulate it. High performers like tier-one special operators or elite climbers often describe pushing emotions aside during peak performance, yet in their personal lives they remain loving and emotionally healthy. The ideal, he suggests, is not suppressing emotion altogether but knowing when to let it run and when to hold it back.
Why Emotions Are Necessary 27:00
Drawing on Antonio Damasio's 1994 work Descartes' Error, Adolphs explains that patients with major frontal lobe damage who stop emotionally responding to things become unable to make decisions or organize their lives, because they no longer feel one way or another about their options. Losing emotion entirely is disabling, even if regulating emotion well is also a valuable skill.
Naming Feelings and Emotional Granularity 29:01
Adolphs credits Lisa Feldman Barrett with pioneering work on emotional granularity, the fine-grained ability to identify and conceptualize your own emotions. He argues you do not need actual words to do this, since people think in words, pictures, or a mix, and even languages lacking a term for an emotion, like the German schadenfreude or the untranslatable concept of something being unbearably cute, still let speakers grasp the underlying concept. Having some way to differentiate what you are feeling, verbal or not, gives you the purchase needed to apply strategies like cognitive reappraisal or acceptance.
The Danger of Emotional Rabbit Holes 31:01
Adolphs describes dynamic attractor states, where dwelling on a feeling pulls more and more mental attention toward it, deepening sadness or anxiety the more you focus on trying to resist it. This spiraling is trainable to avoid, and the most effective strategy is stopping the process at the earliest point, ideally by avoiding situations that trigger unwanted emotions in the first place, since once you're deep into the spiral it becomes very hard to pull out.
Ice Baths and Automatic Regulation 33:00
Adolphs took up ice baths after a colleague mentioned Wim Hof, and now uses a large insulated trough at home with weekly ice deliveries. The first sessions were painful, with heart rate and breathing spiking, but over time his body adapted until entering the ice produced calm rather than stress. He noticed this training generalized: a reflexive anger response to being honked at while parking, once automatic, became flat and unbothered after sustained ice bath practice, suggesting an autonomic training effect on emotion regulation beyond just cognitive strategy.
Ultramarathons and Long-Term Emotional Endurance 37:30
Adolphs, training for a 100-mile race, describes the extreme nonlinearity of ultrarunning, where feeling like complete garbage at mile 40 or 50 can give way to recovery if you keep going, embodying a mindset of relentless optimism rather than just relentless forward momentum. He believes almost anyone healthy could complete such a race if they wanted to, and sees it as a rare, sustained exercise in not giving up on something that initially seems impossible.
Awe as a Uniquely Human Emotion 42:31
Adolphs connects endurance and difficulty to gratitude and awe, describing a felt sense of appreciation simply for existing and talking with another person. Building on Dacher Keller's definition of awe as shifting attention across spatial scale, Adolphs extends it to time, arguing that awe depends on a distinctly human metacognitive ability to mentally time travel and adopt perspectives detached from the immediate present, an ability he sees as central to human consciousness, planning, and storytelling, and largely absent in other animals.
Rethinking Facial Expressions and Fear 46:30
Adolphs recounts his early work on patient SM, who had bilateral amygdala damage and could not perceive fear in facial expressions, and later, with Justin Feinstein, showed she also failed to consciously experience fear herself when exposed to haunted houses, snakes, and horror films, though this reflects a specific deficit rather than a single fear center. He then challenges Paul Ekman's influential claim that six basic emotions are universally read from facial expressions, noting that a paper he co-authored with Lisa Feldman Barrett found real-world facial expressions are far more variable than the posed, exaggerated faces used in classic studies, and that multiple-choice matching tasks overstate how well people actually read emotion from faces, a point echoed by research showing dog owners are no better than chance at detecting canine guilt.
Emotion Reading Is Overconfident 51:01
You tend to overestimate how accurately you can read emotions from faces alone, because in real life you never rely on a single static expression. Instead you probe further, ask questions, and weigh context, which is why lab studies using isolated photos of faces underestimate how emotion recognition actually works.
Reading People Through Text and Change 52:00
Much of how you judge another person's emotional state comes from noticing shifts over time, like a partner being quieter than usual in the morning, rather than from reading a fixed expression. This same pattern shows up in writing: a study found that large language models like GPT or Claude, given only a page of someone's diary-style text, could infer personality traits on the Big Five scale as accurately as a close friend or clinician, showing how much information sits in the words and pacing of what people write.
Prediction and Dynamic Faces 56:00
Lisa Feldman Barrett's recent thinking on emotion centers on prediction, meaning most emotional information comes from mismatches between what you expect and what actually happens. Static face photos used in early studies, including Ralph's own postdoc work with Paul Ekman's images, don't reflect how people actually perceive emotion in real life, since researchers like Philip Schyns and Rachel Jack have shown the brain reads emotion from how a face changes over time, not from a fixed pose.
Patient SM and Separate Fear Systems 57:32
Patient SM, who had bilateral amygdala damage, could still experience autonomic arousal and stress, showing that emotion isn't one single system that simply switches off. She showed no fear toward external threats like haunted houses or snakes, but when she and other amygdala-damaged patients inhaled carbon dioxide, which normally triggers a suffocation-like panic in about half of people by changing blood pH, they had full-blown panic attacks. This means fear of external threats and panic from internal bodily signals rely on distinct brain circuits, with the amygdala needed for the former but not the latter.
Where Emotions Are Felt in the Body 1:00:31
A well-known bodily maps study, led by Lauri Nummenmaa and colleagues in Finland, asked people to mark where on a body outline they would feel emotions like sadness or anger, producing heat-map style images. Ralph points out this measures people's concepts of emotion rather than actual physiological change, since no bodily signals were recorded. Real bodily changes during felt emotion are represented in the insula, a brain region that receives input from organs throughout the body, and Ralph describes his own experience with a kidney stone in Switzerland as an example of an intensely painful, body-driven state that would likely activate this region strongly.
Body Signatures of Emotion Remain Unmapped 1:08:31
Ralph pushes back gently on Lisa Feldman Barrett's claim that there is no systematic link between specific emotions and specific bodily changes, arguing that existing studies only measure a few crude signals like heart rate and blood pressure rather than a detailed, high-dimensional readout of the whole body. He suspects a real bodily signature for each emotion likely exists but hasn't yet been measured, especially given that Barrett and Tor Wager's own neuroimaging work has found distributed but reliable brain activation patterns for basic emotions like anger, fear, and sadness.
Emotion, Attention, and Advertising 1:11:01
Andrew connects this to advertising and media, which are built to capture attention by prioritizing emotion until it drives behavior, whether that means clicking, buying, voting, or forming group loyalties. He notes that attention-grabbing content tends to drift toward primitive, hypothalamus-linked responses like sex and violence, with societal and ethical limits acting as the only brake on how far that drift goes.
Can Art Trigger Shared Emotions 1:14:30
Andrew asks whether abstract art or music, without any faces, scenes, or lyrics, can still trigger the same emotional circuits across many different people. Ralph points to an early PET imaging study by Anne Blood and Robert Zatorre from the mid-1990s, which played the same piece of music to different listeners and measured whether it gave them chills or goosebumps down the spine. The same sound produced different emotional responses in different people, and imaging showed the insula was involved in generating that felt experience.
Abstract Art and Evolved Emotion Circuits 1:16:00
Brain structures like the amygdala activate not just for survival threats but also for chills triggered by music, and Caltech researcher John O'Doherty showed that abstract art can be broken down into basic visual features fed into a neural network to predict whether people find a piece pleasant or unpleasant. Emotions did not evolve to appreciate art, they evolved to detect threats like tigers and bears, but the underlying visual and auditory features that once signaled danger or reward seem to resurface in abstract stimuli, letting art and music evoke real emotional states through features that generalize from the natural world.
Hardwired or Learned Face Processing 1:18:01
Ralph Adolphs explains that face processing in the brain is both innate and shaped by experience, never purely one or the other. He points to Rebecca Saxe's imaging studies of babies to show how face-processing systems develop, and contrasts this with the brain's word-recognition region, which cannot be evolutionarily hardwired since written language is too recent, yet is already wired in preverbal infants in a way that primes it for reading once literacy is learned. Asked whether single neurons respond to smiling versus frowning faces, he notes such neurons exist for almost anything but are noisy and unimportant on their own. What matters is population-level activity across thousands of neurons, which his work with epilepsy patients has shown can reliably decode smiles from frowns, since that distinguishing information must exist somewhere in the brain for people to tell the two apart.
Abstraction as the Core of Emotion 1:23:02
Adolphs describes how raw visual input, just pixels and blobby shapes at the retina, gets built up into increasingly abstract representations, similar to how deep convolutional neural networks recognize objects regardless of angle, distance, or lighting. Emotions work the same way: a bear, a tiger, a swerving pickup truck, a snapped twig, or someone shouting a warning can all funnel into the same central fear state. He and David Anderson describe this as a fan-in fan-out architecture, where many different inputs converge on one internal state, which then fans out into many possible responses like hiding, running, or fighting back. This flexibility is what separates emotions from rigid reflexes, which have only one trigger and one fixed response.
Reading the Room, or Not 1:26:31
Huberman raises the observation that some of the most functional, caring people he knows are not finely attuned to every emotional nuance in a room, while some less functional people track every shift in others' moods. Adolphs responds that emotional intelligence is highly heterogeneous, and that the field has moved from focusing on single regulation strategies like suppression or reappraisal toward emphasizing variability and flexibility, including the interpersonal, bidirectional work of regulating other people's emotions as well as your own. He compares this to heart rate variability, where the goal isn't a single fixed state but the capacity to range widely and adapt to context, and notes this ties to Lisa Feldman Barrett's concept of emotional granularity, the ability to finely differentiate one's own emotional states, which underlies flexible regulation.
Autism, Eye Contact, and Cognitive Load 1:34:32
Turning to autism, excluding severe cases requiring lifelong care, Adolphs frames his interest as understanding individual differences in social awareness generally, with autism spectrum traits offering a well-studied dimension to explore. He discusses being overwhelmed by constantly having to consciously track social rules like eye contact and turn-taking, which is exhausting unless it becomes automatic, something many autistic people struggle to achieve even as some compensate. His lab now uses webcams to measure gaze during videos of people talking, finding that those scoring higher on autism spectrum traits make less eye contact and are more easily distracted by background elements like a television. This gaze pattern appears to be a temporally stable, personality-like trait, consistent across a year of monthly testing and only modestly affected by factors like poor sleep or stress, though whether it predicts anything like career success remains an open question.
Studying Faces and Longitudinal Change 1:41:31
Ralph Adolphs describes a longitudinal study that tracks people over time with extensive questionnaires on personality and mood, aiming to see which traits can shift and which stay fixed. The goal is to figure out whether looking at faces a lot has measurable consequences, and more broadly whether traits like extroversion or neuroticism can be changed, or whether they form an unchangeable core of personality.
Bodily Movement and Thought 1:43:00
Adolphs and his guest discuss how some people seem to run at a higher internal RPM, moving and gesturing constantly, while others stay physically still, and they compare this to differences among dog breeds like bulldogs versus Yorkshire terriers. They agree that bodily movement is not separate from thinking but interfaces with it, since hand and mouth representations sit close to language centers in the brain, meaning gestures likely co-evolved with speech and help structure thought rather than just accompany it.
Zoom Versus Face-to-Face Interaction 1:46:01
Adolphs recalls the abrupt shift to Zoom-only interaction during the 2020 lockdowns and how strange it felt to lose face-to-face contact. His lab is now running experiments that parameterize the degree of social realness in an interaction, comparing in-person contact, video calls, and video with the camera off, especially to see how these differences affect people with autism.
Solitude and Emotional Training 1:48:00
The conversation turns to the value of spending time alone, in flow, or in one's own head, using examples like long solo runs and climber Alex Honnold's intense focus. Both speakers suggest that tolerating one's own internal dialogue without needing to engage the outside world trains emotional regulation and metacognitive control, and that this kind of solitary training, whether through meditation, running, or cold exposure, builds resilience that constant social media use does not allow.
Lab Meeting Silence and Task Switching 1:52:31
Adolphs explains that his lab now begins every Monday meeting with a few minutes of silence, a practice that started in 2020 as a moment of remembrance for George Floyd and evolved into a short meditation to help people mentally reset before a research talk. He connects this to the broader topic of task switching, noting there is a well-documented cost every time the brain shifts from one task to another, and that anticipating and managing that switch is a trainable skill relevant to both human performance and AI design.
Discipline, Meditation, and Resilience 1:57:02
Drawing on an anecdote about a former Navy SEAL who mentally rehearses his transition from work mode to home life, the speakers argue that the ability to switch states deliberately is a trainable and valuable skill, one that meditation practices, including those studied by Richard Davidson in long-term meditators, show measurable health benefits for even in short daily doses.
Facing Illness Through a Neuroscience Lens 2:02:00
Adolphs shares that he recently overcame an illness, describing how the diagnosis brought fear, sadness, and worry that he and his wife worked through together, which gradually gave way to gratitude, awe, and acceptance. He reflects that the experience sharpened his sense of how much time he has left and how to use it, including his decision to train for a hundred-mile ultramarathon despite treatment side effects, and he closes with a neuroscience-based reframing of death: rather than the world losing you, you lose the world, since the world as experienced is a construction of your own brain.
Closing thanks and farewell 2:05:31
Andrew Huberman closes the conversation by thanking Ralph Adolphs for his openness and for the years of high-level scientific work he has done, noting how rare it is for scientists to speak this candidly. Adolphs responds warmly, saying he hopes to report back on finishing his hundred mile race and finishing his book.
Podcast support requests 2:07:01
Huberman thanks listeners and asks them to subscribe on YouTube, follow on Spotify and Apple, leave a five star review, leave comments, and check out the episode sponsors as ways to support the podcast at no cost.
New book and social media 2:07:31
Huberman announces his first book, Protocols: An Operating Manual for the Human Body, five years in the making and based on over thirty years of research, covering sleep, exercise, stress, focus, and motivation, available for presale at protocolsbook.com. He also points listeners to his Huberman Lab accounts across social platforms.
Neural Network newsletter mention 2:08:30
He describes the free monthly Neural Network newsletter, which includes podcast summaries and short protocol PDFs on topics like sleep, dopamine, cold exposure, and fitness, available by signing up at hubermanlab.com without email sharing.
AI-generated summary. It can be wrong or incomplete - check anything that matters against the original.
