Understand & Improve Memory Using Science-Based Tools

High-Level Overview

In this solo episode, Andrew Huberman sets out to answer a deceptively simple question: why do we remember anything at all? Every day we are flooded with sensations, perceive only a fraction of them, and remember a tiny subset of those perceptions. Huberman explains that "memory is simply a bias in which perceptions will be replayed again in the future", a bias in the likelihood that a particular chain of neurons will fire again in the same sequence. He then shows how that bias is created, either by repetition or by a single intense experience, and how an understanding of the chemistry involved can be turned into practical tools for learning faster and remembering longer.

The tone is enthusiastic and teacherly, with personal stories (losing a coin flip to another Andrew in third grade, being upset to learn he had been caffeinating at the wrong time for years) woven through classic experiments and recent studies. The conversation moves from the basics of neural circuits and types of memory, through the famous patient HM, to the landmark work of James McGaugh and Larry Cahill on adrenaline, then to exercise and bone hormones, photographic and mental-snapshot memory, deja vu, and finally a short daily meditation practice that enhances memory, with an important caveat about when to do it.

The Discussion Explores:

• Memory as a biased replay of neural circuits, built by repetition or strong one-time activation
• Ebbinghaus's learning curves and Hebb's postulate that neurons firing together strengthen their connections
• Short-term, working and long-term memory; explicit versus implicit memory; and the hippocampus
• The case of patient HM and what it revealed about forming versus storing memories
• McGaugh and Cahill's discovery that adrenaline after learning stamps in memories, and why timing matters
• Practical ways to spike adrenaline after learning, and why chronic stress backfires
• Exercise, neurogenesis and osteocalcin, a hormone from bone that supports the hippocampus
• Mental snapshots, deja vu, and a 13-minute daily meditation that improves memory and attention

Deep Dive into Main Ideas

What a Memory Actually Is

Huberman begins with the basics. The nervous system converts non-negotiable physical events, photons, sound waves, touch, smells and tastes, into electrical and chemical signals. But we only perceive a small subset of what we sense; if you are listening to him, you are probably ignoring the feeling of your skin against your chair. A memory is a bias in the likelihood that the specific chain of neurons active during an experience will be activated again. If you can remember your name, it is because a particular sequence of neurons fires when you do.

Memories also exist in context. Huberman's own name carries a story: in third grade there were two Andrews, he lost the coin flip, and he was called Andy until about 12th grade before reclaiming Andrew. Each memory is linked to others through close, medium or distant associations, which is how memory champions remember dozens of names by building mental associations. Most of us find that effortful, but there are ways to leverage the brain's natural biology instead.

Repetition, Ebbinghaus and Hebb

The most basic and reliable tool is repetition. In the late 1800s and early 1900s, Hermann Ebbinghaus quantified learning by repeatedly reading sequences of words or numbers and writing down as many as he could recall. Early on it took many repetitions; later, far fewer. He called that reduction savings, and the resulting learning curves, Huberman says, liberated the study of learning from the philosophers. Remarkably, the material got no more interesting with each repetition, yet sheer repetition was enough.

Later, psychologist Donald Hebb proposed that when neurons fire at about the same time, the connections between them strengthen, which decades of experiments have since confirmed. Huberman stresses that most learning does not come from new neurons or new connections but from existing neurons strengthening their connections, either through repetition or through a single, very strong activation, known as one trial learning. That is why we can remember something forever after experiencing it once, whether it is a terrible event or a wonderful one like first seeing a romantic partner or a newborn child.

Types of Memory and the Lessons of Patient HM

For the episode, Huberman focuses on short-term memory, including working memory (holding a phone number or a six-digit security code in mind just long enough to use it), and long-term memory for cognitive or motor information. He also distinguishes explicit memories, things you can declare you know, such as your name or how to walk step by step, from implicit memories, things your nervous system knows without you thinking about them, like simply getting up and walking. Explicit declarative memories are formed in the hippocampus, which he thinks looks less like a seahorse and more like two halves of a cinnamon roll pushed together, while implicit memories are formed and stored in areas such as the cerebellum and neocortex.

The patient HM had severe, frequent grand mal seizures originating in the hippocampus. Surgeons destroyed his hippocampus on both sides, and afterward he could not form new explicit memories: he could introduce himself to you, turn away, turn back and ask your name again. Yet he kept old memories, showing the hippocampus forms but does not store memories, and he retained implicit skills like walking and making coffee. Tellingly, if you told him the same joke repeatedly, he found it a little less funny each time, even though he couldn't recall hearing it. Huberman uses this to introduce the central theme: emotion, and the chemistry underlying it, is the lever that can strengthen memory.

McGaugh and Cahill: Adrenaline Stamps In Memory

Huberman describes the decades-long work of James McGaugh and Larry Cahill. In a key experiment, people read a 12-sentence paragraph that was either mundane or included emotionally intense content, positive or negative, then later received a surprise quiz. The emotional paragraph was remembered far better, echoing Francis Bacon's 1620 observation that "memory is assisted by anything that makes an impression on a powerful passion, inspiring fear, for example, or wonder, shame or joy." What made McGaugh and Cahill's work groundbreaking came next. In animals, one trial learning of where a shock occurred (conditioned place aversion) or where food or a mate was found (conditioned place preference) depends on epinephrine; block it pharmacologically and the animal doesn't learn. In humans, people who read a boring paragraph and then put their arm into ice water, which releases adrenaline, remembered it as well as emotional material. Giving a beta blocker erased the benefit of both emotional content and adrenaline-raising drugs.

The conclusion, in Huberman's words: "It is not the emotion. It is the neurochemical state that you go into as a consequence of the emotion." Adrenaline is released in the body by the adrenal glands and in the brain by the locus coeruleus, which sprinkles the brain with epinephrine and norepinephrine; cortisol can also cross into the brain. Across many studies, adrenaline increases ranged from zero to 10 percent in some people to six- or seven-fold in others, and the greater the increase, the better people remembered, for languages, math, word lists and physical skills alike.

Timing Is Everything: Spike Adrenaline After Learning

The discovery that most upset Huberman was about timing. Like many people, he had always caffeinated before studying. Caffeine blocks adenosine, which builds up the longer we're awake and causes sleepiness; increases epinephrine release in body and brain; and makes dopamine receptors more effective. But experiments that raised adrenaline at various times, from an hour before learning to 30 minutes after, found the best window is late in the learning session or immediately after, up to about 5, 10 or 15 minutes. So if you use caffeine, alpha-GPC or other stimulants to enhance memory, he suggests taking them near the end of or right after a learning bout.

He clarifies that sleep, naps of about 20 to 90 minutes, and non-sleep deep rest (NSDR) remain essential, since that is when neural circuits are actually rewired, but they can come an hour or several hours later. You don't need pharmacology to spike adrenaline: a cold shower, ice bath or circulating cold bath that is uncomfortably cold but safe, enough to quicken breathing and make you wide-eyed, or a hard run, all work. He urges safety: if you're not used to caffeine, don't suddenly down several espressos, especially if you're prone to panic attacks; start with the minimal effective dose, which can be zero, and check with your doctor. He is opposed to non-prescribed use of drugs like Ritalin, Adderall or modafinil, and those who are prescribed them should consider timing with their physician.

Why Chronic Stress Backfires, and How the Amygdala Links Events

What matters is not the absolute amount of adrenaline but the change relative to the previous hour or two. Staying constantly amped up, for example with alpha-GPC and a double espresso before work and another spike afterward, blunts the effect. Drawing on the work of the late Bruce McEwen and Robert Sapolsky, Huberman notes that chronic stress impairs learning, memory and immune function, while acute spikes enhance them. The ideal is to learn in a calm but alert, focused state, then spike adrenaline afterward. He cites a 2022 review in the journal Neuron, "mechanisms of memory under stress", which opens by describing medieval communities throwing young children into a river after important events so they would remember them for life.

The mechanism centers on the amygdala, which Huberman describes as associated with threat detection and with linking novel events to emotional states of both kinds. "It's a correlation detector": when epinephrine is elevated and a brain circuit is strongly active, it strengthens those connections, like an AND gate in computer programming that needs both conditions. Because the amygdala is more of a lumper than a splitter, and epinephrine is the same molecule whatever triggers it, we tend to generalize. That underlies PTSD, where fear spreads from one room to a whole city block or to any crowded place, but it also means positive experiences can generalize to people, places and things.

Exercise, Neurogenesis and Hormones From Bone

Exercise is another powerful tool. Cardiovascular exercise appears to increase neurogenesis, the birth of new neurons, in the dentate gyrus of the hippocampus, an important exception to the rule that learning doesn't rely on new neurons, though Huberman acknowledges debate over how much adult neurogenesis humans have. He recommends a minimum of 180 to 200 minutes a week of zone 2 cardio, steady effort where you can just barely hold a conversation, which likely helps the hippocampus indirectly by improving blood and lymphatic flow.

A second mechanism involves bone. Bones release osteocalcin, a hormone with endocrine effects, and work from Eric Kandel's lab at Columbia and others shows cardiovascular and especially load-bearing exercise triggers osteocalcin release, which travels to the hippocampus and supports its electrical activity and connections. Kandel, now in his 90s, swims a half mile to a mile a day and credits it for his sharp mind. Huberman argues this is how the brain knows the body is actually moving, since so much brain real estate is devoted to vision and movement. He credits John Ratey of Harvard Medical School, author of the book Spark, with the idea, and shares Ratey's example of sea creatures that settle permanently on a rock and then digest much of their own brain. On timing, Wendy Suzuki's lab at NYU found learning and memory improved for at least two hours after exercise. Huberman's logic: do moderate exercise within one to three hours before learning to boost blood flow and osteocalcin, but place very intense, adrenaline-spiking exercise after a learning bout.

Photographic Memory, Face Recognition and Mental Snapshots

True photographic memory exists but is not necessarily desirable, Huberman says, since such people often struggle with auditory memory and physical skills. Super recognizers, who can match a face from a photo to low-resolution video of a crowd, are prized by government agencies; this reflects the fusiform gyrus, a face recognition area, rather than memory per se. At the other extreme, some people are face blind; he recalls a colleague who would ask whether he was Rich or Andrew.

More useful for most people is a study on how volitional photo taking affects memory for the visual and auditory aspects of an experience. When people chose what to photograph, their memory for those objects, including detailed museum pottery, improved, whether or not they ever looked at or even kept the photos. But their memory for sounds from the same moments got worse, suggesting the visual system can outcompete the auditory system. Most strikingly, simply deciding to take a mental snapshot boosted visual memory almost as much as a real camera. Huberman has done this since childhood and can still picture a New York street scene, with a man in a yellow shirt and some construction, that he deliberately snapshotted from an Uber two years ago.

Deja Vu and a Short Daily Meditation

On deja vu, Huberman cites work by Susumu Tonegawa at MIT, a Nobel laureate for his immunology work who also studies memory, and Mark Mayford at Scripps and UC San Diego. By tagging hippocampal neurons active during learning, they could reactivate them later in the original sequence, in a different sequence or all at once, and all versions evoked the same behavior and, in a sense, the same memory. That is the most mechanistic explanation of deja vu: different firing patterns producing the same sense of familiarity. Since the hippocampus doesn't have infinite sequences available, overlap is to be expected, and he knows of no pathology linked to deja vu.

Finally, he describes a study from Wendy Suzuki's lab titled "a brief daily meditation enhances attention, memory, mood and emotional regulation in non-experienced meditators". Adults aged 18 to 45 either did a 13-minute daily meditation, a body scan with attention returning to the breath, or listened to an equal-length episode of the podcast Radiolab. After eight weeks, but not four, meditators showed significant improvements in attention, memory, mood and emotion regulation. The twist: meditators slept worse, and most had meditated between 8 and 11 p.m. or even midnight to 3 a.m. Huberman explains that meditation is calm but attention-heavy, engaging the prefrontal cortex, unlike NSDR or yoga nidra, which lower attention. He plans to meditate about 15 minutes a day, early, ideally after morning sunlight and not past about 5 p.m.

Conclusion

Huberman's answer to why we remember anything is that repetition or a strong neurochemical signal, above all adrenaline, stamps particular neural patterns into memory. As he puts it, "adrenaline is the final common pathway" by which experiences are stamped into memory, and it matters less how you get there than when. The practical protocol follows: learn in a calm, focused state; spike adrenaline late in or right after the session with caffeine, cold water or hard exercise, if safe for you; avoid chronic stress, which blunts the effect; get good sleep, naps or NSDR later to consolidate; do regular zone 2 and load-bearing exercise to support the hippocampus through blood flow and osteocalcin; take deliberate mental snapshots of what you want to remember visually; and consider 13 to 15 minutes of daily meditation for at least eight weeks, done early in the day so it doesn't disrupt sleep.

This is a summary of "Understand & Improve Memory Using Science-Based Tools" from Huberman Lab, giving you the GIST so you can decide if you want the REST. Listen to the full episode.