Food and Memory: How Diet Shapes Cognition

Food and memory are linked through the gut-brain axis, and the evidence runs in both directions. Rodent studies show that early-life sugar or junk-food diets produce hippocampal memory deficits that persist after the animals return to healthy chow, while human studies tie habitual sugar and saturated fat intake to weaker performance on hippocampal memory tasks. The same vagus-nerve and hippocampal circuits that register a meal also appear to shape how much you eat next.

Food and memory: the core connection

Food and memory are linked because the hippocampus, the brain region that builds episodic memory, also receives meal-related signals from the gut and uses them to help regulate how much gets eaten. Dr. Scott Kanoski, a professor of human and evolutionary biology and neurobiology at the University of Southern California, has spent his career on this circuit, and his lab's work has partly defined it.

The direction of the effect runs both ways. Memory of a recent meal shapes intake at the next meal, and what you eat shapes how well the memory system works. Kanoski describes the hippocampus as a canary in the coal mine: it appears more sensitive to dietary insult than other memory systems that do not depend on it.

The pathways carrying that information are not a single wire. Stretch receptors in the stomach report distension. Chemoreceptors in the intestines report nutrients. Circulating hormones such as insulin and leptin reach the brain through the bloodstream. The vagus nerve carries a mixed electrical signal that gets relayed through the brain stem to the medial septum and then to the hippocampus, where acetylcholine appears to be the main currency.

Kanoski's argument is evolutionary. When mammals evolved, food was not reliably available, so encountering nutrients was an important event. Remembering what was eaten, what followed, and where it was found would have been worth encoding. On that view, the satiation systems and the memory systems are tuned to each other, and the hippocampus is where the two meet. Modern environments make the system less necessary for survival, but the wiring is still there.

The Huberman Lab episode featuring Kanoski was published in October 2026. The mechanistic work behind it spans roughly two decades of rodent research, some of it still in preprint form, and the human counterpart remains mostly observational.

How GLP-1 drugs suppress appetite

GLP-1 drugs suppress appetite by raising glucagon-like peptide 1 signaling far above physiological levels, and the naturally secreted hormone works largely through the vagus nerve. GLP-1, or glucagon-like peptide 1, is an incretin hormone released from the distal intestine around meals, and GLP-1 receptor agonist drugs such as semaglutide raise its signaling far above physiological levels to reduce appetite and body weight. The FDA-approved prescribing information for Wegovy (semaglutide) describes the label's appetite-reduction mechanism and indications.

In its natural form, GLP-1 is secreted from intestinal L cells and acts in a paracrine fashion on receptors on the terminals of the vagus nerve, which contacts the gastrointestinal tract directly. Paracrine means the hormone acts on receptors immediately next to the cells that released it rather than traveling through the bloodstream to a distant target. A second, largely independent source of GLP-1 sits in the brain stem, in the nucleus tractus solitarius. Kanoski notes that the peripheral and central systems both suppress appetite but appear to operate without a strong direct connection.

The vagus nerve is not an optional accessory here. Lesion the sensory vagus as it ascends from the gut, and native GLP-1 no longer reduces food intake in rodents. That work comes from Wolf Langhans and colleagues in Zurich, and it establishes the gut-to-brain relay as part of the satiation mechanism rather than an epiphenomenon. The same lab used short hairpin RNA interference to dial down GLP-1 receptor expression on the vagus nerve, and those animals ate larger meals.

GLP-1 also does more than suppress appetite. It helps regulate blood glucose and stimulates insulin secretion when blood glucose is high, which is why the first GLP-1 based drugs were approved for diabetes rather than obesity. The weight loss came later, and it is now the better-known application.

Whether the long-acting drugs cross the blood-brain barrier remains contested. First-generation GLP-1 drugs, studied from roughly 2005 to 2012, produced preclinical evidence of direct brain action. For newer agents such as semaglutide, Kanoski says the picture is less clear and the field is split. Some preclinical studies tag the peptides with a fluorescent molecule and find most of the signal clumping in the circumventricular organs, the regions where no blood-brain barrier separates the brain from circulation. Those studies are hard to read because fluorescent molecules fade and brain tissue has its own autofluorescent signal.

The better method is to radio-label the peptide with carbon 14 and measure directly whether it reaches the brain. Elizabeth Ria, Bill Banks and colleagues at the University of Washington are running those studies, and so far the early-generation drugs seem to enter the brain readily while the newer ones are less clear. Kanoski's summary is that the jury is still out on where these drugs act.

The uncertainty matters because GLP-1 receptors are not confined to appetite circuits. They are densely expressed across the brain, including the mesolimbic reward pathway (midbrain VTA, striatum, nucleus accumbens) and the hippocampus itself. That distribution is the basis for the preclinical interest in GLP-1 receptor agonists for substance use disorder and possibly memory disorders and dementia. The clinical data are only starting to emerge.

One practical signal Kanoski cannot explain yet is reduced "food noise." People on these drugs often report that the background pull toward food quiets down. He does not experience food noise himself and says the field has no answer for whether the effect comes from receptor binding in higher-order regions such as the hippocampus and prefrontal cortex or from the ordinary gut-to-brain satiation route.

Do artificial sweeteners change appetite?

Artificial and low-calorie sweeteners have mixed evidence for appetite and weight, and no randomized trial has settled the question. Kanoski's own position is that the literature is all over the place and that definitive answers do not exist for any of the three camps: sweeteners are harmful, sweeteners are inert, or sweeteners are suspect because they are humanmade.

One mechanism under study is the cephalic phase response. Sweet taste in the mouth triggers insulin release before nutrients reach the gut, preparing the body for a glucose load. When sweet taste arrives without the calories, the relationship between taste and metabolic consequence is disrupted, and preclinical work suggests that this mismatch can alter appetite and glucose regulation. Kanoski's phrasing for it is that you are tricking the wisdom of the body: sweet taste signals that calories are coming, and then they do not arrive.

Kanoski's lab ran one study that has become a reference point. Juvenile and adolescent rats received daily artificial sweetener at doses inside the FDA's acceptable daily intake, roughly equivalent to about six packets a day for a human, and they consumed it voluntarily rather than having it mixed into their drinking water. As adults, with the sweetener withdrawn, the animals showed memory impairments and a larger insulin spike when glucose was infused past the mouth. Sweet taste receptor expression on the tongue also changed. Three sweeteners were tested, including stevia, Ace-K and saccharin, and the effect did not depend on which one was used. Saccharin is no longer common in human beverages, but rodents like it, which is why so much of the older rodent literature is built on it.

The caveat is blunt: this is a rodent model, the effects were not enormous, and the authors did not identify the mechanism behind the memory finding. Kanoski does not make dietary recommendations from it and drinks Coke Zero himself. The honest summary is that the human evidence remains unresolved, mostly preclinical in origin, and unlikely to be settled by a large trial.

How meal memories regulate later eating

Memory of a recent eating occasion changes how much you eat at the next one, and this has been demonstrated in controlled human experiments. Suzanne Higgs and colleagues had people eat a pizza lunch, then primed one group to think about that lunch a few hours later. Primed participants ate fewer cookies at a subsequent taste test than people primed toward something else.

A second line of work comes from Jeff Brunstrom. Participants ate tomato soup from a bowl connected to a peristaltic pump that could add or remove soup without their knowledge, so the volume they believed they had eaten diverged from the volume they actually ate. Someone could see 300 ml in the bowl and consume 500 ml, or see 500 ml while the pump pulled soup out so the actual intake was 300 ml.

Immediately after eating, hunger ratings tracked actual intake. A couple of hours later, the ratings flipped and tracked what participants thought they had eaten. The study measured self-reported hunger, not later food intake, so the extension to a subsequent meal is an inference rather than a measured result. Still, the finding is a clean demonstration that the memory of a meal, not just the metabolic consequence, influences appetite regulation hours later.

Kanoski's lab studies the neural side of this. Using fiber photometry to record acetylcholine release into the hippocampus while animals are awake and eating, they see levels ramp up while the animal holds, chews and swallows food, dip when it pauses to look around, and end the meal higher than it started. The lab proposes that the elevated postprandial level lets the animal encode where the meal happened and what it contained so it can find the location again.

The same profile does not appear when animals consume stevia alone, which suggests that taste plus postoral nutrient sensing is required to drive the signal. The reverse experiment, infusing nutrients without taste, has not been run yet, and Kanoski expects it to show the same requirement for the combination.

Sugar, junk food and lasting memory deficits

High-sugar and junk-food diets produce hippocampal memory impairments in rodents that persist after the diet changes. Kanoski's lab first used an adult model with a Western-style diet of about 35 percent of calories from fat and 15 percent from simple sugar; hippocampal-dependent memory suffered within weeks, and the hippocampus proved more vulnerable than memory systems that do not depend on it. Some studies show effects after a week on the diet, though most run for a few months.

The lab then shifted to development. Juvenile and adolescent rats got sugar solution with a calorie share comparable to a soft drink, alongside healthy chow and water. As adults, they showed hippocampal-dependent memory deficits that persisted after the sugar was removed. A junk-food model using chocolate peanut butter cups, ruffled potato chips and a high-fat high-sugar chow produced the same pattern.

Effect sizes in these tests are substantial but not equivalent to a lesion. In a maze roughly the size of a table, with holes around the edge and visual-spatial cues in the room, control animals perform above chance at remembering where food was found. Diet-treated animals perform at chance. That is a large functional difference in the specific memory test, and it is worth noting that a separate intervention, such as exercise, might still rescue it.

The lab uses non-food memory tasks as well. In one, animals are placed in an arena and must remember which objects they have seen in which locations. An animal that has seen an object before but not in that spot will preferentially investigate it. Diet-treated animals lose that preference, which points to the same hippocampal system rather than to a food-specific effect.

Exercise is one candidate for reversal. Kanoski cites rodent evidence that wheel or treadmill running can reverse diet-induced memory deficits in part through changes in the brain. He avoids swim-based memory tasks because the physical demand confounds the memory measurement: the animal's exercise capacity would feed directly into its score.

Human data are thinner but point the same way. Dick Stevenson and colleagues studied healthy young adults and found associations between habitual sugar and saturated fat intake and performance on hippocampal-dependent memory tasks. That is correlational, and Kanoski notes that humans are poor at reporting what they actually eat, which limits the precision of any dietary study in people. Rodent work has near 100 percent accuracy on intake because the researchers control and measure it; phone-based food logging has improved human accuracy but has not fixed it.

Midlife diet and metabolic status are also associated with later dementia risk in prospective studies, including visceral adiposity as one marker. In transgenic rodent models of Alzheimer's disease, a Western diet pushes the amyloid and tau pathology and the behavioral symptoms earlier than they would otherwise appear. Kanoski's read is that the connection is more prospective than treatable: a healthy dietary pattern across life matters more than switching diets once symptoms start.

The gut microbiome and hippocampal function

The gut microbiome influences hippocampal function in rodent models, though no one can yet name which bacteria are beneficial. Kanoski entered the field by accident about ten years ago when his lab found that the early-life sugar model produced a completely different bacterial population. Principal component analysis showed the sugar-fed animals clustering far from controls, and the shift could come from the sugar itself or from the fact that those animals were getting less fiber and fewer nutrients from chow.

The lab then looked for individual populations. Parabacteroides was elevated roughly twofold in the sugar group and correlated negatively with memory performance. In a separate study, the researchers used antibiotics to deplete the bacterial population, then enriched Parabacteroides and found that the animals showed memory impairments despite never consuming sugar. The enrichment was about eightfold, larger than the sugar diet produced, and in many cases the deficits were more severe.

The preprint on heat-treated food points the other way. Heat treatment of otherwise healthy chow, which drives the Maillard reaction, produced an advanced glycation end product (AGE)-enriched diet. The Maillard reaction is the chemistry behind browning and flavor development in cooked food, and it generates AGEs as byproducts. AGEs circulate in normal physiology and have their own receptor, RAGE, and the system is generally associated with diabetes and aging. Animals on the AGE-enriched diet showed memory impairments, and one bacterial population, a lactic-acid producer called Lactobacillus, was eliminated by the heat treatment. Reintroducing it rescued the memory impairments.

Context matters here. Kanoski is not a microbiologist by training, and he is explicit that the same bacterial population can look protective in one study and harmful in another, depending on treatment and context. The abundance of one population also shifts the abundance of hundreds or thousands of others, which makes single-bacterium claims fragile.

The human microbiome field is young. PubMed hits for microbiome research went from roughly 7,000 about a decade ago to around 250,000, and the Human Microbiome Project only concluded around 2012. Before that point there was no clear picture of the genetic makeup of the human gut bacterial population, so pre-2010 this was barely a field.

At least one targeted microbiome therapy is established. Fecal microbiota transplantation is FDA-approved for recurrent Clostridioides difficile infection, and a single transplant corrects it in a high percentage of patients. That is the exception rather than the rule, and Kanoski says the field is in its early stages.

Vagus nerve, acetylcholine and intraceptive memory

The vagus nerve carries meal-related signals from the gut to the hippocampus, and disrupting it produces memory deficits in rodents and is associated with dementia risk in one historical human cohort. Kanoski's lab traces a three-tier pathway: gut to brain stem, brain stem to medial septum, and medial septum cholinergic projections to the hippocampus.

The signal is heterogeneous rather than unified. Stretch receptors in the stomach fire on distension at one timing, intestinal chemoreceptors at another, and the vagal neurons in the nodose ganglion express different receptors and respond to different stimuli. Kanoski describes it as nuanced, not a single message that food has arrived. His comparison is that the gut does not announce that you just ate lentil soup; the soup produces effects, and electrical currents carry the result upward. It is a song, not a gong.

Human evidence comes from an unusual historical source. In Taiwan in the 1980s and into the 1990s, truncal vagotomy was used to treat stomach ulcers, and long-term medical records allowed researchers to look at later dementia diagnoses. One small study found a positive correlation between vagotomy and later dementia risk. Kanoski's lab is pursuing the question in a Swedish national registry cohort.

Rodent work goes further. Kanoski's lab has a transgenic rat model of Alzheimer's disease that develops amyloid and tau pathology in the hippocampus with age, and implants vagus nerve stimulation devices sutured to the stomach. The device fires only when the animal eats, triggered by the natural expansion and contraction of the stomach. These devices have rescued memory impairments caused by an unhealthy diet, and the lab is testing whether they improve outcomes in the Alzheimer's model.

A separate line of human evidence shows what happens when the hippocampus is damaged. In densely amnesic patients, whose memory span runs roughly 10 to 40 seconds depending on the individual, an experimenter can feed a meal, leave, return with a second meal, and repeat; the patients do not remember eating. Asked to rate hunger on a scale of 0 to 10, they still report about a seven after three meals. The gut signal arrives, but the hippocampus is needed to interpret it. Kanoski calls this intraceptive agnosia: the sensory input is detected, but it no longer means what it is supposed to mean, so it cannot guide behavior.

There is a rodent parallel. A normal animal can learn that a period of energy restriction predicts an upcoming event such as a foot shock or a food pellet. Lesion the hippocampus and that learning disappears. Interoceptive awareness of hunger versus fullness also appears to be plastic and vulnerable to a high-fat high-sugar diet, which means the sense the system depends on can itself be degraded.

Reward, craving and how plasticity resets preferences

The mesolimbic dopamine pathway, running from the midbrain ventral tegmental area to the nucleus accumbens, is engaged by palatable food, but context decides how much. The same system responds to bland healthy food in a food-restricted animal, to water in a water-restricted animal, and to sodium in a sodium-depleted animal. Palatable foods simply tickle it harder.

That state dependence is why the system can recalibrate. People who cut back on sugar often find that foods they used to enjoy now taste too sweet, which shows the preference is not fixed. Kanoski's lab records dopamine release directly in awake animals using the same fiber photometry approach they use for acetylcholine, though they have not yet run the experiment that would measure how plastic the threshold is.

The extreme case of that plasticity is conditioned taste aversion. An animal that normally releases dopamine to sucrose, either at a cue that predicts it or at consumption, will stop releasing dopamine to sucrose after the taste is paired with malaise. The most potent natural reward can be converted into a signal the brain no longer values. Kanoski notes the darker version of the same logic: mid-century attempts to use aversion learning on human behavior, including electrical shock and conversion therapy, did not work and are now illegal.

Whether highly processed food qualifies as addictive is openly contested. Food shares a common currency with drugs of abuse in the brain, and by the DSM criteria for substance use disorder one can construct an argument for food addiction, but eating is not optional the way drug use is and the comparison has limits.

What the GLP-1 drugs mean for nutrition

The GLP-1 drugs reduce how much people eat without addressing what they eat, and most of the nutritional reassurance comes from trials that embedded diet and exercise guidance. Kanoski notes that most clinical trials run a year or two, embed nutritional and physical activity guidelines, and have not reported micronutrient deficiencies in that window.

Discontinuation is common, on the order of a majority of users within about two years, driven by side effects such as nausea and vomiting, by cost, and in elderly patients by lean mass loss. Because weight loss reduces both fat and lean mass, the loss of muscle is a recognized reason to stop. Kanoski is careful to say the drugs do not preferentially Stripe lean mass; any weight loss does that.

Kanoski's concern is not micronutrient deficiency in clinical trial populations but GLP-1 use in lean individuals who were never studied. There are no trials of that group and probably will not be. He also notes that B12 deficiency is historically a concern in modern populations when intake falls, and that a weight loss program without nutrient and activity components is not the optimal way to lose weight. If GLP-1 use becomes widespread and long-term, the combination of muscle loss without exercise plus micronutrient shortfalls is a plausible route to a population that is thin but metabolically and nutritionally worse off, which for hippocampal function could rival obesity as a problem.

Microdosing and non-weight-loss uses are pure anecdote at this point. Kanoski is not aware of any data on microdosing, placebo effects are strong, and randomized controlled trials are expensive enough that no one is likely to run them soon. Microdosing also works against the commercial incentive, since lower doses mean less product sold. The same logic applies to the healthy-weight users chasing cognitive or energy benefits: there is no trial evidence, only reports.

Food processing, the food supply and what changes the trend

Ultra-processed food supplies more than 60 percent of the calories many people in the United States consume, and the engineering that made that possible could be pointed the other way. Kanoski agrees the technology exists to fortify food with more micronutrients without sacrificing taste; what has not happened is a shift in demand or a top-down regulatory change. Even two pieces of fruit and one vegetable a day loaded with more micronutrients would be an improvement for people whose diets are otherwise poor, though it would not be sufficient on its own.

The NOVA classification system, which sorts foods into four processing categories from unprocessed to ultra-processed, is the standard research tool, but it is not mechanistic. It says a food is ultra-processed without saying which process mattered. Kanoski's lab tried to isolate a single variable by heat-treating otherwise healthy food, baking it for an hour at high temperature to drive the Maillard reaction, which is the reason cooked casserole tastes better than the raw ingredients. That single change was enough to produce memory impairments and wipe out a Lactobacillus population in the preprint model.

There is a structural reason the positive side of this literature is thin. The standard rodent diet is already healthy, with 10 to 15 percent of calories from fat and carbohydrate from complex sources, and those animals do not show memory impairments to rescue. It is easier to make an already healthy system fail than to make a healthy system measurably better. Kanoski argues that gap left room for diet books and confident claims that never had to answer to a literature.

The obesity picture explains the demand for drugs. Over 40 percent of US adults are obese, up from roughly 20 percent in the 1990s, and type 2 diabetes is no longer called adult-onset because children and teenagers now develop it. A class of drugs that produces meaningful weight loss could not stay dormant the way the market did after the fenfluramine-phentermine era, when cardiovascular harm took the drugs off the market and pharmaceutical companies lost interest in obesity for years. GLP-1 drugs revived the field around 2010 and beyond, and they remain the most effective pharmacological option available.

FAQ: food, memory and appetite

Does diet affect memory?

Yes, in both directions. Rodent studies show that high-sugar and junk-food diets produce hippocampal-dependent memory deficits that persist after the diet improves, and human studies associate habitual sugar and saturated fat intake with weaker performance on hippocampal memory tasks. The human evidence is correlational rather than causal. Prospective studies also link midlife diet and metabolic status to later dementia risk, which is an association, not a direct line.

How does the gut tell the brain about a meal?

The vagus nerve carries stretch and nutrient signals from the stomach and intestines to the brain stem, which relays through the medial septum to the hippocampus using acetylcholine. Circulating hormones such as insulin and leptin reach the hippocampus through the bloodstream as a second pathway, and the hippocampus expresses receptors for them much as the hypothalamus does. The vagal signal itself is not one message: stomach stretch receptors, intestinal chemoreceptors and different nodose ganglion neurons all contribute on different timings.

Do GLP-1 drugs improve memory?

The receptor is expressed throughout the brain, including the hippocampus, and there is preclinical evidence for potential applications in memory disorders, though clinical data are only starting to emerge. Kanoski is careful to note that the ability of newer GLP-1 drugs to cross into the brain remains contested. Carbon 14 labeling studies suggest early-generation drugs enter readily and newer ones are less clear.

Are artificial sweeteners bad for memory?

One rodent study from Kanoski's lab found memory impairments and larger insulin spikes in adult animals exposed to artificial sweeteners during development at doses within FDA acceptable daily intake, roughly six packets a day for a human. Three sweeteners were tested, including stevia and Ace-K, and the effect did not depend on which one was used. The study is preclinical and the mechanism is unknown, so the finding does not support a recommendation for humans.

Can you change what you crave?

Yes, and the plasticity runs in the direction of healthier preferences. People who reduce sugar intake often find previously enjoyable foods too sweet, and conditioned taste aversion experiments in rodents show that the dopamine response to a food can be eliminated entirely by pairing it with malaise. Kanoski's lab can record that dopamine release directly in awake animals, though the experiment measuring how far the threshold can move has not been done.

What does an amnesic patient tell us about hunger?

It separates the gut signal from the interpretation of the gut signal. Densely amnesic patients will eat a meal, forget it, and eat two more, yet still rate their hunger around seven out of ten because the gut feedback reaches the brain. What they cannot do is attach meaning to that signal, which Kanoski calls intraceptive agnosia. Rodent hippocampal lesions produce a similar failure to use energy restriction as a cue.

Can exercise reverse diet-induced memory deficits?

In rats, yes. Wheel and treadmill running have been shown to reverse the memory deficits caused by an unhealthy diet, in part through changes in the brain. The lab avoids swim-based memory tasks because the animal's physical capacity would confound the memory measure. Whether exercise helps in humans on the same timeline is not settled by the rodent work.

Why can't researchers just name the good gut bacteria?

The field is too young and the system too interconnected. PubMed went from roughly 7,000 microbiome papers about a decade ago to around 250,000 today, and the Human Microbiome Project only concluded around 2012. The same bacterial population can look harmful in one study and protective in another, and changing one population shifts hundreds of others. Fecal microbiota transplantation for recurrent C. difficile is the one FDA-approved targeted therapy, and it is the exception.

What is the single most reliable dietary signal for brain health?

Kanoski's own framing avoids precise recommendations, but the direction is consistent: excess sugar and excess saturated fat are the inputs with the clearest negative associations, exercise and a healthy dietary pattern across life are the inputs with the clearest positive ones, and omega-3s, polyphenols and adequate protein are the established nutritional players other researchers have documented. A weight loss program with nutrient and activity components built in beats a drug alone.

Source video

The same gut-to-brain circuit that decides how much you eat is also the one that files away where the meal came from, and Kanoski's lab has spent two decades mapping it. If you have recorded conversations, explainer videos, or interviews like this one and the knowledge in them is sitting unwatched, Skala Blog turns a YouTube link into a written article you can publish and share. Paste the URL, let it transcribe the video, and get a draft you can edit.

For more tools and resources, visit crazystack.com.br. Thanks to Dev Doido do canal do youtube for pointing viewers toward the underlying research.