Who Decided Three Meals a Day?

Someone on Reddit asked a simple question: "Why did the 3 meals a day pattern become standard? Is there evidence it's optimal, or is it influenced by governments and corporations?"

It's a better question than it sounds. The answer touches on factory schedules, a PR campaign for bacon, a Nobel Prize in cell biology, and the fact that your great-grandparents in Sicily and your great-grandparents in Osaka were eating completely different foods and somehow both doing fine. Until we decided to standardize everything.

My guess is that different cultures and parts of the world had different eating patterns. Men and women had different eating patterns. Food scarcity also likely necessitated that those eating patterns were not very consistent in a lot of places. The fact that we have common guidelines for RDVs, macro balances, and calories for all different backgrounds (except gender) is crazy. Paleo, Mediterranean, Keto, Intermittent Fasting, and other "fad" diets all probably have some truth to them, and work for some people better than others. So I don't think there is much evidence that "3 meals a day" is optimal, or that grazing is optimal, or that IF is optimal. Different strategies are optimal for different people, and probably vary throughout the year (and years) even. The best we can probably do for now is try to listen and look at what our bodies are telling us. Bloating, inflammation, acne, frequent colds, allergies, bowel issues, weight gain (or rarely weight loss) are all signs we're doing something wrong.

That was my initial answer. This post is what happened when I started pulling on the threads behind it: the history of how we standardized eating, what the evidence says about the diets claiming to fix it, and why the government's one-size-fits-all nutrition recommendations might not fit you at all.

How humans actually ate

For most of human history, there was no such thing as a structured meal schedule. People ate when food was available, which was unpredictable and varied enormously by geography and season.

Modern hunter-gatherer groups (the Hadza of Tanzania, the !Kung San of the Kalahari) eat opportunistically: snacking during the day while foraging, then sharing a larger communal meal in the evening when hunters return (Marlowe, Evolutionary Anthropology, 2005). There's no schedule. There's no breakfast. There's "I found something edible" and "the hunting party came back."

The ancient Romans typically ate two meals. Their main meal, the cena, happened mid-to-late afternoon (around 2-3 PM). A light morning bite and a midday snack existed, but they were considered minor. Pliny the Elder reportedly thought eating more than one real meal a day was excessive.

Medieval Europeans followed a similar pattern: a late-morning meal around 10-11 AM and an evening meal around 5-6 PM. Breakfast was considered a sign of weakness or gluttony by the medieval Church (Henisch, Fast and Feast, 1976). Monks and laborers who ate in the morning were breaking a religious fast, and the word itself (break-fast) carries that origin.

Breakfast became socially acceptable gradually during the 1500s-1700s, helped along by the Protestant Reformation (which weakened Catholic fasting norms) and the arrival of coffee, tea, and chocolate as morning beverages. But the idea of a structured, three-meal day with fixed times? That required something entirely different.

Industrialization and Fixed Meal Schedules

Before industrialization, agricultural workers ate when the work allowed it, usually a large midday "dinner" and a lighter evening "supper." Time was task-oriented, not clock-oriented.

The British Industrial Revolution (1760s-1840s) changed that. E.P. Thompson's landmark essay "Time, Work-Discipline, and Industrial Capitalism" (Past & Present, 1967) documents how factory schedules imposed rigid clock-time on workers for the first time. You needed fuel before the shift, a break in the middle, and a meal when you got home. Three meals wasn't a nutritional decision. It was a labor logistics decision.

Gas lighting (widespread by the 1830s-1840s) and later electric lighting (1880s onward) pushed the evening meal later, from mid-afternoon toward 6-8 PM. That created space for a distinct "lunch" in the middle of the day. By the 1850s-1880s, breakfast-lunch-dinner was the established norm in both the US and Britain (Schivelbusch, Disenchanted Night, 1988).

How Cereal Companies Shaped Breakfast

Once the three-meal pattern was established as a social norm, there was a commercial opportunity to define what people should eat at each one. Several industries took full advantage.

Cereal. In 1894, the Kellogg brothers accidentally invented flaked cereal (initially wheat) at the Battle Creek Sanitarium, a Seventh-day Adventist health facility; corn flakes followed in 1898. By the 1920s-1950s, Kellogg's and General Mills were spending heavily on radio and television advertising to position cereal as the default American breakfast. The phrase "breakfast is the most important meal of the day" originated as cereal marketing copy, not nutritional science (Carroll, Three Squares, 2013).

Bacon and eggs. In the 1920s, Edward Bernays (the "father of public relations" and Sigmund Freud's nephew) was hired by the Beech-Nut Packing Company to increase bacon sales. He got a physician to write to 5,000 doctors asking whether a "heavier breakfast" was better for health. The predictable affirmative responses were then publicized as medical consensus. This is one of the earliest examples of manufacturing nutritional authority for commercial purposes (Tye, The Father of Spin, 1998).

The Food Pyramid. The USDA published the Food Guide Pyramid in 1992, recommending 6-11 servings of grains at the base. Luise Light, the USDA nutritionist who led the original design team, later stated publicly that her team recommended 2-3 whole grain servings, but the grain industry lobbied to increase the number. She documented this in her 2006 book What to Eat. The degree of industry influence on the final pyramid has been alleged by multiple sources, though the specifics of what was changed and why remain debated. Light's account is primarily her own recollection.

None of this means breakfast is bad or grains are poison. But it does mean that much of what we consider "normal" eating was shaped by commercial interests, not by evidence about what human bodies actually need.

The obesity timeline

If three meals a day were the main problem, you'd expect obesity to have spiked in the 1880s when the pattern solidified. But obesity rates stayed relatively stable for nearly a century after that. Something else was going on.

The pattern is striking: Americans were told to eat less fat, the food industry replaced fat with sugar to keep things palatable, corn subsidies made sweeteners cheap, and caloric intake climbed by roughly 500 calories per day. Snacking between meals (not the meals themselves) drove much of the increase (Popkin et al.).

It's hard to point to a single cause because government policy, corporate incentives, food engineering, and a cultural shift toward constant eating all happened at roughly the same time. That context matters for understanding why so many people are now looking to restrictive diets for answers.

So that's the context for the original question. If three meals a day isn't biologically optimal, is any specific eating pattern actually better? There are six popular diets that get brought up constantly in these discussions, and they have varying levels of evidence behind them.

Analysis of Popular Diets

Every popular diet is, at some level, a reaction to the food environment described above. They each point to a different culprit (carbs, meal timing, grains, animal products, processed food) and build a framework around avoiding it. The evidence for each varies a lot.

The Mediterranean Diet

The Mediterranean diet has the strongest evidence base of any dietary pattern, and it's probably not particularly close in terms of the quality and quantity of supporting research.

The PREDIMED trial (Estruch et al., originally NEJM 2013, retracted and republished 2018 after a randomization issue at one site was corrected; the re-analysis confirmed the original findings) enrolled 7,447 people and followed them for about 4.8 years. A Mediterranean diet supplemented with extra-virgin olive oil or nuts reduced major cardiovascular events by roughly 30% compared to a low-fat control diet.

The Lyon Diet Heart Study (de Lorgeril et al., Circulation, 1999) found even larger effects: a Mediterranean-style diet in post-heart-attack patients reduced cardiac death and nonfatal heart attacks by 50-70% over 4 years. The effect was so large the trial was stopped early.

Why does it work? Probably multiple mechanisms: polyphenols from olive oil (oleocanthal has anti-inflammatory properties similar to ibuprofen; Beauchamp et al., Nature, 2005), omega-3 from fish, high fiber from legumes and whole grains, and the overall pattern of displacing processed food. It's likely the combination, not any single ingredient.

It also happens to be the most sustainable. The Mediterranean diet doesn't eliminate entire food groups, doesn't require special products, and has millennia of cultural precedent. The "best diet" is the one a person can actually maintain, and by that measure, Mediterranean wins again.

Intermittent fasting: the evidence

Intermittent fasting (IF) comes in several forms: 16:8 (eat within an 8-hour window), 5:2 (eat normally 5 days, restrict heavily for 2), and OMAD (one meal a day). The appeal is simplicity: instead of counting calories, you just watch the clock.

The evidence is more mixed than social media suggests.

The landmark review by de Cabo and Mattson (NEJM, 2019) concluded that IF produces comparable weight loss to continuous calorie restriction, with possible additional metabolic benefits (improved insulin sensitivity, reduced inflammation). But they noted most human studies were small and short.

Since then, larger trials have been less impressive. Lowe et al. (JAMA Internal Medicine, 2020) ran a 16:8 trial with 116 people for 12 weeks and found no significant difference in weight loss between time-restricted eating and three meals a day. The fasting group also lost more lean mass, which was concerning. Liu et al. (NEJM, 2022) ran an even larger trial (139 people, 12 months) and found no difference in weight loss, body fat, or metabolic markers between calorie-restricted IF and calorie restriction alone.

There is an interesting wrinkle: early time-restricted eating (eating earlier in the day, finishing by mid-afternoon) may have metabolic benefits independent of calorie intake. Sutton et al. (Cell Metabolism, 2018) found improved insulin sensitivity and blood pressure with an early eating window, likely because it aligns with circadian rhythm. But this was a small crossover study (n=8) and hasn't been replicated at scale.

The Paleo Diet

The paleo diet eliminates grains, legumes, dairy, and processed food, arguing that our genome hasn't adapted to agricultural foods. There are a few problems with the premise.

First, there was no single "paleo" diet. Paleolithic humans ate dramatically different foods depending on geography and season. Cordain et al. (American Journal of Clinical Nutrition, 2000) estimated hunter-gatherers derived roughly 45-65% of calories from animal sources, but the range was enormous. Second, significant genetic adaptation has occurred in the last 10,000 years. Lactase persistence and amylase gene variation are clear examples (more on this later). Third, archaeological evidence shows starch granules on Neanderthal teeth, suggesting grain consumption predates agriculture (Henry et al., PNAS, 2011).

That said, the clinical evidence is modestly positive. A meta-analysis by Manheimer et al. (American Journal of Clinical Nutrition, 2015) pooled 4 randomized controlled trials (total of 159 people) and found paleo improved waist circumference, triglycerides, blood pressure, and fasting blood sugar more than control diets. But the studies were small and short (2 weeks to 3 months).

One concern: Genoni et al. (European Journal of Nutrition, 2020) found that long-term paleo dieters had significantly higher TMAO (trimethylamine N-oxide) levels, a biomarker associated with cardiovascular risk, potentially from reduced gut microbiome diversity when whole grains and legumes are eliminated.

Paleo gets the broad strokes right: eat real food, avoid processed junk. But the specific prohibitions (no legumes, no whole grains, no dairy) aren't well-supported by the evidence. Lentils and yogurt aren't the enemy.

The Ketogenic Diet

The ketogenic diet restricts carbohydrates below roughly 20-50 grams per day, forcing the body into ketosis (burning fat and producing ketone bodies for fuel instead of glucose). It was originally developed in the 1920s at Johns Hopkins for children with drug-resistant epilepsy, where it remains well-supported (Cochrane review, Martin-McGill et al., 2020: about 40-50% of children achieve a 50% or greater seizure reduction).

For weight loss, keto produces faster initial results, but much of the early loss is water (glycogen depletion pulls water with it). A meta-analysis by Bueno et al. (British Journal of Nutrition, 2013) found keto produced about 0.9 kg more weight loss than low-fat diets at 12+ months. Statistically significant, clinically modest.

The NIH metabolic ward study (Hall et al., Nature Medicine, 2021) is worth noting. In a tightly controlled crossover design, a low-fat plant-based diet and a low-carb animal-based diet produced similar fat loss when participants ate freely. The low-fat group ate about 550-700 fewer calories per day, but the low-carb group lost more weight initially (mostly water from glycogen depletion). Over the 2-week crossover periods, fat loss was similar between groups.

The LDL question. Keto raises LDL cholesterol in a substantial subset of people. The KETO trial (Norwitz et al., JACC Advances, 2024) looked at "lean mass hyper-responders" (lean, metabolically healthy people on keto who develop very high LDL, often above 200 mg/dL). The trial found no significant difference in coronary plaque burden compared to matched controls with much lower LDL, but it was cross-sectional (a snapshot, not tracking change over time) and small. The authors themselves noted it could not rule out long-term cardiovascular risk. Mendelian randomization studies consistently show LDL is causally related to atherosclerosis regardless of mechanism of elevation. This remains an unresolved concern.

Adherence is the real problem. Kirkpatrick et al. (Journal of Clinical Lipidology, 2019) found that carbohydrate intake tends to drift upward, with most participants no longer in ketosis by 12 months. Dropout rates in keto trials commonly range from 30-50%. If someone can sustain it and their lipids look fine, it can work. For most people, it's a diet they try for 3 months.

The Carnivore Diet

The carnivore diet (exclusively animal products: meat, organs, eggs, sometimes dairy) is the most restrictive option on this list and has the thinnest evidence base.

There are essentially no published randomized controlled trials. The largest study is Lennerz et al. (Current Developments in Nutrition, 2021): a self-reported survey of about 2,000 carnivore dieters recruited from online communities. Participants reported high satisfaction and improvements in various conditions. This is extremely low-quality evidence: self-selected, no control group, no biochemical verification.

The vitamin C question. Fresh meat does contain small amounts of vitamin C (roughly 10-30 mg/kg in muscle meat, more in organs). Clinical scurvy requires below 10 mg/day for weeks, so it's plausible that sufficient meat intake avoids frank deficiency. Some advocates argue that lower carb intake reduces vitamin C requirements because glucose and vitamin C compete for the same cellular transporters (GLUT). This is biologically plausible but unproven in clinical trials.

The fiber question. Zero fiber intake on carnivore profoundly reduces gut microbial diversity (Sonnenburg & Sonnenburg, Nature, 2016). Some traditional low-fiber populations (like the Inuit) appeared to have low rates of certain diseases, but this evidence is heavily confounded by short lifespans and limited epidemiological data.

Carnivore's appeal is partly as an elimination diet: by removing everything except meat, people with undiagnosed food sensitivities may feel dramatically better. That's a real phenomenon. But it doesn't mean the optimal long-term diet is exclusively steak.

Prolonged fasting and autophagy

Prolonged fasting (24-72+ hours) is the most extreme intervention on this list, and its claims center on autophagy: the cellular recycling process where damaged proteins and organelles are broken down and reused.

Autophagy is real biology. Yoshinori Ohsumi won the 2016 Nobel Prize in Physiology or Medicine for elucidating its molecular mechanisms (primarily in yeast). Animal studies consistently show that fasting upregulates autophagy. But here's the problem: we cannot reliably measure autophagy in living humans. There is no validated blood test or imaging method. Claims that "autophagy peaks at 16 hours" or "autophagy kicks in at 24 hours" are extrapolated from animal models and cell cultures, not confirmed in human tissue.

Valter Longo's fasting-mimicking diet (ProLon, a 5-day protocol at 750-1100 kcal/day) has more human data than complete fasting. Wei et al. (Science Translational Medicine, 2017, n=100, 3 monthly cycles) showed reductions in body weight, blood pressure, fasting glucose, IGF-1, and C-reactive protein. Effects were most pronounced in people who were already at risk. The study was reasonably designed but industry-funded (Longo has a financial interest in ProLon).

Real risks. Muscle loss accelerates beyond about 24 hours of complete fasting as the body breaks down protein for gluconeogenesis (making glucose). For lean individuals, the risk-to-benefit ratio worsens considerably. Refeeding syndrome (a dangerous shift of electrolytes when food is reintroduced after extended fasting) is a genuine medical risk, typically after 5+ days, but possible earlier in malnourished individuals (Mehanna et al., BMJ, 2008).

The underlying biology here is genuinely interesting, but the evidence that deliberately inducing multi-day fasts is net beneficial for already-healthy humans is essentially absent.

Common Ground Across Diets

The adherence data supports this interpretation. A 2014 meta-analysis by Johnston et al. (JAMA) compared named diets head-to-head and found minimal differences in weight loss at 12 months. The dominant predictor of success was whether people could actually stick with the diet, not which one they chose.

If the specific diet framework matters less than adherence and food quality, there's a natural follow-up question: has the food itself changed? There's actually a fair amount of research on this, and the answer is more complicated than you might expect.

Nutrient Density in Modern vs. Historical Crops

Even if someone eats "real food" and avoids the processed stuff, there's a legitimate question about whether the real food itself has changed.

Davis et al. (Journal of the American College of Nutrition, 2004) compared USDA nutrient composition data for 43 garden crops between 1950 and 1999. They found statistically reliable declines in protein (-6%), calcium (-16%), phosphorus (-9%), iron (-15%), riboflavin (-38%), and vitamin C (-20%). These are median declines across all 43 crops. Some individual crops showed larger drops; some showed none.

The most likely mechanism isn't worn-out soil (though that plays a role regionally). It's selective breeding. Modern crop varieties are optimized for yield, pest resistance, appearance, and shelf life. A tomato that grows faster and bigger distributes the same minerals from the same root system across more biomass. Davis called this the "genetic dilution effect" and published further work on it (HortScience, 2009). Direct experiments comparing old and new wheat cultivars grown side by side in identical soil show nutrient differences that track with yield (Fan et al., Journal of Trace Elements in Medicine and Biology, 2008).

There's a newer mechanism that will make this worse: atmospheric CO2. Myers et al. (Nature, 2014) grew crops at elevated CO2 levels (~550 ppm, projected mid-century) and found significant reductions in zinc (-5-10%), iron (-5-8%), and protein content in wheat, rice, field peas, and soybeans. Higher CO2 accelerates plant growth but dilutes mineral uptake. This is the same dilution effect as selective breeding, now driven by the atmosphere.

What about meat? Grass-fed beef has a more favorable fatty acid profile than grain-fed: roughly 2-5x more omega-3 and 2-3x more conjugated linoleic acid (CLA) (Daley et al., Nutrition Journal, 2010). But the absolute numbers are small. Even grass-fed beef has about 80 mg of omega-3 per 100g serving, compared to 1,000-2,000 mg in salmon. The difference between grass-fed and grain-fed is real but clinically modest unless beef is the primary protein source.

The pesticide question

The standard reassurance goes like this: the USDA Pesticide Data Program tests thousands of food samples every year and consistently finds residue levels well below EPA tolerance levels. Often 100-1,000x below. Case closed.

But that argument assumes the EPA tolerance levels themselves are the right benchmark. Where did those numbers come from?

How tolerance levels are set

Under FIFRA (the Federal Insecticide, Fungicide, and Rodenticide Act), pesticide manufacturers submit their own safety studies to the EPA. The manufacturer funds the toxicology research, conducts it, and submits it for review. The EPA evaluates those studies but generally does not run independent testing. The raw data is typically claimed as confidential business information, limiting outside scrutiny.

The EPA identifies a No Observed Adverse Effect Level (NOAEL) from animal studies, then applies safety factors: 10x for animal-to-human extrapolation, 10x for human variability. The Food Quality Protection Act of 1996 added an additional 10x safety factor for children, but the EPA has frequently reduced this to 3x or 1x based on available data, which has been a point of contention among toxicologists and public health researchers.

Many current tolerances trace their origins to registrations from the 1950s through 1980s. The FQPA mandated reassessment of all existing tolerances by 2006. The EPA completed high-priority classes (organophosphates, carbamates) but the broader reassessment has been criticized as incomplete by the GAO and EPA's own Office of Inspector General, and many registration reviews remain behind schedule.

How US limits compare internationally

The US and EU use fundamentally different regulatory philosophies. The US uses a risk-based framework: a pesticide is registered unless the EPA determines it causes "unreasonable adverse effects," and economic benefits can be weighed against risks. The EU uses the precautionary principle: if there is scientific uncertainty about safety, the default is to restrict or ban. For certain hazard classes (carcinogens, endocrine disruptors), the EU will not approve a substance regardless of dose.

The practical result: several pesticides widely used in the US are banned in the EU.

Specific limit comparisons tell a similar story. The EU drinking water standard for any individual pesticide is 0.1 ppb. The US EPA maximum contaminant level for atrazine alone is 3 ppb (30x higher). Japan applies a default limit of 0.01 ppm for any pesticide without a specific tolerance, far more restrictive than the US approach.

What the long-term evidence shows

Children's neurodevelopment has the strongest evidence. Three independent birth cohort studies (CHAMACOS at UC Berkeley, the Columbia Center for Children's Environmental Health, and the Mt. Sinai study) all found associations between prenatal organophosphate exposure and cognitive deficits in children. The CHAMACOS study found approximately a 7-point IQ deficit comparing the highest vs. lowest exposure quintiles (Eskenazi et al., Environmental Health Perspectives, 2007; additional CHAMACOS findings in Bouchard et al., 2011). The Columbia study found brain structural changes visible on MRI (Rauh et al., PNAS, 2012). These consistent results across independent cohorts studying different populations were a major factor in the eventual chlorpyrifos ban.

Cancer. The NutriNet-Santé cohort (Baudry et al., JAMA Internal Medicine, 2018) followed ~69,000 French adults and found 25% lower overall cancer risk among those eating the most organic food. But this is observational data with serious healthy-user bias: people who buy organic also tend to exercise more, smoke less, and have higher incomes. A UK Million Women Study (Bradbury et al., British Journal of Cancer, 2014) found no association between organic food and overall cancer risk. The honest answer is that we don't know whether pesticide residues at dietary levels cause cancer in humans.

Glyphosate is the highest-profile example of the uncertainty. IARC (the WHO's cancer agency) classified it as "probably carcinogenic" in 2015. The EPA concluded it's "not likely to be carcinogenic" in 2017. They looked at overlapping evidence and reached opposite conclusions, partly because IARC evaluates hazard (can it cause cancer at any dose?) while the EPA evaluates risk (does it cause cancer at real-world doses?), and partly because the EPA weighted unpublished industry studies that IARC didn't consider. A 2019 meta-analysis (Zhang et al., Mutation Research) found a 41% increased risk of non-Hodgkin lymphoma among high-exposure users. Internal Monsanto documents released through litigation (the "Monsanto Papers") revealed ghostwriting of published safety studies and communications with EPA officials about suppressing outside reviews. Bayer, which acquired Monsanto for $63 billion, set aside $10-16 billion to settle Roundup lawsuits.

Untested Interactions and Cumulative Exposure

Perhaps the most important structural problem: the EPA assesses pesticides individually. A person eating a normal diet consumes residues of multiple different pesticides in a single day. The EPA has completed cumulative risk assessments for a few chemical families that share a mechanism (organophosphates, carbamates, triazines), but does not assess the combined effect of pesticides with different mechanisms consumed simultaneously. The National Academy of Sciences flagged this gap in 2012 ("Exposure Science in the 21st Century"), and it remains unresolved.

Traditional toxicology also assumes linear dose-response: more chemical, more effect. But the Endocrine Society (2009, 2015 scientific statements) has documented that some chemicals, particularly endocrine disruptors, show non-monotonic dose-response, meaning effects at very low doses that aren't predicted by high-dose animal studies. If this applies to dietary pesticide exposure, the standard safety-factor approach may not be protective.

Pesticides are one dimension of food quality, but macronutrient composition matters too. Fiber is a good case study: it's one of the largest gaps between ancestral diets and modern ones, and the evidence for its health effects is stronger than most people realize.

Fiber Intake and Health Outcomes

The official recommendation for fiber is 25 grams per day for women and 38 grams for men (IOM, 2002). Average American intake is about 15-16 grams. Only about 5% of Americans hit the target. That gap is worth understanding.

The IOM set the fiber recommendation as an Adequate Intake (AI), not a Recommended Dietary Allowance (RDA). That distinction matters technically (an AI means the evidence wasn't sufficient to establish an Estimated Average Requirement, the prerequisite for an RDA), but it doesn't mean the evidence is weak. The AI was based on consistent epidemiological data showing that fiber intakes of roughly 14 grams per 1,000 calories were associated with reduced coronary heart disease risk. And the evidence has gotten stronger since 2002, not weaker.

Heart disease: Multiple large cohort studies consistently link higher fiber intake to reduced cardiovascular risk. Pereira et al. (Archives of Internal Medicine, 2004) found roughly 10-30% reduced coronary events per 10g/day increase. RCTs confirm that soluble fiber from oats and psyllium reliably lowers LDL by about 5-10% (FDA allows a health claim for this). Large randomized trials with hard cardiovascular endpoints for fiber specifically don't exist, but the consistency across dozens of cohorts, combined with well-understood mechanisms (bile acid binding, SCFA production), makes this one of the more solid areas in nutrition.

Blood sugar and type 2 diabetes: This is where fiber has some of its strongest evidence. Soluble fiber slows gastric emptying and glucose absorption, reducing postprandial blood sugar spikes. Multiple meta-analyses link higher fiber intake to reduced type 2 diabetes risk (roughly 15-30% reduction in prospective cohorts), and RCTs confirm that supplemental soluble fiber improves glycemic control in diabetic patients. This is mechanistically well-understood and clinically reproducible.

Cancer: More mixed. Denis Burkitt's famous 1970s hypothesis (that high-fiber African diets explained low colon cancer rates) didn't hold up cleanly in two major RCTs. The Polyp Prevention Trial (Schatzkin et al., NEJM, 2000) randomized about 2,000 people to a high-fiber diet and found no significant difference in adenoma recurrence after 4 years. The Phoenix Wheat Bran Fiber Trial (Alberts et al., NEJM, 2000) found the same. But both trials used insoluble wheat bran fiber specifically, over a relatively short period. More recent pooled analyses of prospective cohorts (Aune et al., BMJ, 2011) do show a significant association between higher fiber intake and reduced colorectal cancer risk, particularly from cereal and whole grain sources. The picture is more nuanced than "fiber prevents cancer" or "fiber doesn't prevent cancer."

Gut health: This is the most active research area. Gut bacteria ferment fiber into short-chain fatty acids (butyrate, propionate, acetate), and butyrate is the primary fuel for colon cells with anti-inflammatory properties. The Sonnenburg lab at Stanford showed that low-fiber diets reduce microbiome diversity in mice, with some species lost irreversibly across generations (Sonnenburg et al., Nature, 2016). A follow-up Stanford study comparing high-fiber vs. high-fermented-food diets found that the fermented food group showed greater microbiome diversity increases (Wastyk et al., Cell, 2021), suggesting diversity and fiber are related but not the same thing.

Ancestral context. Eaton and Konner's estimate of ancestral fiber intake (NEJM, 1985, later revised) ranges from 45-100 grams per day, primarily from wild tubers, roots, and fibrous plants. The Hadza of Tanzania reportedly consume around 100-150 grams daily, mostly from tubers and baobab fruit. By this measure, 38 grams isn't an aggressive target. It's a fraction of what humans ate for most of our history. The variation across populations was enormous (the traditional Inuit and Maasai ate very little fiber, though confounders like high physical activity, no processed food, and shorter lifespans make direct comparisons tricky). But most ancestral diets were far higher in fiber than modern ones.

One distinction the recommendation could do a better job with: soluble fiber (oats, beans, psyllium) and insoluble fiber (wheat bran) have different evidence profiles. The cancer trials that came back negative used wheat bran. The cholesterol and blood sugar evidence is strongest for soluble fiber. They're doing different things physiologically, and a single number doesn't capture that.

Hunger, Satiety, and Circadian Cues

There's a popular idea that cravings reflect nutritional deficiencies. Craving chocolate means you're low on magnesium. Craving red meat means you need iron. It's an appealing concept, but the evidence doesn't support it for most food cravings.

Reviews of the craving literature consistently find that food cravings are better predicted by stress, habit, and sensory exposure than by micronutrient status (Boswell & Kober, Annals of the New York Academy of Sciences, 2016). If the body were truly nutritionally wise about cravings, we'd crave leafy greens and sardines, not chocolate and chips. The one clear exception is pica (craving non-food substances like ice, clay, or starch), which is clinically associated with iron deficiency (Young, Craving Earth, Columbia University Press, 2011).

The protein leverage hypothesis is a more interesting version of the "body knows" argument. David Raubenheimer and Stephen Simpson (Obesity Reviews, 2005) proposed that humans have a strong appetite specifically for protein and will overconsume total calories to reach a protein target if available food is protein-dilute. Animal data is strong (demonstrated across insects, mice, and other species). In humans, Gosby et al. (PLoS ONE, 2011) found that participants on a 10% protein diet consumed about 12% more total energy than those on 15% or 25% protein diets over 4 days. The implication: if modern processed food is lower in protein density than the whole foods it replaced, people may be eating more total food to hit the same protein target. This is increasingly taken seriously in nutrition science (published in Nature and Cell Metabolism), though the full extrapolation to the obesity epidemic remains a hypothesis.

Energy levels are more informative than cravings. Persistent fatigue has well-established associations with iron deficiency (even without anemia), vitamin D deficiency, B12 deficiency, and magnesium insufficiency. The postprandial spike-and-crash pattern (feeling tired after meals) correlates with glycemic variability, and the PREDICT studies (Berry et al., Nature Medicine, 2020) showed that glycemic responses to identical meals vary enormously between individuals, driven by gut microbiome, genetics, sleep, and prior exercise.

The practical takeaway: specific food cravings are mostly habit and hedonic reward. But chronic fatigue, brain fog, or energy crashes after meals are worth investigating through bloodwork and dietary tracking. And if the protein leverage hypothesis is correct, making sure meals have adequate protein may reduce total calorie intake without trying.

Body signals are useful, but they're not precise enough to catch specific deficiencies. For that, the data is more informative than the symptoms.

Common Nutrient Deficiencies

Before debating keto vs. paleo vs. Mediterranean, it's worth asking a simpler question: are you getting the basics?

Sources: NHANES data, Forrest & Stuhldreher 2011, King et al. 2005. EAR = Estimated Average Requirement, AI = Adequate Intake.

The thing that jumps out is how widespread these gaps are. The majority of Americans are below recommended intakes for at least several of these nutrients, regardless of which "diet" they follow. Someone could be perfectly adherent to keto or paleo or Mediterranean and still be deficient in magnesium or vitamin D.

This is why tracking micronutrients (not just calories and macros) matters. Most calorie-counting apps track 4 macros and call it a day. The gaps above are in vitamins, minerals, and fatty acids that those apps don't show.

But even tracking against standard recommendations has a problem: the recommendations themselves vary depending on who wrote them and who they were written for.

Limitations of Universal Dietary Guidelines

The Recommended Dietary Allowances (RDAs) were first published in 1941 by the National Research Council, driven by wartime concerns about military nutrition. They've been updated periodically since, and were expanded into the Dietary Reference Intakes (DRIs) starting in 1997. The RDA for any given nutrient is set to meet the needs of 97-98% of healthy individuals in a given age and sex group.

Two things to understand about how these numbers are set. First, the historical threshold was preventing clinical deficiency disease (scurvy, rickets, pellagra), not optimizing long-term health. The vitamin D RDA of 600 IU for adults under 70 targets bone health, not the higher serum levels that some researchers associate with reduced cancer and cardiovascular risk. Second, the recommendations assume a generic human. They don't account for ancestry, skin color, genetic variants, or the traditional diet your lineage adapted to.

The gap between countries is revealing:

Iron is the most dramatic example. WHO recommends 29.4 mg/day for women, assuming a plant-based diet with low absorption. Japan recommends 10.5 mg, assuming a mixed diet with good absorption. Same nutrient, same goal, nearly 3x difference. The "right" number depends on what you eat alongside it.

These aren't fringe disagreements. These are the world's top nutrition authorities looking at the same evidence and reaching different numbers because their populations eat differently, have different genetics, and live at different latitudes.

The RDA discrepancies between countries hint at a deeper issue: human populations have spent thousands of years adapting to different diets, and those adaptations are written into our genes.

Genetic and Regional Variation in Ancestral Diets

Beyond the RDA discrepancies, there are well-documented genetic variations in how different populations process nutrients. These aren't small differences.

These aren't obscure genetic curiosities. They're mainstream, well-replicated findings. A person of West African descent living in Minnesota has fundamentally different vitamin D needs than a person of Scandinavian descent in the same city. A person of East Asian descent processes alcohol differently than a person of European descent. Someone with 5 copies of the amylase gene handles rice differently than someone with 12 copies.

The personalized nutrition field is growing. The PREDICT studies (Berry et al., Nature Medicine, 2020) showed that even identical twins have dramatically different glycemic and lipemic responses to the same meals, driven by gut microbiome, genetics, sleep, and exercise patterns. José Ordovás at Tufts has argued since the early 2000s that population-level dietary recommendations are "a blunt instrument."

The 2020 Dietary Guidelines Advisory Committee acknowledged these genetic variations but said the evidence was insufficient for genotype-specific recommendations. That's fair for government policy. But for an individual trying to optimize their own health, the data suggests that paying attention to your own body's responses (bloodwork, energy, digestion) is more useful than following a generic recommendation chart.

Which brings us back to the Reddit question: is three meals a day optimal? After going through all of this, I think the answer is basically what I said initially: it depends on who you are, where your ancestors came from, what your body is telling you, and what you're actually eating. The evidence really doesn't support a single optimal pattern for everyone, and the genetic variation data suggests it probably never will.

Summary

1. Three meals a day is a labor convention, not a biological one. Humans thrived on irregular eating for most of our history. If skipping breakfast or eating two meals works for someone, there's no evidence that it's harmful.
2. The obesity epidemic wasn't caused by meal timing. It was caused by a system-level shift: government policy, food engineering, corn subsidies, and a culture of constant snacking that added ~500 calories/day to the American diet between the 1970s and 2000s.
3. Every popular diet works in the short term, mostly by eliminating processed food. The common thread is that all of them displace the worst parts of the modern food supply with actual food, not any particular stance on carbs, meal timing, or ancestral eating.
4. Mediterranean has the strongest evidence by a wide margin. Large trials, long follow-up, hard endpoints, and it's actually sustainable. If forced to pick one label, this is the one the data supports.
5. Autophagy is real biology, but the popular claims about fasting timelines are mostly extrapolated from animal models. We cannot measure autophagy in living humans. The 16-hour or 24-hour thresholds floating around social media are estimates, not confirmed human data.
6. Most people are deficient in several key nutrients regardless of which diet they follow. Vitamin D, magnesium, omega-3, potassium, and choline are widespread gaps that calorie-counting apps don't track.
7. One RDA for everyone is a starting point, not an answer. Genetic variations in lactose tolerance, vitamin D synthesis, folate metabolism, omega-3 conversion, and starch digestion are well-documented and can significantly affect what a specific person actually needs.
8. Track your own data. Until genotype-informed nutrition recommendations go mainstream, the best available tool is tracking intake, bloodwork, and energy patterns over time. Your body's responses are more informative than any population-level guideline.

Key references

Not exhaustive. Additional studies are cited inline throughout the post.