When to Eat

TL;DR. Your metabolism is not the same at 8am and 10pm. Insulin sensitivity is highest in the morning and falls through the day, so the same meal produces a larger glucose rise at night. Eating late is associated with worse metabolic outcomes in both observational and controlled work, and this is the best-supported timing finding there is. Intermittent fasting works about as well as ordinary calorie restriction and no better, which makes it a useful tool for people who find it easier to follow, not a metabolic trick. Almost everything else about meal timing (eating fruit on an empty stomach, not eating after 6pm, the anabolic window, six small meals) is folklore.

Key takeaways

• Insulin sensitivity is highest in the morning. An identical meal eaten in the evening produces a substantially higher glucose response in controlled studies.
• Late eating is associated with worse outcomes and, in tightly controlled crossover trials, with increased hunger and altered energy expenditure at identical calorie intake.
• Intermittent fasting produces weight loss equivalent to calorie restriction in head-to-head trials. The evidence for benefits beyond that is much thinner in humans than in mice.
• Meal frequency has little effect on metabolic rate. The "eat six small meals to stoke your metabolism" advice is not supported.
• Eating vegetables and protein before carbohydrate in the same meal measurably lowers the glucose response, which is the most practical timing finding in the chapter.

The body clock is a real thing

In short: Nearly every tissue runs a roughly 24-hour molecular clock, the master clock is set by light, and the peripheral clocks in liver and gut are set largely by food.

Your cells contain molecular oscillators: a feedback loop of clock genes (CLOCK, BMAL1, PER, CRY) that cycles over roughly 24 hours. The master clock sits in the suprachiasmatic nucleus of the hypothalamus and is synchronised by light hitting specialised retinal cells. Peripheral clocks sit in liver, pancreas, muscle, and gut, and they are entrained substantially by feeding time rather than by light.

That split is the key to the whole subject. Light sets the brain clock; food sets the metabolic clocks. Eat at times that conflict with your light-set clock and the two disagree, a state called internal desynchrony. This is what shift workers live in, and shift work is associated with higher rates of obesity, type 2 diabetes, and cardiovascular disease in consistent epidemiology, with the International Agency for Research on Cancer classifying night shift work involving circadian disruption as probably carcinogenic.

Things that follow a daily rhythm and matter here:

Melatonin is the mechanistic link that ties this together. Melatonin receptors are present on pancreatic beta cells, and melatonin suppresses insulin secretion. So eating a large meal close to bedtime, when melatonin is already rising, means a glucose load arriving exactly when insulin release is being damped. Carriers of a common variant in the melatonin receptor gene MTNR1B, which is one of the strongest known type 2 diabetes risk variants, show this effect more strongly, which is a satisfying piece of mechanistic convergence.

Late eating: the best-supported finding

In short: The same food eaten late produces higher glucose, and controlled trials show late eating increases hunger and changes energy expenditure at identical intake.

Controlled evidence. A 2022 crossover study by Vujović, Scheer and colleagues at Brigham and Women's Hospital is the cleanest demonstration. Participants ate identical meals, in identical amounts, with identical composition, either early or shifted four hours later, under strictly controlled laboratory conditions. Late eating:

• Increased hunger and altered the appetite hormones leptin and ghrelin in the direction of greater hunger.
• Reduced energy expenditure by a small but measurable amount.
• Shifted gene expression in fat tissue toward storage rather than breakdown.

All at identical calories. That is a rare and valuable design.

Other consistent findings:

• Identical test meals produce substantially higher glucose and insulin responses in the evening than in the morning, reproduced across many studies.
• Front-loading calories (larger breakfast, smaller dinner) produces greater weight loss than the reverse at the same total intake in several trials, though not all, and the effect size is modest.
• Night-eating patterns are associated in cohort studies with higher BMI, worse glycaemic control, and higher cardiovascular risk.
• Eating close to bedtime worsens reflux and, at large volumes, disrupts sleep, which has its own metabolic consequences.

Practical reading: finishing the main eating window two to three hours before bed is sensible and low cost. "Never eat after 6pm" is arbitrary; what matters is the relationship to your sleep, not the clock on the wall.

Intermittent fasting: what the trials actually show

In short: Roughly equivalent to continuous calorie restriction for weight and metabolic outcomes; the human evidence for unique benefits is much weaker than the mouse evidence.

The main protocols:

The head-to-head trials. When intermittent fasting is compared against continuous calorie restriction with matched calorie intake, results converge: similar weight loss, similar changes in blood pressure, lipids, and glycaemic markers. A well-designed 2022 trial in NEJM by Liu and colleagues found time-restricted eating plus calorie restriction produced weight loss no different from calorie restriction alone over 12 months. Multiple meta-analyses agree.

Where TRE alone (no calorie counting) has been tested, results are modest. The 2020 TREAT trial found a 16:8 window produced about 1 kg of weight loss over 12 weeks, no better than the control condition, and raised a concern about loss of lean mass. Other trials have found more benefit. The overall picture: some people spontaneously eat less within a shorter window, and some do not.

The mechanisms people cite, assessed:

• Autophagy. Cellular self-cleaning, upregulated by fasting. Real, important, and overwhelmingly studied in yeast, worms, and mice. The timing in humans is not well established, and the confident claims that autophagy "kicks in at 16 hours" are not based on human measurement.
• Ketosis. Real after roughly 12 to 16 hours of fasting depending on prior glycogen. Not in itself a health outcome.
• Metabolic switching between glucose and fat as fuel, promoted by Mark Mattson and colleagues as the core mechanism. Plausible, and the human outcome data do not yet distinguish it from simple calorie reduction.
• Circadian alignment. This is the most promising strand: early time-restricted eating (window in the morning, e.g. 8am to 4pm) has outperformed late windows in several controlled studies for insulin sensitivity and blood pressure, independent of weight. This suggests that when the window sits matters as much as how long it is, and that the popular pattern of skipping breakfast and eating late may be the worst version of TRE.

Who should not do it: people with a history of eating disorders (fasting protocols can trigger relapse, and this is not a minor caution), people who are pregnant or breastfeeding, underweight people, children and adolescents, people with type 1 diabetes or on insulin or sulfonylureas without medical supervision because of hypoglycaemia risk, and people taking medicines that must be taken with food.

The honest summary: intermittent fasting is a legitimate way to reduce calorie intake for people who find eating windows easier than counting. It is not metabolically magic. If you do it, an earlier window is better supported than a later one.

Meal frequency

In short: Total intake matters; how many meals you split it into matters very little.

The idea that frequent small meals raise metabolic rate rests on a misunderstanding of the thermic effect of food. The thermic effect is a percentage of what is eaten, so dividing the same food into six meals produces six small thermic bumps instead of three larger ones, and the total is identical. Controlled studies confirm no difference in 24-hour energy expenditure between two and six meals at matched intake.

What frequency does affect:

• Appetite control, individually and unpredictably. Some people find grazing prevents overeating; others find it prevents ever feeling satisfied. Trials show no consistent advantage either way.
• Glycaemic variability in people with diabetes, where more even distribution can help.
• Protein distribution, where three to four servings across the day beats one large one for muscle protein synthesis (Chapter 12).
• Dental health, where frequency genuinely matters: each exposure to fermentable carbohydrate produces an acid attack on enamel lasting 20 to 30 minutes, so six snacks are worse than the same food at three meals.
• Snacking on ultra-processed food, which is where the practical harm of frequent eating usually lives.

Meal order: the most useful practical trick

In short: Eating vegetables and protein before carbohydrate in the same meal substantially lowers the glucose response.

This one is well replicated and easy to apply. Studies in both healthy people and those with type 2 diabetes find that consuming vegetables and protein 10 to 15 minutes before the carbohydrate portion of a meal reduces the post-meal glucose peak by 30 to 40 percent compared with the reverse order, with the identical food.

The mechanism is a combination of slowed gastric emptying, increased GLP-1 release, and a physical barrier to rapid starch access.

Related, similarly supported interventions:

• Vinegar (a tablespoon in a dressing) with a starchy meal reduces the glucose response measurably, apparently by slowing gastric emptying and inhibiting some starch digestion.
• A 10 to 15 minute walk after a meal lowers the glucose peak substantially, because muscle contraction moves glucose transporters to the cell surface without insulin. Trials find short post-meal walks outperform a single longer walk elsewhere in the day for glucose control.
• Adding fat, protein, or fibre to a carbohydrate meal flattens the curve, which is why a slice of bread with butter and cheese behaves very differently from bread alone.

These are small effects individually and they are free, immediate, and cumulative.

Breakfast, and the argument about it

In short: The observational case for breakfast is strong and heavily confounded; the trial evidence is much weaker.

Breakfast eaters have lower BMI, better diet quality, and better cardiometabolic markers in cohort studies. They also smoke less, drink less, exercise more, and are more likely to be employed in regular hours. This is the standard problem.

Randomised trials assigning people to eat or skip breakfast find much smaller effects. The Bath Breakfast Project found breakfast eaters had higher total energy intake and more spontaneous physical activity, with no net weight difference. A 2019 BMJ meta-analysis of trials found that eating breakfast was associated with slightly higher total daily energy intake and slightly higher weight, though the trials were short and mostly small.

So: breakfast is not metabolically obligatory. It is genuinely useful for people who otherwise under-eat protein, for children (where the school-performance evidence is better), and for anyone who overeats later without it. And a large morning meal fits the insulin sensitivity rhythm better than a large evening one, so if you eat three meals, making breakfast or lunch the largest is the better-aligned choice.

Timing around exercise

In short: Total daily intake dominates; a few timing effects are real and small.

• Carbohydrate before endurance exercise improves performance in sessions over about 60 to 90 minutes. For shorter sessions it makes little difference.
• Fasted training increases fat oxidation during the session and does not increase fat loss over weeks when total intake is matched. It may reduce performance in hard sessions.
• Protein timing: as Chapter 12 set out, the anabolic window was oversold. Protein within a few hours either side of training is sufficient.
• Post-exercise carbohydrate matters for glycogen replenishment if you are training again within about eight hours. Otherwise, the day's total is what matters.
• Caffeine 30 to 60 minutes before exercise has good evidence as a performance aid (Chapter 88).

The myths

In short: A short list of timing claims with nothing behind them.

"Eat fruit only on an empty stomach, or it rots in your stomach behind other food." Physiologically incoherent. The stomach is a mixing chamber at pH 2; nothing rots in it. Fruit eaten with a meal is digested perfectly well, and eating it after protein and vegetables produces a lower glucose response, not a problem.

"Never combine protein and carbohydrate." Food combining diets have been tested and produce no advantage beyond the calorie reduction that comes from the restriction itself. Your pancreas secretes the full enzyme set for every meal.

"Carbohydrates after 6pm turn to fat." Fat storage depends on energy balance, not the clock. The genuine finding, that late eating produces higher glucose responses and is associated with worse outcomes, is about the whole meal and about circadian alignment, not about carbohydrates specifically.

"You must eat within 30 minutes of waking." No.

"Fasting puts you in starvation mode and wrecks your metabolism." Metabolic rate does fall with sustained weight loss (adaptive thermogenesis, a real and well-measured effect of roughly 10 to 15 percent below predicted), but this is a response to weight loss and energy deficit over weeks, not to skipping a meal. Short fasts slightly increase resting metabolic rate, via noradrenaline.

"Detox by fasting." Your liver and kidneys detoxify continuously. Fasting does not accelerate it, and prolonged fasting slightly impairs some liver detoxification pathways, which is the reverse of the claim.

The bottom line

• Metabolism follows a daily rhythm. Insulin sensitivity is highest in the morning and lowest at night, and melatonin suppressing insulin release is a large part of why.
• Late eating is the best-supported timing problem, demonstrated in controlled crossover trials at identical calorie intake. Finishing two to three hours before bed is a reasonable, cheap rule.
• Intermittent fasting matches calorie restriction and does not beat it. If you use it, an earlier window is better supported than a later one, and it is unsuitable for several groups including anyone with a history of disordered eating.
• Meal frequency barely affects metabolic rate. It affects appetite unpredictably and dental health predictably.
• The most useful practical timing tricks are meal order (vegetables and protein first), vinegar with starchy meals, and a ten-minute walk afterwards.

Sources and notes

Circadian physiology and peripheral clock entrainment by feeding follow Panda's and Bass's reviews. Diurnal variation in insulin sensitivity and the melatonin-insulin interaction follow Poggiogalle et al. and the MTNR1B diabetes genetics literature. The controlled late-eating crossover is Vujovic et al., Cell Metabolism, 2022, from Brigham and Women's Hospital. Intermittent fasting versus continuous restriction follows Liu et al., New England Journal of Medicine, 2022, and the TREAT trial, Lowe et al., JAMA Internal Medicine, 2020. Early time-restricted eating benefits follow Sutton et al., Cell Metabolism, 2018. Metabolic switching is Mattson's framework. Meal frequency and energy expenditure follows Bellisle et al. Meal order effects follow Shukla et al., BMJ Open Diabetes Research and Care, 2015 and 2017. Vinegar and glycaemic response follows Johnston's work. Post-meal walking follows Reynolds et al., Diabetologia, 2016. Breakfast trials follow the Bath Breakfast Project (Betts et al.) and the BMJ meta-analysis by Sievert et al., 2019. Anabolic window revision follows Schoenfeld and Aragon's meta-analyses.

Open questions. Whether intermittent fasting has benefits independent of calorie reduction in humans is unresolved, and the autophagy claims that circulate rest almost entirely on animal work with no established human timing. Long-term trials of early time-restricted eating are still small.

👉 That is the machinery. Now the food itself, starting with the fruit that has an entire proverb attached to it: apples.