Often, yes. When researchers measure what people really eat and compare it with what they write down, the written record usually comes up short: by 47% on average in a group of dieters who reported very low intakes1, and by about 19% even in athletes3.
This is not about lying. It is about how hard it is to see portions, count snacks and remember drinks. This article walks through the studies that measured the gap, how big it is, and what it does and does not mean for your own log.
How do researchers know what people really eat?
You cannot follow someone around with a scale. So the studies measure the other side of the equation: energy burned. When body weight holds steady over a couple of weeks, energy in roughly equals energy out. That means measured energy burned is a fair stand-in for what a person truly ate.
The reference tool is doubly labelled water. You drink water whose hydrogen and oxygen carry harmless, rare natural tags. Urine samples over one to two weeks show how fast the tags leave, and that gives the energy your body used in normal life. Reviews of athlete nutrition describe it as the "gold standard" for checking self-reported intake3.
The self-reported side comes in three common forms:
| Method | What the person does |
|---|---|
| Food record (food diary) | Writes down everything eaten and drunk, for several days, as it happens |
| 24-hour recall | Tells an interviewer everything eaten in the previous day |
| Food frequency questionnaire (FFQ) | Ticks how often they eat each food over months |
In plain terms: the studies below compare what people wrote down with an objective measure of what their body used.
The 1992 study of "diet-resistant" dieters
The best-known study was published in the *New England Journal of Medicine* in 19921. Among 224 consecutive patients with obesity, the team looked closely at 10 people (nine women, one man) who said they kept failing to lose weight even though they reported intakes below 1,200 kcal a day. A comparison group of 80 patients without that history served as controls.
Over 14 days, the researchers measured energy burned and actual intake, using indirect calorimetry (breathing measures that read oxygen used) and body-composition analysis1.
The results, in order:
- Their metabolism was normal. Total energy burned and resting metabolic rate were within 5% of the values predicted for their body composition. The energy cost of digesting food and of exercise did not differ from the controls1.
- Their food records were short. On average they under-reported what they actually ate by 47% (standard deviation 16%)1.
- Their activity records were long. They over-reported physical activity by 51% on average, with a standard deviation of 75%, so individuals varied enormously1.
- Nothing unusual about them as people. The group had no distinct psychological disorder. They did describe their eating as relatively normal and more often believed their weight had a genetic cause1.
The authors concluded that the stalled weight loss came from intake higher than reported and an overestimate of activity, "not to an abnormality in thermogenesis" (heat and energy production)1.
In plain terms: the engine was working as expected. The fuel gauge, the food diary, was reading low.
It is not only dieters
A 47% gap in 10 carefully selected people is an extreme case. The broader literature finds a smaller but consistent gap in almost every group studied.
Middle-aged adults. The OPEN study measured 484 men and women aged 40 to 69 in Maryland with doubly labelled water2. On 24-hour recalls, men under-reported energy by 12–14% and women by 16–20%. On food frequency questionnaires the gap was larger: 31–36% for men and 34–38% for women.
Athletes. A 2017 systematic review pooled 11 studies that compared athletes' food records (kept for 4 to 10 days) with doubly labelled water, each with 7 to 19 athletes3. Reported intake was about 19% below measured expenditure on average, ranging from 0.4% to 36% across studies. Body mass was stable in most studies, so the authors judged the gap "likely due to misreporting" rather than real under-eating3.
Adults in general. A 2019 review gathered 59 studies with 6,298 free-living adults4. Most reported significant under-reporting compared with doubly labelled water, and few found over-reporting. Women misreported more often than men on recall-based methods, and 24-hour recalls varied less than the other methods.
How far reported intake fell below measured intake or expenditure · %
In plain terms: across very different people, the written record tends to sit below the truth, and the size of the gap depends on the person and the method.
How common is it in national surveys?
A 2025 study built a prediction equation for daily energy burned from 6,497 doubly labelled water measurements in people aged 4 to 965. The equation gives a plausible range for someone's intake from weight, age and sex. The team then checked two national diet surveys, one from the UK and one from the US, with 18,567 people in total.
A large share of adult reports fell below the lowest plausible intake5:
- US survey (2017–2018): 32.8% of men and 28.4% of women.
- UK survey (years 1–11): 38% of men and 29.7% of women.
The average shortfall was 3.5 MJ a day in the UK survey (about 840 kcal) and 1.8 to 2.8 MJ a day in the US survey (about 430 to 670 kcal), by our unit conversion5. Collecting more days did not fix it. The share of plausible reports was the same across repeated survey days, which the authors read as consistent under-reporting rather than people eating less while being watched5.
Adult food reports below the plausible intake range · %
One more pattern matters for macros. People who under-reported total energy tended to report a higher share of protein and a lower share of fat5. That fits the idea that some foods go missing more easily than others.
Why food goes missing
The athlete review lists the usual suspects3:
- Snacks. Frequent small eating occasions are easy to forget.
- Portions. People write "a bowl" or "a serving" and guess the size.
- Eating away from home. Meals someone else cooked are hard to estimate.
- Drinks. Energy in drinks is easy to leave out.
- The burden itself. Writing everything down for days is tiring, and effort fades.
The 2025 paper adds a quieter source: the person converting a diary into numbers has to assume sizes, "for example, by assuming that all apples are the same size"5.
In plain terms: most missed calories are small, frequent and ordinary: a bigger portion than written, a snack, a drink. Not hidden meals.
What this does and does not mean
It does not mean you are lying. The 1992 dieters had no distinct psychological disorder1. Estimating food is simply hard.
It does not mean everyone misses 47%. That number came from 10 people chosen because their reports and their weight did not match. In broader groups the average gap is smaller, and some people report accurately.
It does not mean metabolism never differs. People of the same size and age do differ in energy use. The 1992 study only shows that in this group, a slow metabolism was not the explanation.
It does mean a food log is an estimate. Your log is useful for trends, but its total is not a measurement. A steady bias, the same missed item every day, is less of a problem than it sounds, because an app that learns from your weight trend absorbs it (see how accurately you need to track calories and how apps estimate energy expenditure).
The trials and reviews above point in one direction: weighing food, logging as you eat, and counting drinks and extras all narrow the gap. That is our reading of their findings, not an intervention any of these studies tested. For why weight can stall even when a log looks right, see why has my weight loss stalled.
What the studies found
| Study | Who was studied | Key result | Year |
|---|---|---|---|
| Lichtman et al., *N Engl J Med*1 | 10 adults with obesity reporting diet resistance, plus 80 controls; 14 days | Energy burned within 5% of predicted; intake under-reported by 47%, activity over-reported by 51% | 1992 |
| Subar et al., *Am J Epidemiol* (OPEN)2 | 484 adults aged 40 to 69; doubly labelled water | 24-hour recalls under-reported energy by 12–20%; questionnaires by 31–38% | 2003 |
| Capling et al., *Nutrients*3 | Athletes in 11 studies vs doubly labelled water | Food records about 19% below measured expenditure (range 0.4–36%) | 2017 |
| Burrows et al., *Front Endocrinol*4 | 59 studies, 6,298 adults | Most studies found significant under-reporting; few over-reporting | 2019 |
| Bajunaid et al., *Nature Food*5 | 18,567 people in two national surveys, screened with an equation from 6,497 measurements | About 28–38% of adult reports below the plausible range | 2025 |
Limits worth knowing
- The famous study is tiny. Ten people, selected because their reports did not add up. We read its abstract only; the full paper is closed.
- Different reference tools. The 1992 study used indirect calorimetry and body composition, not doubly labelled water. The later studies used doubly labelled water.
- "Below the plausible range" is not the same as lying or under-eating. The 2025 method cannot tell the two apart for one person, and the authors caution that some deviations reflect real under- or overeating5.
- The 2025 paper was corrected. Its original abstract gave a different headline misreporting figure. We use only the numbers in its Results section.
- Small athlete studies. The review found few robust studies, with 7 to 19 athletes each3.
- Abstracts only for the OPEN study and the 2019 review.
Latest research
The newest study we opened is the 2025 *Nature Food* paper5. Its equation, built from doubly labelled water measurements in 6,497 people, flagged roughly a third of adult reports in two national surveys as implausibly low, more than older screening methods caught. It also showed that under-reporting skews the apparent make-up of the diet, not just the total. We found no newer study that reverses the direction: when intake is measured objectively, self-reports tend to fall short.
The short version
- When measured, what people eat is usually more than what they record.
- In 10 dieters who reported very low intakes, the gap was 47%, and their metabolism was normal.
- In athletes the gap was about 19%; in broad groups of adults it was often 12–20% on recalls.
- It is a measurement problem, not a character flaw: portions, snacks and drinks are hard to see.
- Read the log as an estimate, and let your weight trend correct it.
Educational overview only — not medical advice.
