Category: Measurement

  • Who Invented BMI? The History of Measuring Body Weight

    Who Invented BMI? The History of Measuring Body Weight

    In short: BMI was invented by a Belgian astronomer and statistician, Adolphe Quetelet, who noticed in the 1830s that adult weight rises roughly with the square of height. It was a tool for describing the “average man,” not for judging anyone’s health. The name “body mass index” came much later, in 1972, from American physiologist Ancel Keys, and the familiar cutoffs of 25 and 30 were adopted by the NIH and the World Health Organization in the late 1990s. Since 2023, major medical groups have said BMI works best as a first screening number, used alongside waist size and other measures.

    This page tells the story of how a 19th-century statistics project became the number on your chart, and why doctors are now adding measures beside it. If you just want your number, our BMI calculator works it out in seconds; for the wider story of weight and health, see our history of weight loss.

    How BMI came to be, at a glance

    1. 1830sQuetelet finds that adult weight rises about with the square of height (described 1832, in his 1835 book).
    2. Before BMIDoctors use height-weight tables built from life-insurance data (revised as late as 1983).
    3. 1972Ancel Keys tests several weight indexes in 7,424 men and names the winner the "body mass index".
    4. 1985An NIH consensus conference reviews the health effects of obesity.
    5. 1997-1998A WHO consultation and new NIH guidelines set 25 for overweight and 30 for obesity.
    6. 2004WHO experts note higher risk at lower BMI in many Asian populations.
    7. 2022-2025Waist-to-height ratio joins BMI in UK guidance; the AMA and a Lancet Commission say BMI alone is not enough.

    1830s: an astronomer counts people

    Adolphe Quetelet (1796 to 1874) was a Belgian mathematician, astronomer and statistician who wanted to apply the tools of astronomy to human beings. He measured births, deaths, heights and weights to describe what he called the “average man.” According to a 2008 history in the journal Nephrology Dialysis Transplantation, he described the weight-height relationship in 1832, and it appeared in his principal work, Sur l’homme et le développement de ses facultés, published in 1835 and translated into English in 1842 as A Treatise on Man and the Development of His Faculties.

    In the English edition, Quetelet explains that if people grew equally in every direction, weight would rise with the cube of height, but that is not what he saw. Between infancy and puberty, he wrote, weight increases nearly as the square of height, and when he compared fully grown adults of different heights, he found their weight was “nearly as the square of the stature.” That observation, weight divided by height squared, is the whole of the BMI formula.

    Two things are worth knowing. First, Quetelet was describing averages in a population, not diagnosing anybody; his treatise is about statistics, growth and social patterns, not about obesity. Second, his ratio was known for over a century as the “Quetelet index.” He went on to organize the first International Statistical Congress in 1853, which pushed for a common list of causes of death, an early ancestor of today’s International Classification of Diseases.

    Before BMI: insurance height-weight tables

    For much of the 20th century, doctors did not use Quetelet’s ratio. They used height-weight tables. A 2000 review in Annals of Pharmacotherapy explains that these tables were built to compare groups by relative weight, and that because the weights in them were linked to lower death rates among insured people, they came to be called “desirable” or “ideal” weights. The search for a simpler index, the 2008 history adds, began after life-insurance actuaries reported higher death rates among their heavier policyholders.

    The tables had quirks. The Metropolitan Life tables of 1983 split people into small, medium and large frames by elbow width, which was supposed to put about 25%, 50% and 25% of people in each group. When Canadian researchers tested this in 19,305 people in 1989, the split did not hold, and very few people came out as large-framed. The “ideal body weight” formulas used to dose some medicines, such as Devine’s equations, grew out of these same tables; the 2000 review found that the various formulas agree because the tables behind them agree. Our ideal weight calculator shows these formulas side by side and explains their limits.

    1972: Ancel Keys gives it a name

    After World War II, researchers studying heart disease needed a way to compare body weight across very different countries. Ancel Keys of the University of Minnesota, working with colleagues in Italy, Finland and Japan, tested several “indices of relative weight” in 7,424 healthy men in 12 groups across five countries, plus body-density measurements in two smaller groups. They judged each index by two tests: it should not depend much on height, and it should track body fat as closely as possible.

    The ponderal index, which uses the cube root of weight, did worst. Percentage of average weight for height worked, but only within a given population. Weight divided by height squared came out slightly better than weight divided by height, and in their 1972 paper in the Journal of Chronic Diseases the authors wrote that “the ratio of weight to height squared, here termed the body mass index,” seemed preferable because it was simple to calculate and could be applied to all populations at all times. Note what they did and did not claim: BMI was the best of several simple options for comparing groups, not a precise measure of an individual’s fat.

    The studies behind the number

    Checked on October 7, 2026

    1985 to 1998: the cutoffs we use today

    BMI moved from research into the clinic over the next quarter century. In February 1985, the National Institutes of Health held a consensus development conference on the health implications of obesity, published that December in Annals of Internal Medicine. In June 1997, a World Health Organization consultation in Geneva met on preventing and managing obesity; its report, Obesity: Preventing and Managing the Global Epidemic, followed in 1998 and was later issued in WHO’s Technical Report Series. WHO’s definitions today are simple: for adults, overweight is a BMI of 25 or more and obesity a BMI of 30 or more.

    In September 1998, the NIH’s National Heart, Lung, and Blood Institute published its Clinical Guidelines on the Identification, Evaluation, and Treatment of Overweight and Obesity in Adults. It defined overweight as a BMI of 25 to 29.9 and obesity as 30 or above, estimated that 97 million American adults were in those ranges, and noted that a BMI of 30 equals about 221 pounds for someone 6 feet tall and 186 pounds at 5 feet 6. The guidelines also paired BMI with a second number: waist circumference above 40 inches for men or 35 inches for women signaled higher risk. The same report admitted that the cutoffs are “somewhat arbitrary,” because disease risk rises gradually as weight rises rather than jumping at one number.

    Those data also show how quickly things changed. Using the new definitions, national surveys found that obesity rose from 10.4% to 19.9% of men and from 15.1% to 24.9% of women between 1960 and 1994. By August 2021 to August 2023, the CDC found that 40.3% of American adults had a BMI of 30 or more, and 9.4% had severe obesity (a BMI of 40 or more).

    U.S. adults with obesity by BMI, 1960 and 1994Values in %
    U.S. adults with obesity by BMI, 1960 and 1994
    ItemMenWomen
    196010.4%15.1%
    199419.9%24.9%

    Obesity = BMI of 30 or more. By 2021-2023 the CDC put the figure at 40.3% of all adults (men and women together).

    Source: NHLBI Clinical Guidelines on Overweight and Obesity in Adults (1998), citing NCHS/CDC surveys (checked on October 7, 2026)

    The categories, as used today

    The CDC’s adult categories, for ages 20 and up, are below. The CDC calls BMI a screening measure and says it should be considered with other factors when assessing a person’s health; it is not a diagnosis.

    BMICDC category (adults 20+)Lower thresholds NICE uses for some groups*
    Below 18.5UnderweightSame
    18.5 to under 25Healthy weight18.5 to under 23
    25 to under 30Overweight23 to 27.4
    30 to under 35Obesity, class 1Obesity from 27.5
    35 to under 40Obesity, class 2Classes 2 and 3 about 2.5 points lower
    40 or moreObesity, class 3 (severe)
    *NICE (England) advises lower thresholds for people of South Asian, Chinese, other Asian, Middle Eastern, Black African or African-Caribbean background. Sources: CDC adult BMI categories; NICE NG246. Checked on 2026-10-07.

    2004: one number does not fit every body

    Keys had hoped BMI would apply to all populations, but doubts appeared quickly. In 2004, a WHO expert consultation published in The Lancet reviewed evidence that many Asian populations have more body fat and higher health risks at the same BMI than European populations. It found that the BMI at which risk rises varies from 22 to 25 in different Asian populations, kept the international cutoffs, and suggested extra “public health action points” at 23, 27.5, 32.5 and 37.5. England’s NICE now uses lower thresholds for several groups, as shown in the table above.

    There are other blind spots. The 2025 Lancet Commission described below notes that BMI can both underestimate and overestimate body fat: because it only sees weight, a very muscular person can score high, while someone who has lost muscle can score in the “healthy” range and still carry a lot of fat around the organs. That kind of deep belly fat, called visceral fat, matters for heart and metabolic health; our visceral fat guide explains why, and the belly fat guide covers what helps.

    2012 to 2022: the waist joins the conversation

    Because BMI can’t see where fat sits, researchers looked for a simple measure that can. A 2012 meta-analysis in Obesity Reviews pooled 31 studies covering more than 300,000 adults in several ethnic groups and found that waist-to-height ratio was better than both waist size and BMI at picking out people with high blood pressure, diabetes and other heart and metabolic risk factors. Waist size alone improved on BMI by about 3%, and waist-to-height ratio by about 4% to 5%.

    In 2022, England’s National Institute for Health and Care Excellence (NICE) added waist-to-height ratio to its guidance for adults with a BMI under 35, with a memorable message: keep your waist to less than half your height. Its current guideline sorts a ratio of 0.4 to 0.49 as healthy, 0.5 to 0.59 as increased risk and 0.6 or more as high risk. You need only a tape measure: measure around your waist and divide by your height in the same units.

    BMIWaist-to-height ratio
    What it isWeight (kg) divided by height (m) squaredWaist divided by height, same units
    Where it came fromQuetelet (1830s); named by Keys (1972)Backed by a 2012 meta-analysis; added to NICE guidance in 2022
    What it seesTotal weight for heightFat around the middle (central adiposity)
    Common cutoffs25 overweight, 30 obesity (lower for some groups)Under 0.5 healthy; 0.5 to 0.59 increased; 0.6+ high
    Main blind spotCan’t tell muscle from fat or see where fat sitsNeeds a careful tape measurement; NICE uses it for adults with a BMI under 35
    Best usePopulation tracking and first screeningAdding a risk signal alongside BMI
    Sources: Keys et al. 1972; Ashwell et al. 2012; NICE NG246; CDC. Checked on 2026-10-07.

    2023 to 2025: “useful, but not on its own”

    In June 2023, the American Medical Association adopted a policy calling BMI an imperfect way to measure body fat. The policy notes that BMI was based mainly on data from earlier generations of non-Hispanic white people, that it has been used in ways that caused historical harm, and that it loses predictive power at the individual level. The AMA now suggests using BMI together with other measures such as waist circumference, body composition, visceral fat and metabolic factors, and says BMI alone should not be the reason to deny insurance coverage.

    In January 2025, a Lancet Diabetes and Endocrinology Commission of 58 experts, including people with lived experience of obesity, went further. It recommended using BMI only as a population-level or screening measure, confirming excess body fat with a direct measurement or at least one other body measure such as waist circumference, waist-to-hip or waist-to-height ratio. It also proposed separating “preclinical obesity” (excess fat with normal organ function) from “clinical obesity” (excess fat that is already affecting how organs or daily activities work). For people with a BMI above 40, it said, excess fat can be assumed.

    What this history means for your number

    Almost two centuries on, Quetelet’s ratio is still the quickest way to sort large groups and to start a conversation, and many guidelines still use it. FDA labels for weight-management medicines, for example, describe their approved use in terms of BMI, as our FDA-approved medications comparison shows, and bariatric surgery criteria also start from BMI. But the history is clear that it was never meant to be the whole story about one person’s health. A gentler, more complete picture adds your waist, how you feel and function, and your blood pressure, blood sugar and cholesterol.

    If you want more than one number, our body fat calculator estimates body fat from tape measurements, and our comparison of smart scales, DEXA and tape measures explains how accurate each method is. If you track at home, our best smart scales list explains what their body-fat readings can and can’t tell you, and if you are building muscle while losing fat, the body recomposition guide explains why the scale can stall while your shape changes. For how weight got counted in calories, read the companion story, the history of calorie counting.

  • The History of Calorie Counting: From Atwater to AI Apps

    The History of Calorie Counting: From Atwater to AI Apps

    In short: Calorie counting began as a 19th-century physics unit that an American chemist, Wilbur O. Atwater, turned into a way to measure food. His U.S. Department of Agriculture bulletins of the 1890s and 1900s gave the 4-9-4 rule (calories per gram of protein, fat and carbohydrate) that is still written into food-label law. Physician Lulu Hunt Peters made counting a diet tool in 1918, Congress put calories on almost every package in 1990, and phones and AI apps now estimate them from a photo. The method has always worked best as a rough guide: labels, menus, memory and apps can all be off by a meaningful amount.

    This page follows the calorie from the lecture hall to your phone. If you want to work out your own numbers, our calorie calculator uses a modern equation, and our calorie deficit guide explains how a deficit works in practice. For the bigger story, see our history of weight loss.

    Calorie counting at a glance

    1. 1819-1824Nicolas Clément uses "Calories" in lectures on heat engines in Paris.
    2. 1887-1902Atwater brings the Calorie to American food, with the 4-9-4 factors.
    3. 1918Lulu Hunt Peters' Diet and Health makes counting a weight-loss method.
    4. 1918-1990Equations estimate how many calories a body burns (Harris-Benedict, then Mifflin-St Jeor).
    5. 1990-1994The Nutrition Labeling and Education Act puts calories on packaged food.
    6. 2016-2021A new label makes "Calories" bigger; chain menus list calories from 2018.
    7. 2010s-2020sTracking moves to apps, then to photo-based AI estimates.

    Before food: the calorie as a unit of heat

    The calorie started in physics, not nutrition. A 2006 history in the Journal of Nutrition by James Hargrove traces it to Nicolas Clément, who used “Calories” in lectures on heat engines in Paris between 1819 and 1824. The word was defined in a French dictionary by 1845 and entered English in 1863 through a translated physics textbook, which defined a Calorie as the heat needed to raise 1 kilogram of water by 1 degree Celsius. That big “Calorie” is the same as the kilocalorie (kcal) we use for food today.

    1887 to 1902: Wilbur Atwater measures food

    The capital-C Calorie on American labels, Hargrove writes, comes from Wilbur O. Atwater’s 1887 article on food energy in Century magazine and his 1894 USDA Farmers’ Bulletin No. 23, Foods: Nutritive Value and Cost. Atwater, a chemistry professor at Wesleyan University, wanted working families to get the most nourishment for their money. In that 1894 bulletin he also proposed a daily standard: 3,500 calories of energy for a man at moderate muscular work, higher than the German standard of about 3,050 because, he wrote, Americans led a more active life.

    Atwater then built the evidence. With Charles D. Woods he published The Chemical Composition of American Food Materials (USDA Bulletin No. 28, 1896), a large set of food tables. With physicist E. B. Rosa he built a respiration calorimeter, a sealed room in which a person could live while every bit of heat and breath was measured, described in USDA Bulletin No. 63 in 1899. Those experiments let him check how much of the energy in food the body actually uses.

    The result is the rule every nutrition student still learns. In Farmers’ Bulletin No. 142, Principles of Nutrition and Nutritive Value of Food (1902), Atwater gave the fuel value of protein as 4 calories per gram, fats 8.9 calories per gram and carbohydrates 4 calories per gram, counting only what is digested and burned. He noted that older figures of 4.1 and 9.3, which did not allow for losses, had come into common use. More than a century later, the federal rule for Nutrition Facts labels still lets food makers calculate calories with “the general factors of 4, 4, and 9” or with Atwater’s more specific factors.

    Calories per gram: Atwater 1902 vs today's food label
    Calories per gram: Atwater 1902 vs today's food label
    ItemAtwater, USDA Bulletin 142 (1902)General factors in U.S. label rules (21 CFR 101.9)
    Protein44
    Fat8.99
    Carbohydrate44

    Food makers may also use Atwater's food-specific factors or lab measurements. Alcohol and fiber have their own values.

    Source: Atwater WO, USDA Farmers' Bulletin No. 142 (1902); 21 CFR 101.9(c)(1)(i) (checked on October 7, 2026)

    1918: Lulu Hunt Peters makes it a diet

    Atwater measured food for economy and nutrition. Physician Lulu Hunt Peters turned his unit into a weight-loss method in Diet and Health, With Key to the Calories (1918), now in the public domain. She told readers to use the word calorie as often as “foot, yard, quart, gallon,” and to stop thinking in slices and pieces: you would now eat “100 Calories of bread, 350 Calories of pie.” She wrote that she had once carried 70 pounds more than her usual weight, walked readers through cutting an example maintenance diet of 2,200 calories down to 1,200, and supplied tables of 100-calorie portions. The book’s warm, funny tone and simple arithmetic made counting a household habit.

    Some of Peters’ ideas have aged well, such as noting that a small daily excess adds up over a year. Others have not; her very low fixed targets would not be advised for everyone today. Modern guidance from the NIH points to an initial goal of losing 5% to 10% of starting weight over about six months, with a calorie target set for the person, and our history of diet fads shows how her book set the stage for a century of plans that either embraced counting or promised you could skip it.

    1918 to 1990: estimating the other side of the equation

    Counting what you eat is only half of the math; you also need an idea of what you burn. In 1918, the same year as Peters’ book, J. Arthur Harris and Francis G. Benedict of the Carnegie Institution’s Nutrition Laboratory published “A Biometric Study of Human Basal Metabolism” in the Proceedings of the National Academy of Sciences, the basis of the Harris-Benedict equations. In 1990, Mark Mifflin, Sachiko St Jeor and colleagues published a newer equation in the American Journal of Clinical Nutrition, built from measurements in 498 healthy adults, both normal weight and with obesity, aged 19 to 78. The Mifflin-St Jeor equation is the main one our TDEE calculator uses today.

    The old shortcut that “a pound equals 3,500 calories” has been revised too. NIH researcher Kevin Hall and colleagues showed in The Lancet in 2011 that the body adapts as weight changes, so the 3,500-calorie rule overestimates long-term loss. Their rule of thumb for an average adult with overweight: every lasting change of about 10 calories a day changes weight by about a pound in the end, with half of that change coming in about a year and nearly all of it in about three years. Our metabolism guide explains the adaptation.

    1990 to 2021: calories on every label and menu

    For most of the 20th century, a calorie count on a package was optional. The Nutrition Labeling and Education Act, Public Law 101-535, was signed on November 8, 1990. According to the FDA, it required all packaged foods to carry nutrition labeling, and according to the USDA’s Economic Research Service, the Nutrition Facts label it mandated dates from 1994. The FDA notes that the serving sizes behind that label were published in 1993.

    In 2016, the FDA finalized an updated label, its first big change in more than 20 years. “Calories” and the serving size became larger and bolder, serving sizes were updated to reflect how much people actually eat and drink, and the “Calories from Fat” line was dropped because the type of fat matters more than the amount. Large manufacturers had to switch by January 1, 2020, and smaller ones by January 1, 2021. Meanwhile, chain restaurants and similar food sellers with 20 or more locations have had to show calories on menus and menu boards since May 7, 2018, along with the reminder that “2,000 calories a day is used for general nutrition advice, but calorie needs vary.”

    How accurate are the numbers?

    Labels are close, not exact. Federal rules treat a packaged food as misbranded if its calories are more than 20% above the label. When Tufts University researchers measured food in the lab, the average stated calories of 269 restaurant items from 42 restaurants matched well overall, but 19% of items had at least 100 calories more than stated. In an earlier study of reduced-calorie meals, restaurant dishes averaged 18% more calories than stated and supermarket frozen meals 8% more.

    The bigger gap is usually our own memory. In a 1992 New England Journal of Medicine study of people who said they could not lose weight on fewer than 1,200 calories a day, careful measurement showed their metabolism was normal; they underreported what they ate by an average of 47% and overreported their activity by 51%. That isn’t a character flaw. It is simply very hard to estimate portions by eye, which is why a kitchen scale helps so many people who count.

    How far calorie numbers can driftValues in %
    How far calorie numbers can drift
    ItemValue
    Reduced-calorie restaurant meals, above stated18%
    Supermarket frozen meals, above stated8%
    Self-reported food, below actual (diet-resistant group)47%
    AI photo estimates, average error (two chatbots)35.8%

    Different studies, methods and directions of error; shown together only to give a sense of scale.

    Source: Urban LE et al., J Am Diet Assoc 2010; Lichtman SW et al., NEJM 1992; Fridolfsson J et al., Curr Dev Nutr 2025 (checked on October 7, 2026)

    From paper diaries to phones and AI

    Writing down what you eat is, in researchers’ words, the centerpiece of behavioral weight-loss programs. A 2011 systematic review in the Journal of the American Dietetic Association looked at 22 studies published from 1993 to 2009: the paper diary was the most common tool, with the internet, handheld computers and digital scales appearing in five studies. Across the studies, self-monitoring was consistently linked with weight loss, though the authors rated the evidence as weak because of study limitations.

    The 2010s moved the diary into phone apps with barcode scanners and big food databases, and the 2020s added photo-based estimates. Accuracy is still a work in progress. A 2023 review of 52 studies of image-based dietary assessment found calorie errors ranging from 0.1% to 38.3%, lowest for simple single foods. In a 2025 test of 52 standardized food photos, two AI chatbots misjudged calories by about 36% on average and underestimated more as portions grew. Our article on how accurate AI calorie counting is covers these studies in detail, and our best weight-loss apps list compares today’s trackers.

    Calorie counting by the numbers

    • 4-9-4calories per gram of protein, fat and carbohydrate, from Atwater's work, still allowed on U.S. labels21 CFR 101.9
    • 20%how far above the label a packaged food's calories can be before it counts as misbranded21 CFR 101.9(g)
    • 47%average underreporting of food in a 1992 study of "diet-resistant" adultsLichtman SW et al., NEJM 1992

    Checked on October 7, 2026

    What a century of counting teaches

    Atwater’s numbers have lasted because energy balance is real: trials of named diets keep finding that the calories, more than the label on the diet, drive weight loss. What history adds is humility about precision. Treat every number, from the label to the app to your own estimate, as an approximation, and watch the trend in your weight and waist over weeks rather than the total for one day. Counting is also not the only route: our comparison of intermittent fasting vs calorie counting shows that structured eating windows can work about as well for many people, and the macro calculator helps if you would rather track protein than every calorie.

    If you are keeping weight off, the habit of self-monitoring matters even more; see our habits of people who keep weight off and the maintenance guide. And for the story of the other number in most weight conversations, read who invented BMI.