Fluids

The weighing that is mostly a measurement of air

To measure how much of a body is fat, a person is weighed in air and then again lying under water after breathing out as hard as possible. The difference is Archimedes' principle applied to a human being, and it gives the body's volume and so its density. The reading under water for a 75 kg person is about three kilograms — and the air that no amount of breathing out can expel from the lungs is holding up 1.3 kilograms of it. Every tenth of a litre of that air misjudged moves the answer by two-thirds of a percentage point of fat. The weighing is an exquisite measurement of something else.

Assumes: The weight of the water that is not there · The depth past which it must sink

Archimedes is supposed to have found how to tell whether a crown was pure gold by thinking about the water his body displaced in a bath. The method that bears his principle and is still used in sports science and nutrition labs reverses the story: it weighs the body in the bath. A person sits on a chair hung from a scale in a tank of warm water, breathes out as far as they can, bends forward until fully submerged, and holds still for a few seconds while the scale settles. The reading, compared with the weight in air, gives the volume of the body; the volume and the weight give its density; and the density, through a rule, gives the share of the body that is fat.

It is a beautiful use of a two-thousand-year-old principle, and for decades it was the reference against which other methods of measuring body fat were judged. It is also a measurement in which the quantity the scale responds to most is not the person but the air inside them. Following the numbers through for one person shows how a measurement can be very precise and still be dominated by a correction.

Archimedes for a person

The weight of the water that is not there put the principle in its plainest form: a body submerged in a fluid is pushed up by a force equal to the weight of the fluid it displaces. A person’s weight under water is their weight in air minus the weight of the water their volume displaces. If MM is the mass in air and MwM_w the reading under water, the volume of everything submerged is (M−Mw)/ρw(M - M_w)/\rho_w, with ρw\rho_w the density of the water at the tank’s temperature — about 0.994 kilograms per litre at 35 °C.

Everything submerged includes gas. Lungs that have been emptied as far as possible still hold about a litre and a quarter of air in an adult, the residual volume, which cannot be expelled because the small airways close before the alveoli empty. The gut holds a variable amount, conventionally taken as a tenth of a litre. Neither has any weight to speak of, and both displace water. So the body’s own volume, the one that belongs in its density, is

V=M−Mwρw−VRV−Vgut.V = \frac{M - M_w}{\rho_w} - V_\text{RV} - V_\text{gut}.

Three kilograms on the scale, and the lungs account for 1.3 of them. The weight under water of a 75 kg person who is 18% fat, broken into its parts. The body's tissue alone would weigh 4.51 kg submerged; the 1.3 L of air the lungs cannot empty holds up 1.29 kg of that, and gas in the gut another 0.10, so the scale reads 3.12 kg. The measurement of body fat is made from this reading, and the lungs' air alone is worth 41% of it.
Fig. 1 The weight under water of a 75 kg person who is 18% fat, broken into its parts. The body’s tissue alone would weigh 4.51 kg submerged; the 1.3 L of air the lungs cannot empty holds up 1.29 kg of that, and gas in the gut another 0.10, so the scale reads 3.12 kg. The measurement of body fat is made from this reading, and the lungs’ air alone is worth 41% of it.

The figure follows a person of 75 kilograms who is 18 per cent fat. Their tissue has a density of 1.058 kilograms per litre, very nearly that of water, and if they were solid tissue they would weigh four and a half kilograms under water. The air in their lungs holds up almost a third of that, and the scale reads 3.12 kilograms. The body’s density differs from water’s by six per cent; the reading is the difference; and the correction for the lungs is two-fifths of the reading. The measurement is a small difference between two large numbers, with a correction comparable to the difference.

The crown was the easy case

The legend’s crown makes a useful comparison, because it shows what makes a density measurement easy. Gold has a density of 19.3 grams per cubic centimetre and silver 10.5. A one-kilogram crown with a fifth of its gold replaced by silver has a volume of 60.5 cubic centimetres instead of 51.8, and weighed under water it is 8.7 grams lighter than pure gold would be — nearly one per cent of its weight, easily seen on a balance of Archimedes’ time with the crown and an equal mass of gold hung from the two arms and lowered into water together. The story of the overflowing bath, in which the difference would be a film of water a fraction of a millimetre thick round a basin, is almost certainly not how it was done; the balance is.

A body is the hard case on every count. Its two components differ in density by about a fifth, not by a factor of two, so the signal per unit of composition is small. Its average density is within six per cent of the water’s, so the underwater weight is a small remainder of a large number. And it contains gas that has nothing to do with its composition and is comparable to the remainder. A crown has none of these problems, and the same principle that settled the crown in an afternoon needs a laboratory protocol to give a body’s fat to within a few points.

A density read through a rule

The volume and the mass give the density, and the density is not yet a fat fraction. Turning it into one needs a model of what a body is made of. The simplest, still the one most used with underwater weighing, divides the body into two compartments: fat, with a density of about 0.900 kilograms per litre, and everything else — water, protein, bone mineral — lumped together as fat-free mass with a density of about 1.100. A body that is a fraction ff fat then has 1/D=f/0.900+(1−f)/1.1001/D = f/0.900 + (1 - f)/1.100, and solving for ff gives William Siri’s equation of 1961, %fat=495/D−450\%\text{fat} = 495/D - 450. Josef Brozek’s slightly different assumptions give 457/D−414.2457/D - 414.2.

A percentage of fat is a density read through a rule. Body fat inferred from whole-body density by Siri's rule, 495/D − 450, and Brozek's, 457/D − 414.2. Both assume the body is two substances of fixed density, fat at 0.90 kg/L and everything else near 1.10. The worked person, 18% fat, has a density of 1.0577 kg/L. Near there the line falls by 0.44 percentage points for every 0.001 kg/L, so a density good to three decimal places is a percentage good to about half a point.
Fig. 2 Body fat inferred from whole-body density by Siri’s rule, 495/D−450495/D - 450, and Brozek’s, 457/D−414.2457/D - 414.2. Both assume the body is two substances of fixed density. The worked person, 18% fat, has a density of 1.0577 kg/L; near there the line falls by 0.44 percentage points for every 0.001 kg/L.

The rule is steep. A change in density of one part in a thousand, 0.001 kilograms per litre, moves the answer by 0.44 of a percentage point. To know a person’s fat to within a point, their density has to be known to about two parts in a thousand — the precision of a decent laboratory measurement of the density of a liquid, applied to a living person in a tank who has just breathed out as far as they can and is trying not to move. That the method works at all is a tribute to how well Archimedes’ principle holds, and to how carefully it is applied.

Every tenth of a litre of air is two-thirds of a point

The correction for the lungs is where the difficulty concentrates. Residual volume is not the same in everyone; it is larger in older people, in taller people and in men, and it varies by several tenths of a litre between people of the same age and size. It can be measured, by having the subject breathe a known gas mixture and seeing how much it is diluted, or it can be predicted from age, sex and height by an equation fitted to a population.

Every tenth of a litre of air is two-thirds of a point of fat. The body fat inferred for the 75 kg, 18% person from the same underwater reading, against the error in the residual lung volume assumed when working it out. Each 0.1 L the residual volume is overestimated lowers the answer by 0.66 points, and each 0.1 L underestimated raises it. Ignoring the lungs altogether gives 26.6%. Residual volume predicted from age and height instead of measured is typically uncertain by several tenths of a litre (shaded), which is several points of fat.
Fig. 3 The body fat inferred for the 75 kg, 18% person from the same underwater reading, against the error in the residual lung volume assumed when working it out. Each 0.1 L the residual volume is overestimated lowers the answer by 0.66 points; each 0.1 L underestimated raises it. Ignoring the lungs gives 26.6%. Residual volume predicted from age and height is typically uncertain by several tenths of a litre (shaded), which is several points of fat.

An error of a tenth of a litre in the residual volume is an error of a tenth of a litre in the body’s volume, which for a 71-litre body is 0.14 per cent of the volume and so 0.0015 in density, and through the steep rule 0.66 of a percentage point of fat. A predicted residual volume, uncertain by a few tenths of a litre in an individual, puts a few points of uncertainty on the answer before anything else is considered. Leaving the lungs out altogether — treating the person as if their lungs were empty — would put this 18 per cent person at 26.6 per cent, an error larger than the difference between an athlete and someone of average build.

The person being weighed contributes another error that is hard to control: whether they really breathed out to residual volume. Breathing out “as far as possible” under water, bent over, is uncomfortable, and a subject who stops half a litre short has added half a litre of buoyancy that the calculation does not know about. Labs repeat the weighing several times and take the highest readings, on the reasoning that the heaviest reading is the one with the least air.

The water’s own density is a correction too

The water in the tank enters the volume through ρw\rho_w, and water’s density depends on its temperature. Tanks are kept warm, around 33 to 36 °C, so that the subject can sit still in them, and over that range water’s density is about 0.994 kilograms per litre, falling by about a part in three thousand for each degree. Using the density of water at room temperature, 0.997, for water that is actually at 35 °C shrinks the computed volume by 0.21 litres — which, through the same chain, lowers the inferred fat by 1.4 points. The anomalous way water’s density varies near its freezing point is what the lake that has to cool all the way down first followed in a lake; at bath temperatures the variation is ordinary and steady, and small, and still worth more than a point of fat.

The depth matters too. The subject’s chest sits half a metre to a metre down, where the pressure is a twentieth to a tenth of an atmosphere above the surface’s, by the rule the pressure that only knows depth set out, and the residual air is compressed in proportion: 1.3 litres measured on land becomes between 1.18 and 1.24 under water. The difference, sixty to a hundred and twenty millilitres, is worth half a point of fat or more, and a protocol that ignores it biases every subject the same way. None of these corrections is large by itself. Together with the residual volume they decide whether the answer is good to one point or to five.

Why the lungs must be emptied

Breathing out is not an optional refinement. With more air in the lungs most people would not stay under at all.

A full breath floats almost anyone. The underwater weight of a 75 kg person against their body fat, with the lungs emptied to a residual volume of 1.3 L, relaxed at 3.0 L, and full at 6.5 L. A full breath holds up 6.5 kg and floats everyone on the chart, even at 5% fat (−0.09 kg); relaxed, a body floats above about 27% fat. Breathed out as far as possible, a lean person sinks with a few kilograms to spare, and the margin shrinks with fat until, at about 39%, a body floats on its residual air alone and needs a weight to be measured at all.
Fig. 4 The underwater weight of a 75 kg person against their body fat, with the lungs emptied to a residual volume of 1.3 L, relaxed at 3.0 L, and full at 6.5 L. A full breath floats everyone on the chart, even at 5% fat; relaxed, a body floats above about 27% fat. Breathed out as far as possible, the margin shrinks with fat until, at about 39%, a body floats on its residual air alone and needs a weight to be measured at all.

A body in water is in the position of the balloon that floats at a density, which rises or sinks according to whether its average density is below or above the fluid’s, and a person can move their average density across the water’s with one breath. A full breath, six and a half litres, adds six and a half kilograms of buoyancy, more than the whole three-kilogram margin of an average body, and floats almost anyone. At the end of a normal relaxed breath the lungs hold about three litres, and a body floats above roughly 27 per cent fat. Only after a maximal exhalation does an average body sink, and then by a few kilograms. A person more than about 39 per cent fat floats even then, on residual air alone, and has to be weighed with a belt of known weight added — another correction, though a well-known one.

The same arithmetic is why swimmers float on a breath and sink on an exhalation, and why a diver’s lungs work as a buoyancy control: the depth past which it must sink showed that a compressible body loses buoyancy as it goes deeper and its air is squeezed, so that neutral buoyancy is an unstable balance. In a weighing tank the depth is small and the compression slight, but not zero, as the section above found. The pressure on the chest is also why a subject cannot simply breathe through a tube while being weighed: the snorkel that cannot be made longer found that even a few tens of centimetres of water on the chest make breathing from the surface hard work, and the weighing is done on a held exhalation for that reason.

The rule’s own assumption

Suppose the air were handled perfectly. The two-compartment rule would still be only as good as its assumption that the fat-free part of everyone has the same density.

The rule's own assumption is worth four points either way. The fat Siri's rule infers for a body that is truly 18% fat, against the actual density of everything that is not fat, which the rule takes to be 1.100 kg/L. Lean tissue lighter by 0.015 — less bone mineral, more water — reads as 23.1%; denser by 0.010 reads as 14.7%. No care with the lungs removes this: the weighing measures a density, and turning a density into a fat fraction needs a fact about the person the scale cannot see.
Fig. 5 The fat Siri’s rule infers for a body that is truly 18% fat, against the actual density of everything that is not fat, which the rule takes to be 1.100 kg/L. Lean tissue lighter by 0.015 — less bone mineral, more water — reads as 23.1%; denser by 0.010 reads as 14.7%.

It is not the same. The fat-free mass is a mixture of water at 0.99 kilograms per litre, protein at about 1.34 and bone mineral at about 3.0, and its density depends on their proportions. Children have more water and less mineral, so their lean tissue is lighter; older adults lose bone mineral; some athletes carry denser skeletons and more muscle protein; and the proportions differ between populations. A lean density of 1.085 instead of 1.100 makes the same 18 per cent person read 23 per cent; 1.110 makes them read under 15. This error does not shrink with care in the tank. It is a fact about the person that the scale cannot see, and it is why underwater weighing, once called the gold standard, is now one compartment of models that combine it with measurements of body water and bone mineral to pin down the composition of the lean part.

There is something worth noticing in the structure of this. The body that displaces two things found a floating object’s buoyancy depending on more than one fluid; here the problem is reversed, a body made of more than one substance, and a single measurement of density cannot separate more than two. One number, the density, can determine one unknown, the split between two components of known density. A third component, or an uncertain density for either, needs a second independent measurement. That is not a weakness of the tank. It is counting.

A weighing done with air instead of water

The modern alternative to the tank is a sealed chamber, an egg-shaped pod in which a person sits while the volume of air they displace is measured from pressure changes. A diaphragm between the person’s chamber and a reference chamber oscillates, and Boyle’s law relates the pressure swings to the volume of air in the chamber; subtracting from the empty chamber’s volume gives the body’s.

It replaces a hydrostatic correction with a thermodynamic one, and the correction is the one the ball that weighs how a gas stores heat used: air compressed quickly heats up and resists more than air compressed slowly. The diaphragm moves fast enough for the air in the open chamber to be compressed adiabatically, with the exponent 1.4, but air close to the skin, in the hair and clothing, and in the lungs is held at body temperature by contact and is compressed isothermally, with exponent 1. Isothermal air is more compressible, so it looks like less air — like more body. The instrument corrects by estimating the skin area and the lung volume, and the corrections, once again, are comparable to the effects being measured. The physics of the two methods is entirely different and the shape of the difficulty is the same: a body’s volume can be measured very precisely, and the gas that is with it, in it and around it decides how much the precision is worth.

What the figures cannot show

The worked numbers are for one invented person — 75 kilograms, 18 per cent fat, a residual volume of 1.3 litres, water at 35 °C — and the errors scale with them. The residual-volume figure shows the effect of a pure error in the volume; in practice errors in breathing, in the gut gas, in the water temperature and in the scale reading all add. The two-compartment densities are the conventional values, and the variation of lean density among real people is described here only by its range, not by any population’s distribution.

The deeper limit is that density alone fixes only a split between two components. Everything the figures say about fat assumes the body is fat and one other thing of known density, and it is that assumption, not the physics of buoyancy, that fails first. Archimedes’ principle holds to far better than the precision any of this needs.

Still open: how much a single number can say about a body

Multi-compartment methods combine underwater weighing or air displacement with a measurement of total body water by isotope dilution and of bone mineral by X-ray absorptiometry, and so solve for three or four components instead of two. They are more accurate and more laborious, and each added measurement brings its own assumptions — about the hydration of lean tissue, about how much of an isotope exchanges with non-water hydrogen. How small the error in a person’s fat fraction can be made, and how much of the remaining spread between methods is error and how much is real variation in what lean tissue is made of, is still being argued in the field, and the answer differs for children, for the elderly, and for athletes.

Underwater weighing gives a body’s volume by Archimedes’ principle, but a 75 kg person who is 18% fat reads only 3.12 kg under water, and the 1.3 L of air the lungs cannot expel holds up 1.29 kg of that — so every 0.1 L misjudged in the residual volume moves the inferred fat by 0.66 points, ignoring it reads 26.6%, a full breath floats almost anyone, and a breathed-out body floats on its residual air alone above about 39% fat; and even with the air perfect, turning density into fat assumes the lean part is 1.100 kg/L, where 1.085 makes the same person read 23.1%. The scale weighs the person with exquisite care, and the answer depends most on what is not the person.

Part 14 of 14

This essay is one argument about Buoyancy. The others:

The objects named here

The third axis, after the field and the reading path: the things themselves, and every essay that touches each one.

Archimedes principleDensityError propagationResidual volumeTwo-compartment model