Some of the water in your body was air this morning. The O₂ you breathe in ends up as water inside your cells, and the CO₂ you breathe out carries oxygen away from that water. Bone, teeth and eggshell grow from body water, so whatever breathing writes there can outlast an animal by millions of years. This post follows one breath in and one breath out, with a few experiments on my own lungs along the way.

The bigger question, how respiration and the use of body water show up in the triple oxygen isotopes of animals, is the subject of ongoing work by Cloe Knudsen, a graduate student at the UNM Center for Stable Isotopes (CSI).

Air has a fingerprint

Oxygen comes in three stable isotopes, ¹⁶O, ¹⁷O and ¹⁸O. Nearly every process on Earth sorts them by mass, so ¹⁷O shifts about half as much as ¹⁸O and the two move in lockstep. Air O₂ breaks that rule slightly. Reactions among ozone, O₂ and CO₂ in the stratosphere leave air O₂ a little short of ¹⁷O for its ¹⁸O (Luz and Barkan 2005). Rain and rivers carry no such deficit, so an animal that makes part of its body water from air O₂ inherits a trace of the fingerprint. That trace, written Δ′¹⁷O, is how researchers estimate how much of an animal’s water comes from breathing (Whiteman et al. 2019; Sabat et al. 2021).

Air O₂ is also unusually rich in ¹⁸O, and breathing is part of the reason. Lane and Dole (1956) let bacteria, mushrooms, spinach leaves, crabs, frogs and a person breathe from closed reservoirs of air and found that every one of them used ¹⁶O first. Summed over the planet, that taste for the light isotope helps keep the air’s oxygen heavy.

Holding my breath

To watch that sorting in one body, I held my breath. During a breath hold the lungs keep drawing O₂ from the same parcel of air, so a series of holds traces the uptake step by step. In September 2023 I exhaled into sample tubes after holds of up to 80 seconds. By the longest holds my lungs had used about half the O₂ I had breathed in. The O₂ left behind had grown steadily heavier in both ¹⁷O and ¹⁸O, as Epstein and Zeiri (1988) found for volunteers at rest. It moved away from air along a straight line whose slope matches the one Luz and Barkan (2005) measured for human breath (Interactive 1).

Measuring that oxygen is harder than it sounds, because air’s argon comes along for the ride. Air holds about one argon atom for every 22 O₂ molecules. The two are hard to separate, because they are nearly the same size and boil within three degrees of each other. Left in the sample, argon can distort the small ¹⁷O signal in the mass spectrometer, and ³⁶Ar lands right on top of the rarest form of O₂, ¹⁸O¹⁸O. So the O₂ is first passed through a gas chromatograph column chilled to about −80 °C, cold enough to pull the two apart (Ash et al. 2020).

Interactive 1. My breath holds in ¹⁷O and ¹⁸O. Drag the slope to see how well it fits, and open the data table for every sample. Open the module full screen.

The bottleneck in your lungs

An O₂ molecule’s trip from air to blood has three legs. It diffuses across the thin wall of an air sac, moves through blood plasma into a red blood cell, and binds to hemoglobin. Lighter molecules diffuse a little faster, and a chemical step can also favor one isotope, so each leg can sort. The breath records whichever leg is the bottleneck, because the slowest step controls how much of its sorting gets expressed (Epstein and Zeiri 1988; Zanconato et al. 1992).

Physiologists frame this as diffusion limitation against perfusion limitation. Is uptake held back by how fast O₂ can cross the wall, or by how fast blood flows past to carry it away? At rest the standard answer is blood flow, with diffusion starting to limit uptake in hard exercise or thin air (as summarized by Zanconato et al. 1992). The isotopes add a twist. Zanconato and colleagues had volunteers pedal a cycle ergometer and found the sorting fell by as much as 40% at high intensity, then recovered within two minutes of stopping. People with anemia show the same drop (Epstein and Zeiri 1988). Zanconato and colleagues read this as the hemoglobin reaction, which sorts less than diffusion does, becoming the slow step as blood rushes through the lung.

The ¹⁷O to ¹⁸O slope of the sorting cannot say which step it is. My measurements sit between the slopes expected for diffusion and for binding, so either one, or a mix, fits. O₂ molecules carrying two heavy atoms can tell them apart. Diffusion depends on the mass of the whole molecule, while binding to heme tugs on the bond between its two atoms. So the two steps push the abundance of ¹⁸O¹⁸O in opposite directions (Ash et al. 2020). For holds like mine the predicted difference is more than ten times the precision of today’s best measurements. That experiment is waiting to be run, and it needs the argon gone first.

Interactive 2. How much O₂ each hold used, and how heavy the O₂ left behind became. Move z, the strength of the sorting, across the resting range Epstein and Zeiri measured, and see what the lungs took up. Open the module full screen.

Breathing out changes the water too

Every breath out carries CO₂, and on its way out that CO₂ trades oxygen atoms with body water. Carbonic anhydrase, one of the fastest enzymes known, makes the trade so quick that exhaled CO₂ leaves in isotopic equilibrium with the water (Epstein and Zeiri 1988). CO₂ in equilibrium with water is rich in ¹⁸O (Brenninkmeijer et al. 1983), so each exhaled molecule takes heavy oxygen away and leaves body water lighter.

What that does to the ¹⁷O fingerprint hangs on one number, θ, which sets how far ¹⁷O moves for each step in ¹⁸O as CO₂ and water trade oxygen. Laboratory measurements put it near 0.522 (Hofmann et al. 2012; Barkan and Luz 2012), a calculation puts it a little higher (Cao and Liu 2011), and my model had been running at a higher working value still. The choice decides how much of the air fingerprint the exhaled CO₂ cancels.

My own breath offered a check. I worked back from the CO₂ I exhale to the body water it came from, and compared that with what my model predicts for me. The two agree only near the laboratory value (Interactive 3). It is one sample, and the laser instrument’s reference scale may itself rest on an assumed θ, so it is a reason to measure rather than an answer. The measurement that settles it is breath CO₂ alongside saliva or urine from people drinking different waters.

Interactive 3. Body water worked back from my breath CO₂, against my model’s prediction for me, at any θ. The lines cross where my breath and the model agree. Open the module full screen.

How much of you is air?

The strength of the air fingerprint in body water comes down to one share, the fraction of all the oxygen entering the body that arrives as O₂ from the air. Everything else, drinking water, water in food and the oxygen bound in food, carries no fingerprint. My body water model tracks every one of these fluxes in both ¹⁷O and ¹⁸O (Interactive 4), and that share sorts animals in a satisfying way.

A resting person takes about a fifth of their oxygen from the air. Merriam’s kangaroo rat, the champion water conserver, can live on dry seeds without drinking. Its water comes from its own metabolism, from humid burrow air condensing in its lungs, and from seeds it stores in the burrow to soak up moisture before eating them (Nagy 2004). In my model it takes about half its oxygen from the air. A ground squirrel hibernating on its fat reserves takes nearly four-fifths.

Fish sit at the other extreme. Water pours across their gills so fast that breathing leaves no mark. A dogfish swaps more than half its body water through its gills every hour (Giacomin et al. 2017), which makes fish a built-in zero for any air breather that shares their water.

Lung design matters less than you might think. Birds push air through their lungs in one direction, while mammals breathe in and out of dead-end sacs, and the two extract different shares of each breath. In my model that difference barely registers in body water, unless the two kinds of lung sort isotopes differently, and I know of no measurement of that in a bird. Fuel matters more. Burning fat sends less of the body’s oxygen out as CO₂ than burning carbohydrate does, so diet and fasting leave a readable mark.

Interactive 4. My body water mass balance. Pick an animal, then change its water budget, fuel, body temperature and breathing to see how each term moves body water in ¹⁸O and ¹⁷O. Open the module full screen.

Breath in stone

Teeth and bones lock in the oxygen of the body water they grow from, which is why all of this matters for fossils. Marine mammals and reptiles breathe air but share their water with fish, so the offset between an air breather and the fish it swam with isolates breathing. Feng et al. (2022) found a clear air fingerprint in the teeth of marine mammals and, more puzzling, faint traces in sharks. At the rate a dogfish flushes its body water, my model cannot make that signal by breathing. Either the sharks pick up the fingerprint another way, or they show where the model is incomplete.

What comes next

Three measurements would turn this from a model into a record. Breath CO₂ and body water from the same people would pin down θ in a living body. Breath holds at rest, in exercise and in thin or oxygen-rich air, measured for ¹⁸O¹⁸O too, would find the bottleneck in the lung. Closed-chamber measurements on a rodent, a songbird, a lizard and a seal would show whether other lungs sort isotopes the way ours do.

Credits and code

The breath measurements are my own, and so are any errors here. The topic at large is the subject of ongoing work by Cloe Knudsen at CSI. The interactives run my body water mass balance, so every number here can be pushed on. Corrections and better numbers are welcome by email.

References and data

1. Luz B., Barkan E. (2005) The isotopic ratios ¹⁷O/¹⁶O and ¹⁸O/¹⁶O in molecular oxygen and their significance in biogeochemistry. Geochimica et Cosmochimica Acta 69, 1099–1110. Article
2. Whiteman J.P., Sharp Z.D., Gerson A.R., Newsome S.D. (2019) Relating Δ¹⁷O values of animal body water to exogenous water inputs and metabolism. BioScience 69, 658–668. Article
3. Sabat P., Newsome S.D., Pinochet S., Nespolo R., Sanchez-Hernandez J.C., Maldonado K., Gerson A.R., Sharp Z.D., Whiteman J.P. (2021) Triple oxygen isotope measurements (Δ′¹⁷O) of body water reflect water intake, metabolism, and δ¹⁸O of ingested water in passerines. Frontiers in Physiology 12, 710026. Article
4. Lane G.A., Dole M. (1956) Fractionation of oxygen isotopes during respiration. Science 123, 574–576. Article
5. Epstein S., Zeiri L. (1988) Oxygen and carbon isotopic compositions of gases respired by humans. Proceedings of the National Academy of Sciences 85, 1727–1731. Article
6. Ash J.L., Hu H., Yeung L.Y. (2020) What fractionates oxygen isotopes during respiration? Insights from multiple isotopologue measurements and theory. ACS Earth and Space Chemistry 4, 50–66. Article
7. Zanconato S., Cooper D.M., Armon Y., Epstein S. (1992) Effect of increased metabolic rate on oxygen isotopic fractionation. Respiration Physiology 89, 319–327. Article
8. Brenninkmeijer C.A.M., Kraft P., Mook W.G. (1983) Oxygen isotope fractionation between CO₂ and H₂O. Chemical Geology (Isotope Geoscience) 41, 181–190. Article
9. Hofmann M.E.G., Horváth B., Pack A. (2012) Triple oxygen isotope equilibrium fractionation between carbon dioxide and water. Earth and Planetary Science Letters 319–320, 159–164. Article
10. Barkan E., Luz B. (2012) High-precision measurements of ¹⁷O/¹⁶O and ¹⁸O/¹⁶O ratios in CO₂. Rapid Communications in Mass Spectrometry 26, 2733–2738. Article
11. Cao X., Liu Y. (2011) Equilibrium mass-dependent fractionation relationships for triple oxygen isotopes. Geochimica et Cosmochimica Acta 75, 7435–7445. Article
12. Nagy K.A. (2004) Water economy of free-living desert animals. International Congress Series 1275, 291–297. Article
13. Giacomin M., Schulte P.M., Wood C.M. (2017) Differential effects of temperature on oxygen consumption and branchial fluxes of urea, ammonia, and water in the dogfish shark (Squalus acanthias suckleyi). Physiological and Biochemical Zoology 90, 627–637. Article
14. Feng D., Tütken T., Löffler N., Tröster G., Pack A. (2022) Isotopically anomalous metabolic oxygen in marine vertebrates as physiology and atmospheric proxy. Geochimica et Cosmochimica Acta 328, 85–102. Article
15. Kohn M.J. (1996) Predicting animal δ¹⁸O, accounting for diet and physiological adaptation. Geochimica et Cosmochimica Acta 60, 4811–4829. Article
16. Steele Z.T., Caceres K., David Z.A., Shollenberger L.M., Gerson A.R., Newsome S.D., Whiteman J.P. (2025) Validating a novel capability of assessing pathways of animal water gain and loss. Royal Society Open Science 12, 241942. Article
17. Data. Every breath sample, with its date, hold time, number of cycles and standard errors, is in the data table inside Interactive 1. The archetype parameters and the sources behind them are listed under Interactive 4.