Does Glycogen Hold 3 g of Water? What the 1941 Rat Study Found

Where Does the 3 g Figure Come From?

The 3 g figure goes back more than a century, to early studies of liver glycogen (Shiose et al., 2023). A 2023 review traces the first estimate, 3 g of water per gram of liver glycogen, to Zuntz and colleagues in 1906, who worked from earlier data published by Pavy (Shiose et al., 2023). Later liver studies in rats and cats produced figures between 1.6 and 3.8 g (Shiose et al., 2023).

In 1941 the Columbia University physiologists McBride, Guest and Scott reworked one of those rat data sets, argued that the older method rested on unjustifiable assumptions and offered 2.7 g of water per gram of glycogen deposited in rat liver as their own best value (McBride et al., 1941).

The 2.7 figure is still widely repeated. A 2021 review of bodybuilders’ peak-week practices called 2.7 g the commonly referenced figure, sometimes rounded to 3 g, and traced it to studies of rat liver (Escalante et al., 2021). A 2025 position stand on nutrition for combat sports still states that 1 g of glycogen binds 2.7 g of water, although it credits the figure to a 1972 paper by Bergström and Hultman (Ricci et al., 2025). Both documents use it when discussing athletes’ glycogen and body water, far from the rat livers in which it was measured (Escalante et al., 2021; Ricci et al., 2025).

What Did the 1941 Rat Study Find?

Testing an older assumption

The 1941 study found about 2.7 g of water per gram of rat-liver glycogen, but only where the liver’s other solids stayed constant (McBride et al., 1941). The Columbia team began by checking how earlier workers had calculated the ratio. Most had assumed that the liver’s non-glycogen solids, meaning everything except water and glycogen, keep the same weight while glycogen and water rise and fall (McBride et al., 1941). Reworking MacKay and Bergman’s data on rats fasted for 48 hours and then given glucose or fructose, the authors found those solids growing by up to 32% in glucose-fed males as glycogen peaked (McBride et al., 1941).

Feeding and fasting experiments

For their own work the authors used 100-day-old male rats. Controls ate an agar diet, chosen so that their liver solids needed no correction for glycogen (McBride et al., 1941). Other groups were fasted, given limited or full rations of a carbohydrate-rich stock diet, or fed meat-scrap protein (McBride et al., 1941).

The agar-fed controls set the baseline: each gram of non-glycogen solids came with 2.43 g of water (McBride et al., 1941). Rats fasted for 36 hours and then given half their normal intake had livers that were almost 6% glycogen (McBride et al., 1941). Once the water expected with their other solids was subtracted, 1.21 g of water remained for 0.450 g of glycogen, a ratio of 1:2.7 (McBride et al., 1941).

Only these two groups came close to keeping non-glycogen solids constant (McBride et al., 1941). The authors therefore presented 1 g of glycogen to 2.7 g of water as the most accurate value their data and method could give (McBride et al., 1941). A later repeat series, in which each group’s livers were pooled, again produced 1:2.7 in rats on limited feeding (McBride et al., 1941).

Why Was the Ratio Not a Constant?

The authors’ main conclusion was that the ratio moves with the rest of the liver. As the content of non-glycogen solids rose, the apparent glycogen-to-water ratio generally fell (McBride et al., 1941). They stressed the word “apparent”, because their calculation assumed that newly deposited solids held water in the same proportion as the original ones, an assumption they judged unjustified (McBride et al., 1941).

Fat accounted for the low ratios in fully fed rats on the stock diet. Those rats in the repeat series had 0.21 g more liver fat than controls and an apparent ratio of 1:1.9; recalculated on an equal-fat basis, the figure became 1:2.6 (McBride et al., 1941).

Protein broke the pattern in a different way. Rats fed commercial dry meat-scrap powder behaved unlike those on the stock diet (McBride et al., 1941). In one protein-prefed group fasted for 24 hours, the liver held nearly 1% glycogen, yet there appeared to be no accompanying water and no change in its other solids (McBride et al., 1941).

A ratio measured in rat liver under one diet is a clue to how human muscle behaves, not a constant to multiply by.

What Have Human Studies Found Since?

Human studies point to a similar order of magnitude, but they infer glycogen water rather than measure it directly. In 1970, Olsson and Saltin studied 19 young men and proposed 3–4 g of water per gram of muscle glycogen (Shiose et al., 2023). They measured glycogen in muscle biopsies but tracked water across the whole body by tritium dilution, not in the muscle itself (Shiose et al., 2023; Homer et al., 2024).

A 2015 experiment dehydrated nine aerobically trained adults by 4.6% with 150 minutes of cycling in the heat (Fernández-Elías et al., 2015). In two trials they then took a carbohydrate drink with either 400 mL of water or enough water to replace their fluid losses, about 3.2 litres in total (Fernández-Elías et al., 2015). Muscle stored 3 g of water per gram of glycogen recovered with little fluid and 17 g with full fluid (Fernández-Elías et al., 2015). The authors concluded that each gram is stored with at least 3 g, reading the higher ratio as extra water likely not bound to glycogen (Fernández-Elías et al., 2015).

In a 2016 study, eight people followed a very high-carbohydrate diet for 72 hours after depleting exercise (Shiose et al., 2016). Muscle glycogen measured by 13C magnetic resonance spectroscopy more than doubled, from 72.7 to 169.4 mmol per kilogram of wet muscle (Shiose et al., 2016). Total body water rose by 0.9 kg, and intracellular water increased significantly in the legs but not in the arms or trunk (Shiose et al., 2016).

Not every study agrees. In 38 rodents, Sherman and colleagues found no consistent relation between muscle water and glycogen, and their data did not support the 1:2.7 ratio (Sherman et al., 1982). A 2023 narrative review nonetheless judges that body water can be assumed to rise by about 3–4 g for each gram of muscle glycogen stored in humans, despite inconsistent animal results (Shiose et al., 2023).

Why Does Weight Drop Fast When Carbohydrate Is Cut?

Water leaving with glycogen is one likely reason: a 1992 paper on very-low-calorie diets linked their quick early weight loss to it (Kreitzman et al., 1992). Kreitzman and colleagues described glycogen as stored in hydrated form, with three to four parts water and with potassium (Kreitzman et al., 1992). They warned that this distorts early estimates of weight loss and body composition (Kreitzman et al., 1992).

Tightly controlled data come from a metabolic ward. Seventeen men with overweight or obesity ate a high-carbohydrate diet for four weeks and then an isocaloric ketogenic diet with protein held steady (Hall et al., 2016). The switch brought a rapid extra loss of 1.6 kg, which the researchers judged likely to be mainly body water, since fat mass fell by only 0.2 kg over the next 15 days (Hall et al., 2016). They did not measure glycogen, so linking that water to glycogen stores is an inference. Over the full 28 days on the ketogenic diet, weight fell by 2.2 kg, of which 0.5 kg was fat (Hall et al., 2016).

The liver holds only a small share of the body’s glycogen. The 2025 combat-sports stand puts human glycogen stores at 80–100 g in the liver and 350–700 g in skeletal muscle (Ricci et al., 2025). Most glycogen, and presumably most of the water that moves with it, therefore sits in muscle.

How Strong Is the Evidence?

The evidence behind the 3 g rule is limited. Its best-known source is one set of experiments on male rats, in liver rather than muscle, reporting ratios that the authors called apparent and that shifted with diet (McBride et al., 1941). Its tables also contain minor rounding and printing slips (McBride et al., 1941), none of which changes the central 1:2.7 result on recalculation. Human estimates rest on a few small studies of 8 to 19 people, which infer glycogen water from changes in whole-body or muscle water (Shiose et al., 2023; Fernández-Elías et al., 2015; Shiose et al., 2016). For these reasons the evidence on the article’s central question, how much water each gram of glycogen holds in the human body, is rated limited.

Measurement is a large part of the problem. Muscle biopsy remains the gold standard for glycogen, and 13C magnetic resonance spectroscopy is a reliable non-invasive alternative (Shiose et al., 2023). In the human studies reviewed, water held by glycogen itself was not measured directly, so the extra body water cannot be pinned on glycogen with certainty (Shiose et al., 2023).

The practical claims built on the ratio are weaker still. The weight shifts put down to glycogen water on low-carbohydrate diets and after loading come from small or indirect studies (Shiose et al., 2016; Hall et al., 2016). A method that could tell glycogen-bound water from other water in muscle would settle much of this. Until then, roughly 3–4 g per gram of human muscle glycogen (Shiose et al., 2023), read as an estimate rather than a constant, is a fairer summary than a fixed 2.7 or 3.

Frequently asked questions

Does every gram of glycogen hold exactly 3 g of water?

No. A 1941 rat-liver value of about 2.7 g, sometimes rounded to 3 g (Escalante et al., 2021), held only when the liver’s other solids stayed constant (McBride et al., 1941). Human studies suggest roughly 3–4 g per gram of muscle glycogen (Shiose et al., 2023), and one found far more with full fluid replacement (Fernández-Elías et al., 2015).

Was the 2.7 g ratio measured in people?

No. McBride, Guest and Scott analysed only the livers of 100-day-old male laboratory rats, where glycogen is far more concentrated than in muscle (McBride et al., 1941; King et al., 2018). Even there the apparent ratio moved with diet, from no extra water to 1:3.1, so applying it to people is an extrapolation (McBride et al., 1941).

Why is weight loss so quick in the first days of a low-carb diet?

Much of it appears to be water. On a metabolic ward, men switching to a ketogenic diet quickly lost an extra 1.6 kg while fat mass hardly changed over 15 days (Hall et al., 2016). Glycogen was not measured, but a 1992 paper tied early weight change on very-low-calorie diets to glycogen water (Kreitzman et al., 1992).

How much weight does carbohydrate loading add?

There is no firm figure. A 2023 review gives the generally understood 1.0–1.5 kg, mostly water (Shiose et al., 2023). In one small study, total body water rose by 0.9 kg after three days of very high carbohydrate intake (Shiose et al., 2016). The 2016 joint sports-nutrition statement gives no expected gain (Thomas et al., 2016).

Can glycogen water protect an athlete from dehydration?

Probably very little. Laboratory and modelling work found most glycogen-associated water to be osmotically active rather than a separate store, worth only about 55 mL of plasma volume in a 2-litre sweat-loss model (King et al., 2018). Guidance still centres on avoiding fluid deficits above 2% of body weight (Thomas et al., 2016).

This article is for general information and is not medical advice. It was reviewed for accuracy by a qualified clinician; decisions about your health should be made with your own doctor.