The group that followed every guideline lost. The group that broke the 2% rule won.
They consumed an extra litre per session, stayed under every threshold — and performed worse than the group that just drank when thirsty.
The bottle has hourly marks printed on the side. Or the phone buzzes every fifteen minutes: time to drink. Either way, you force a sip between sets — not because you are thirsty, but because a trainer, a poster, or an app told you thirst arrives too late.
A team led by Eric Goulet and Martin Hoffman pooled seven studies. The group: 82 athletes — mostly cyclists and runners in their twenties and thirties, 22 percent women — working out for about ninety-six minutes in mild-to-warm heat. They asked a simple question: what happens when you follow the schedule versus when you drink only when your body asks?
The scheduled group consumed 1,073 millilitres per hour. The thirst group consumed 505 — a gap of 568 millilitres per hour, or roughly an extra litre of forced fluid over a typical session.
The extra litre did not help.
The thirst group finished 0.98 percent faster. The team called that gap "likely trivial" — on purpose, because normal day-to-day swings in how you perform are already bigger. But even their most careful reading landed the same way: whatever small edge existed went to the group that drank less, not more.
A seven-study review found that athletes who drank on a schedule took in roughly an extra litre per session — and came out slightly worse than those who drank only when thirsty. The feared 2% body mass line was crossed safely. No heat or workout change shifted the result. The scheduled drinking rule had never been properly tested before this study ran the first real match.
- The thirst group lost 2.1 percent of their body mass, crossing the threshold every guideline warns about, and still came out ahead — how far past the line they went made no measurable difference.
- Five major sports science organisations built scheduled drinking protocols around a premise none of them had systematically tested until this meta-analysis ran the comparison.
- All seven studies pointed in the same direction, and no condition — temperature, exercise duration, or intensity — changed the result.
- Thirst is not a late alarm that goes off after the damage is done. The paper argues it is a signal tuned by biology that matched what the body truly needed.
The Line They Were Told to Fear
Every major hydration guideline draws the same line. Do not lose more than two percent of your body mass during exercise. The American College of Sports Medicine locked that number into their Position Stand [1].
Trainers teach it and apps enforce it. Marked bottles are built around it.
The thirst group crossed it. They finished at 2.1 percent body mass loss. The scheduled group finished at 1.0 percent — safely inside the line.
The group that broke the rule outperformed the group that followed it.
Goulet and Hoffman checked whether the amount of body mass lost predicted the result. It did not. Between 0.9 and 3.1 percent body mass loss, how dried out someone got showed no link to whether drinking to thirst came out ahead or behind.
The number on every hydration chart, the line behind every hourly bottle mark, was crossed by the group that won. How far past it they went made no measurable difference.
Who Wrote the Schedule
The two-percent rule did not come from one place. Five sports groups back it — from the American College of Sports Medicine to the Dietitians of Canada [1]. All five say athletes should not lose more than two percent of body mass during exercise.
The NATA's stance: "consuming fluids according to thirst alone does not maximize performance." The data said otherwise.
Those groups were not guessing. They followed a belief decades in the making: thirst is a bad signal, too slow to guard performance. An earlier version of the ACSM guidelines told athletes to "consume the maximal amount that can be tolerated."
But none of them had run a proper test of whether drinking to thirst actually performs worse than drinking on a schedule. They built the rule from a belief about how thirst works — and never ran the head-to-head.
The evidence is medium-strength — seven studies, eighty-two athletes. But the rule they tested was built on no systematic evidence at all.
The Argument That Produced the Data
In 2016, two of the researchers behind this review — Hoffman and Goulet — published their case in Sports Medicine [2]. Their stance: thirst is good enough for exercise. The existing guidelines were not well backed by the science.
In the same issue, Armstrong, Johnson, and Bergeron fired back [3]. Their stance: thirst is not enough. Athletes need a plan.
The journal ran both papers in the same issue, with rebuttals. Two groups of scientists staking their names on opposite answers to the same question.
Armstrong gave ground in his own case. He admitted that "no research has systematically evaluated" whether drinking to thirst works during long exercise [3]. He was asking for the test.
Three years later, Goulet and Hoffman delivered it. The 2019 review — the study this page covers — is that test. It pooled every controlled trial that qualified and ran the exact match Armstrong asked for.
The test sided with thirst.
One money detail is worth noting. Armstrong's co-author, Johnson, disclosed funding from Danone Research — the company behind Evian and Volvic [3]. A BMJ probe found that sports drink firms funded a large share of the science behind scheduled drinking rules.
That is not an accusation — it is the landscape. The industry selling bottled water funded research supporting the argument that people need to drink more water. The data, when someone ran the test, pointed the other way.
What the Evidence Actually Weighs
Seven studies and 82 athletes is not a large pool. The range where a future study would likely land crossed zero — from a 0.36 percent loss to a 2.31 percent gain for drinking to thirst. A new study could still find scheduled drinking slightly ahead.
The researchers were honest about that gap. Their "likely trivial" label was not a dodge — it was the clear call of a team that drew a line for real-world meaning before looking at the data and landed below it.
But the weight is relative. The rule these seven studies tested had never been properly tested before. Armstrong said so himself. The five groups that built drinking schedules relied on a belief about thirst being too slow — not on head-to-head tests of the two approaches.
Medium-strength evidence that cuts against a rule built on no real testing is not a weakness. It is the first real test of a rule that never had one.
A separate review by McCartney and colleagues pooled 64 trials and 643 athletes [4]. Their finding lined up from a different angle: how much you drink and when you drink it did not change the size of the benefit. A different team, a different question, a similar answer.
Goulet and Hoffman also checked whether heat, workout length, or effort level changed the result. None did. Warmer or cooler, longer or shorter, harder or easier — the finding held.
Part of the explanation may sit in a detail no drinking schedule can account for. Deuterium tracking across 5,604 people found daily water turnover ranged from one litre to more than ten — a tenfold spread driven by body composition, activity level, and climate. A fixed schedule is calibrated to an average body in average conditions. The body holding the marked bottle is neither.
The review covered cyclists and runners aged twenty to forty, working out for up to two hours in mild-to-warm heat. It does not cover ultra-long events, older athletes whose thirst cues may shift with age, or anyone starting dried out. The team said both approaches work best as partners, not rivals.
Within that scope, the extra bottle was never needed. The forced sips, the app reminders, the hourly marks — the effort bought nothing you could measure and a slightly worse outcome. The finding comes from a seven-study review, not a single trial. What it means for your next workout is yours to decide.
One question follows on its own. If drinking on a schedule during exercise did not help, does loading up on fluid before you start make a difference? A separate review set out to test just that.
For a gym session lasting about an hour — weights, some cardio, a circuit — the tested population maps directly. The review covered moderate-to-high effort exercise — the kind of intensity a regular gym session runs at. Drinking when thirsty performed at least as well as following a schedule.
For a 90-minute outdoor run in summer heat, the scope still applies. The studies averaged 28°C and no environmental condition within the tested range changed the finding.
For indoor cycling classes, same story. The review included both cycling and running, and neither sport showed a different result.
The evidence runs out past two hours. Marathon training runs, century rides, and ultra-endurance events fall outside the tested scope. So does anyone starting a session already dehydrated, and older adults whose thirst cues may shift with age. Those questions need different data.
What other research found
What this means for you
The main review studied endurance exercise — cycling and running. For those activities, the case for drinking to thirst is clearest. But the McCartney review found the proof for fluid intake helping resistance tasks was less clear: only 7 of 22 tasks across 5 of 9 trials showed a benefit.
The picture for lifting is thinner in both ways — less proof that drying out hurts your sets, less proof that drinking more helps them. The water bottle may matter less for your lifts than for your runs, but the data to say so for certain is not there yet.
The review covered exercise lasting up to about two hours. Marathon training runs, century rides, ultra stages, and long triathlon legs fall outside the tested scope.
Thirst under three or more hours of steady work in heat faces different pressures — the stomach empties more slowly, sweat rates drift, and mental fatigue may blunt the thirst signal itself. The finding that drinking to thirst works for a 90-minute run does not carry over to a 4-hour event in the same heat.
This is an honest evidence gap, not a careful hedge. The data to answer the ultra-long question has not been pooled yet.
Before you change anything
Cyclists and runners aged 20 to 40, 22 percent women, exercising at moderate-to-high intensity for one to two hours in 22 to 34°C conditions. Training levels were mixed — VO₂max averaged 65 mL/kg/min, but the pool included 20 college students and 15 recreationally active individuals alongside trained athletes.
Not tested: ultra-endurance athletes beyond two hours, older adults whose thirst regulation may change with age, team sport athletes with intermittent high-intensity bouts (soccer, basketball, rugby), anyone starting exercise already dehydrated, and military or occupational settings where voluntary access to fluid is restricted.
Small total sample — 82 athletes across seven studies, with individual studies typically running 8 to 15 participants per condition.
Lab conditions — no field studies with wind, sun, terrain variation, or real-world pacing decisions.
Limited female representation — 22 percent women in a meta-analysis covering a population roughly 50 percent female. Thirst regulation may differ across menstrual cycle phases.
No cognitive performance measures — the meta-analysis tested physical output, not decision-making or focus during exercise.
No ultra-endurance events — all studies lasted two hours or less.
No team sports with intermittent high-intensity bouts.
Medium. Seven studies, 82 total participants. Effect size: 0.98 percent (95% CI: 0.11 to 1.84). Heterogeneity was very low — I² of 6 percent, meaning all seven studies pointed in essentially the same direction. No publication bias detected (Egger’s test p = 0.28). Twenty-one null studies would need to exist in file drawers to erase the finding (fail-safe N = 21). The prediction interval crosses zero, meaning a future study could plausibly find scheduled drinking slightly ahead. The authors called the advantage "likely trivial" using a pre-registered threshold for practical significance.
If drinking on a schedule during exercise bought nothing, does loading up on fluid before you start make a difference? A separate review looked at pre-exercise fluid loading — adding extra body water before a session to get ahead of the losses. The answer was less clear-cut than anyone expected, with the payoff hinging on how long you train and how hot it gets.
The sibling question runs the other way. If how dried out you get barely predicts the damage, which costs more — reps or max? A review of 28 studies found that endurance dropped harder than strength, and the way the drying out happened mattered more than the amount.
What This Study Found
All findings from this paper, in plain language.
- Thirst-driven drinkers performed 0.98 percent better than scheduled drinkers, a gap the researchers called "likely trivial" but one that favoured drinking less, not more.
- Scheduled drinkers consumed 568 millilitres per hour more than thirst-driven drinkers — roughly an extra litre over a typical session.
- The thirst group lost 2.1 percent of their body mass, crossing the threshold every guideline warns about, while the scheduled group stayed at 1.0 percent.
- The finding contradicts the NATA position and is not in agreement with the ACSM, ISSN, Academy of Nutrition and Dietetics, and Dietitians of Canada statements on fluid replacement during exercise.
- Ambient temperature did not change the result — thirst-driven drinking performed similarly whether conditions were warmer or cooler (p = 0.37).
- Exercise duration did not change the result — shorter and longer bouts showed the same pattern (p = 0.25).
- Exercise intensity did not change the result — harder and easier sessions showed the same pattern (p = 0.55).
- Heterogeneity was very low (I² = 6 percent), meaning all seven studies pointed in essentially the same direction.
- No publication bias detected — Egger’s test p = 0.28, and 21 null studies would need to exist in file drawers to erase the finding.
- The prediction interval ranged from −0.36 to 2.31 percent, meaning a future study could plausibly find either strategy slightly ahead.
- The paper argues thirst is a built-in biological signal — a calibrated biological signal that matched what the body actually needed, not a lagging indicator.
- Running and cycling showed no significant difference in how the two drinking strategies performed (p = 0.95).