Open any pre-workout label, any supplement forum, any strength coach's intake spreadsheet, and you will find the same number: 3 to 6 milligrams of caffeine per kilogram of bodyweight, taken roughly an hour before training. For an 80 kg lifter that is 240 to 480 mg — somewhere between two and four cups of strong coffee. The dose is treated as settled fact. The mechanism behind it is treated as settled fact too. Both deserve a second look, because the confidence around caffeine ergogenics outpaces the evidence that built it.
The short verdict: caffeine reliably and meaningfully improves strength and power performance, but most of the field's stated reasons why are inherited from an older, weaker story about metabolism that the data abandoned decades ago. The dose works. The popular explanation for it is largely wrong.
Where the belief came from
The modern enthusiasm for caffeine as an athletic aid has a traceable origin, and it is narrower than the belief it spawned. In 1978, Costill, Dalsky, and Fink published a study in Medicine and Science in Sports showing that caffeine ingestion increased free fatty acid availability and improved endurance cycling time. The interpretation that stuck was elegant: caffeine mobilizes fat, the muscle burns fat instead of glycogen, glycogen lasts longer, performance improves. This is the glycogen-sparing hypothesis, and for a generation it was the explanation.
It was a cycling-endurance finding. It got generalized — by coaches, by supplement marketers, eventually by lifters — into a universal theory of caffeine ergogenics, including for activities where glycogen depletion is not the limiting factor at all. A heavy triple takes under ten seconds. There is no glycogen crisis to spare against. Yet caffeine still helps the triple. So whatever the mechanism is, the 1978 story cannot be it.
That is the pattern worth noticing: the belief has a source, and the source is thinner than the belief.
The metabolic story, retired
Later work undid the glycogen-sparing idea even on its home turf. Graham and colleagues, across the 1990s, repeatedly failed to find the predicted muscle glycogen preservation when subjects ingested caffeine, and showed that fatty acid elevation and performance benefit could be decoupled — you could get one without the other. If glycogen sparing drove the effect, blocking the fat mobilization should blunt the performance gain. It did not reliably do so.
So the metabolic explanation survives mostly as folklore. It sounds mechanistic. It is the kind of thing that feels like understanding. But the prediction it makes — that the benefit depends on substrate shifts — does not survive contact with the strength and sprint literature.
Catecholamines and the pump: real, but not the engine
Two other peripheral candidates deserve a fair hearing, because each is partly true.
The first is catecholamines — caffeine does raise circulating adrenaline. Adrenaline does plenty of useful things for a working athlete. But if catecholamine release were the primary driver, you would expect beta-blockade to abolish caffeine's ergogenic effect, and you would expect the performance benefit to track adrenaline levels tightly. Neither holds cleanly. Caffeine still works in conditions where the catecholamine response is muted.
The second is the sodium-potassium pump. Caffeine can reduce the rise in plasma potassium during exercise, which in theory delays membrane fatigue. The trouble is dose. The concentrations needed to act meaningfully on the pump, or on phosphodiesterase — another favorite mechanistic hand-wave — sit well above what a human reaches from three to six milligrams per kilogram. At drinkable doses these pathways contribute at the margins, not at the center.
The mechanism that survives: adenosine antagonism
Strip away the theories that fail their own predictions and one is left standing. Caffeine (1,3,7-trimethylxanthine) is structurally similar enough to adenosine to occupy its receptors without activating them — it sits in the seat and refuses to do the job. Adenosine accumulates during sustained neural and metabolic activity and acts, broadly, as a brake: it dampens neural excitability and contributes to the central nervous system component of fatigue. Block the brake, and the central drive to muscle is harder to suppress.
This is the explanation that fits the awkward facts. It works in efforts too short for glycogen to matter. It works when catecholamines are blunted. It predicts that caffeine should improve the perception of effort and the maintenance of voluntary activation — and that is exactly the pattern measured: lifters complete more reps, produce more force, and rate the same load as easier. The effect is in the nervous system's willingness to drive the muscle, not in the muscle's fuel supply.
A genuine but secondary peripheral mechanism survives alongside it: caffeine can enhance calcium release from the sarcoplasmic reticulum, improving the muscle's contractile response to a given signal. This is real in isolated muscle preparations. At the doses humans actually consume, it is a supporting actor, not the lead.
The mechanisms against their predictions
| Proposed mechanism | If true, we'd expect | Survives the data? |
|---|---|---|
| Glycogen sparing (fat oxidation) | Benefit tied to substrate shift; nothing in <10s efforts | No — works in sprints/lifts |
| Catecholamine release | Beta-blockade abolishes effect | Partly — effect persists |
| Na+/K+ pump, phosphodiesterase | Effect at drinkable doses | No — needs supraphysiological doses |
| Adenosine antagonism | Lower perceived effort, sustained drive, short-effort benefit | Yes |
| Calcium release | Contractile boost in muscle | Yes, but minor at real doses |
The practical numbers are modest and consistent. Meta-analyses (Grgic and colleagues, mid-2010s onward) put caffeine's effect on muscular strength at roughly 2 to 7 percent and on muscular endurance somewhat higher. That is not transformative. For a competitive lifter chasing a one-rep max, a few percent is the difference between a make and a miss. For someone training for general fitness, it is below the noise of a good night's sleep.
Who this is for, and who it isn't
Caffeine ergogenics earns its place for strength and power athletes near a competitive ceiling, where small reliable gains matter and the protocol can be tested in training first. It suits people who tolerate caffeine without it wrecking the night's sleep — and that caveat is not minor, because the sleep cost can erase the performance gain several times over.
It is oversold for novice and intermediate lifters whose progress is gated by technique, total volume, and recovery, not by a 4 percent acute boost. And individual variation is large: a subset of people are slow metabolizers (the CYP1A2 genotype is the usual suspect) who get more side effect than benefit. There is no dose that overrides poor sleep.
Evidence grade — that caffeine improves strength/power: Strong. That it does so primarily by mobilizing fat or sparing glycogen: Weak, effectively retired. That the mechanism is central adenosine antagonism: Moderate-to-Strong.
Back to the number
So return to the 3 to 6 mg/kg on every label. The dose is right — that part of the inheritance holds up. But the story usually told to justify it, the one about burning fat and saving glycogen, is a fifty-year-old endurance finding wearing the wrong outfit. What the caffeine actually does, when you take it before a heavy session, is sit in the adenosine receptor and quiet the brain's instruction to quit. The number survived. The reason behind it had to be replaced.