Caffeine Doesn't Make You Smarter

We all love caffeine.

Whether it’s coffee, tea, yerba maté, guarana, guayusa, kola nut, or pure caffeine — this molecule is loved by all.

Many of us use it in the mornings to wake up… at school to stay awake… and throughout the day when we meet friends and family.

Caffeine has always been tied to social meetups, intellectual conversation, and the working day.

Coffeehouses earned that reputation early. When they first appeared in Mecca around 1511 and spread through Cairo and Ottoman Constantinople over the following century, they became known as mektep-i irfan — “schools of the wise” — because merchants, scholars, artisans and bureaucrats all mixed and argued in the same room. Rulers noticed. Coffee was banned in Mecca in 1512, Cairo in 1532 and Constantinople in 1570, not because anyone thought it was unhealthy, but because the conversations happening over it were considered politically dangerous.

Yes, only men…

Women were almost never admitted. (Perhaps those conversations weren’t quite so wise after all.)

Somewhere along the way that association hardened into a claim: that caffeine is a nootropic — a cognitive enhancer, something that makes you smarter. It gets stacked into nootropic formulas on exactly that premise.

That claim doesn’t survive contact with the evidence, and it has survived it even less well over the last decade. Caffeine is a real, powerful, genuinely useful drug. It just doesn’t do the thing it’s most often sold for.

How Does Caffeine Actually Work?

Caffeine works off the adenosine system.

Adenosine is the molecule that keeps score of how long you’ve been awake. Every hour you’re up, it accumulates. It slows the firing of neurons and dilates blood vessels — including the vessels in your brain.

Your neurons work by firing fast impulses from one region to the next. In an alert, rested person those firings happen quickly and you process information almost immediately.

As the day progresses, adenosine levels rise. When adenosine reaches a neuron it binds to a receptor and acts as a sort of “delay” on that cell. Overall, the messages fired by your neurons slow down. That slowing is what sleepiness is.

Caffeine then comes along and competitively binds those same adenosine receptors without activating them — mainly the A1 and A2A subtypes ^{1Reference 1Fredholm et al. · 1999Actions of caffeine in the brain with special reference to factors that contribute to its widespread use}. Adenosine is still there, still rising, but it can’t get to the receptor. The brake is disconnected, and your neurons carry on firing at their normal rate. Knockout work in mice narrowed the wakefulness effect specifically to A2A receptors, not A1 ^{2Reference 2Huang et al. · 2005Adenosine A2A, but not A1, receptors mediate the arousal effect of caffeine}.

This leaves you feeling alert and awake.

Two details in that sentence matter more than most people realise:

First — “normal rate.” Caffeine doesn’t push your neurons above baseline. It restores baseline by removing an inhibitor. There’s nothing in that mechanism that would make you smarter; there’s only a mechanism that stops you getting duller.

Second — the adenosine is still accumulating. Blocking the receptor doesn’t clear the molecule. Sleep pressure keeps building behind the blockade the whole time, which is exactly why the crash when caffeine wears off can feel worse than the tiredness you started with. Caffeine borrows alertness; it doesn’t create it. Only sleep clears the debt.

There’s a vascular half to this too. Because adenosine dilates cerebral vessels, blocking it does the opposite — caffeine is a cerebral vasoconstrictor. A 250 mg dose cuts global resting cerebral blood flow by roughly 22–30% ^{3Reference 3Addicott et al. · 2009The effect of daily caffeine use on cerebral blood flow: How much caffeine can we tolerate? Human Brain Mapping, 30(10), 3102–3114}. Interestingly, that constriction is blunted in habitual users, which is the first clue that your brain adapts to this drug far more thoroughly than most people assume.

The Study That Named It: Faster But Not Smarter

The cleanest test of this question is also the bluntest. In 2013, Peter Rogers and colleagues at Bristol ran 369 people — 212 medium-to-high caffeine consumers and 157 non-or-low consumers — through a double-blind trial after overnight caffeine abstinence, dosing them with caffeine (100 mg, then 150 mg) or placebo and testing self-reported state alongside computer-based performance tasks ^{4Reference 4Rogers et al. · 2013Faster but not smarter: effects of caffeine and caffeine withdrawal on alertness and performance}.

They titled the paper “Faster but not smarter.”

Two results carried the whole thing:

Caffeine improved motor performance in both groups. Reaction times got quicker regardless of who you were. That effect was real and it was general.

Caffeine improved mental alertness only in the habitual consumers. In the non-and-low consumers, any benefit from reduced sleepiness was cancelled out by increased anxiety and jitteriness. The people who “got smarter” from caffeine were the people who had been deprived of caffeine — the drug was fixing a hole it had dug itself.

The authors’ own conclusion: tolerance develops to caffeine’s anxiety-provoking effect, but tolerance also develops to its sleepiness-reducing effect — so frequent consumption fails to enhance mental alertness and mental performance ^{4Reference 4Rogers et al. · 2013Faster but not smarter: effects of caffeine and caffeine withdrawal on alertness and performance}.

Caffeine can’t make you smarter — but it can make you faster.

The Withdrawal Trap

This is the part people find hardest to accept, because it contradicts a daily lived experience: coffee obviously works, I can feel it.

You can. That feeling is real. The question is what it’s measured against.

If you drink caffeine every day, your baseline isn’t “normal” — it’s mildly withdrawn. Adenosine receptors upregulate in response to chronic blockade, so by morning you’re running a deficit. The first cup doesn’t lift you above a rested person’s alertness; it hauls you back up to it ^{4Reference 4Rogers et al. · 2013Faster but not smarter: effects of caffeine and caffeine withdrawal on alertness and performance}.

Which means the honest version of the story is:

• In a decayed cognitive state — fatigue, sleep loss, withdrawal, boredom, the 3 p.m. trough — caffeine restores something close to normal function for a few hours. That’s genuinely valuable.
• In a rested, non-dependent state — it produces a small bump in speed and attention, and no improvement at all in the things we actually mean by “smart.”

Not everyone agrees withdrawal reversal explains all of the effect; several groups have found benefits in genuine non-consumers too, and the debate is still live. But nobody in it is arguing that caffeine raises intelligence. The argument is about how much of a small, low-level effect is net gain versus repaid debt.

What the Newer Research Says

The post you’re reading was originally written in 2017 off the back of the Rogers trial. Here’s what nearly a decade of further work has done to that conclusion — and the short answer is that it’s held up unusually well.

Attention Improves — But Attention Isn’t Intelligence

The most important update is a 2025 systematic review and meta-analysis in Psychopharmacology, pooling 31 randomised, double-blind, placebo-controlled trials in 1,455 rested, healthy adults ^{6Reference 6Kløve et al. · 2025Meta-analysisA systematic review and meta-analysis of the acute effect of caffeine on attention}.

It found significant effects for caffeine on both reaction time (g = 0.28) and — this is the genuinely new part — accuracy (g = 0.27). Doses of 200 mg and above outperformed lower doses, with a linear dose-response for speed and a curved one for accuracy.

So the strict 2017 framing needs a small correction: caffeine doesn’t only make you faster, it also makes you less error-prone on attention tasks.

But look at the size of that effect. A Hedges’ g of 0.27 is small by any conventional standard — a fraction of what you’d get from a decent night’s sleep. And look at what’s being measured: sustained attention. Not reasoning. Not learning. Not comprehension. Attention is the floor cognition stands on, not the ceiling it reaches. Caffeine keeps you at the desk; it doesn’t do the thinking.

The Genetics Verdict: No Causal Cognitive Benefit

The most powerful evidence against the “coffee makes you smarter” claim came in 2018, and it’s still the study nobody in this debate has answered.

Researchers ran a Mendelian randomisation analysis across 10 European cohorts — up to 415,530 participants, including 300,760 coffee drinkers ^{7Reference 7Zhou et al. · 2018Meta-analysisHabitual coffee consumption and cognitive function: a Mendelian randomization meta-analysis in up to 415,530 participants}. Mendelian randomisation uses inherited genetic variants that predict how much coffee someone drinks as a natural, lifelong randomised trial, which sidesteps the confounding that wrecks ordinary observational nutrition research.

The result was flatly null. No association between genetically predicted coffee consumption and global cognition (β = −0.0007, p = 0.87) or memory (β = −0.001, p = 0.51) — with high consistency between studies ^{7Reference 7Zhou et al. · 2018Meta-analysisHabitual coffee consumption and cognitive function: a Mendelian randomization meta-analysis in up to 415,530 participants}.

If lifelong caffeine consumption raised cognitive function, a study of that size and design would have seen it. It didn’t.

Working Memory & Creativity: Nothing

A 2020 double-blind, placebo-controlled trial gave 80 volunteers 200 mg of caffeine or placebo and tested creative thinking ^{10Reference 10Zabelina et al. · 2020Percolating ideas: The effects of caffeine on creative thinking and problem solving}. Convergent thinking — closed problem-solving, where there’s one right answer — improved. Divergent thinking — generating novel ideas, the thing people actually mean when they say coffee makes them creative — didn’t move. Neither did working memory.

That pattern shows up repeatedly: when studies statistically separate caffeine’s effect on low-order task speed from memory function itself, the apparent working-memory benefit mostly evaporates. You’re completing the task faster, not holding more in mind ^{5Reference 5McLellan et al. · 2016ReviewA review of caffeine’s effects on cognitive, physical and occupational performance}.

What About Memory?

Memory deserves its own section, because it produced the one genuinely exciting result in this whole literature — and then produced the correction.

In 2014, Borota and colleagues published in Nature Neuroscience that 200 mg of caffeine given immediately after learning improved performance on a discrimination task 24 hours later, following an inverted-U dose-response, in caffeine-naive participants ^{8Reference 8Borota et al. · 2014Post-study caffeine administration enhances memory consolidation in humans}. Crucially, the effect was specific to consolidation, not retrieval — dosing after the material was studied. It looked like the first clean demonstration of caffeine doing something genuinely cognitive.

In 2020, Aust and Stahl attempted a direct replication and failed to find it ^{9Reference 9Aust et al. · 2020The enhancing effect of 200 mg caffeine on mnemonic discrimination is at best small}. Their non-superiority testing ruled out medium and large enhancements in both average consumers (~1,200 mg/week) and low consumers (~50 mg/week). Their conclusion: if the effect exists at all it’s substantially smaller than reported, and it may have reflected withdrawal reversal from the abstinence period rather than genuine enhanced consolidation.

That’s the same explanation Rogers landed on seven years earlier, arrived at independently. The pattern in this field is remarkably consistent: a promising cognitive effect appears, and on closer inspection it turns out to be caffeine repairing a caffeine-shaped hole.

What Caffeine Is Genuinely Good At

Now the good news — and it’s substantial. Being wrong about caffeine as a nootropic doesn’t make caffeine a weak drug. It makes it a drug that’s misfiled.

The sleep-loss finding is worth dwelling on. A 2020 systematic review and meta-analysis pooled 45 publications yielding 327 effect estimates on acute caffeine after sleep loss or restriction, and found large effects on performance across cognitive, physical, occupational and driving domains ^{11Reference 11Irwin et al. · 2020Meta-analysisEffects of acute caffeine consumption following sleep loss on cognitive, physical, occupational and driving performance: A systematic review and meta-analysis}.

That is a far bigger effect than anything caffeine does to a rested brain — and it tells you precisely what this drug is for. Caffeine is a fatigue countermeasure, not a cognitive enhancer. It’s the right tool for a night shift, a long drive, a monotonous vigilance task, or the last hour of a hard training session. It is the wrong tool for learning something difficult.

The ability to delay sleepiness makes it genuinely useful for late-night study or work, and for staying alert on long drives. The improvement in motor function makes it excellent for reaction time and for exercise. Just don’t expect the material to go in any better because you drank it.

Meet Paraxanthine, Caffeine’s Busiest Metabolite

Here’s a detail almost everyone misses: by the time caffeine is doing most of its work in your body, a lot of it isn’t caffeine anymore.

The liver enzyme CYP1A2 handles roughly 95% of caffeine metabolism, and about 84% of that output is paraxanthine (1,7-dimethylxanthine). The rest splits into theobromine — the dominant alkaloid in cacao — and theophylline.

Paraxanthine isn’t inert. It’s an adenosine antagonist in its own right, and it does things caffeine doesn’t: it dose-dependently increases intracellular calcium in skeletal muscle, which is part of why methylxanthines improve muscle performance ^{14Reference 14Hawke et al. · 2000Paraxanthine, a caffeine metabolite, dose dependently increases [Ca2+]i in skeletal muscle}. It also has a shorter half-life and greater plasma clearance than caffeine, with less of the anxiety signature.

Recently it’s been tested directly as a standalone compound. A 2021 double-blind crossover trial found 50–200 mg of paraxanthine improved reaction time over a series of trials, with more consistent effects at 100–200 mg ^{15Reference 15Xing et al. · 2021RCTDose-Response of Paraxanthine on Cognitive Function: A Double Blind, Placebo Controlled, Crossover Trial}. A 2024 trial in trained runners reported paraxanthine outperforming caffeine on cognitive measures after a 10-km run ^{16Reference 16Yoo et al. · 2024Paraxanthine provides greater improvement in cognitive function than caffeine after performing a 10-km run}.

Read those with the appropriate squint. Both were small (n = 12) and both came out of research funded by the ingredient’s manufacturer. That doesn’t make them wrong — it makes them preliminary. It’s a genuinely interesting thread, and if paraxanthine does turn out to separate the alertness from the jitters it would be a meaningful advance. It is not yet established.

Why Your Response Isn’t My Response

Part of why caffeine arguments never resolve is that people are honestly reporting different experiences, because they are having different experiences.

CYP1A2 — the enzyme above — comes in fast and slow variants, splitting the population roughly 46/54 ^{17Reference 17Nehlig · 2018Interindividual Differences in Caffeine Metabolism and Factors Driving Caffeine Consumption}. Fast metabolisers clear caffeine quickly and, in at least one trial, showed greater cognitive benefit from it than slow metabolisers. Slow metabolisers hold onto it far longer, which is where the late-afternoon coffee that wrecks your sleep comes from.

ADORA2A — the gene encoding the A2A receptor caffeine blocks — carries a variant (rs5751876) that tracks with caffeine-induced anxiety. Carriers of the sensitive genotype get more jitteriness and agitation at the same dose. If caffeine reliably makes you feel wired and awful while your friend drinks espresso at 9 p.m. and sleeps fine, this is probably why.

A 2025 review of habitual caffeine intake, genetics and cognitive performance found this literature is still limited and mixed ^{18Reference 18Kapellou et al. · 2025Habitual caffeine intake, genetics and cognitive performance} — so don’t go buying a genotype test off the back of it. But it’s a solid reason to trust your own response over anyone else’s protocol.

The Stimulant Nap

Many people are now onto the trend of stimulant naps — and unlike most caffeine folklore, this one has actual trial evidence behind it.

The idea: caffeine takes 20–30 minutes to reach meaningful effect, and an ideal power nap runs about that long. So the two can be stacked. Take the coffee, immediately lie down, and wake as the caffeine arrives.

Loughborough sleep researchers tested exactly this in driving simulators. In one study, sleepy drivers took 200 mg of caffeine only, a 15-minute nap only, or caffeine immediately followed by a nap, before a two-hour monotonous drive. The combination reduced driving incidents to around 9% of placebo levels, against 34% for caffeine alone ^{12Reference 12Reyner et al. · 1997Suppression of sleepiness in drivers: combination of caffeine with a short nap}. An earlier trial found both a short nap and 150 mg of caffeine in coffee significantly reduced driving impairment, subjective sleepiness, and EEG markers of drowsiness ^{13Reference 13Horne et al. · 1996Counteracting driver sleepiness: effects of napping, caffeine, and placebo}.

Notably, even “non-sleep dozing” — lying down without properly falling asleep — produced measurable benefit.

Note what this is, though. It’s a fatigue countermeasure stack — which is the thing caffeine is actually good at. It isn’t a cognitive enhancement protocol.

The Combination That Gets Closer

If you’re genuinely chasing improved mental performance rather than just wakefulness, the more interesting move isn’t more caffeine — it’s caffeine plus L-theanine, which is to say, drinking tea instead of coffee.

L-theanine is an amino acid found almost exclusively in Camellia sinensis. On its own it’s mildly relaxing without being sedating. Combined with caffeine, meta-analysed trials show moderate effects on attention-switching accuracy and multisensory attention in the first two hours post-dose, alongside the expected alertness bump ^{19Reference 19Camfield et al. · 2014Meta-analysisAcute effects of tea constituents L-theanine, caffeine, and epigallocatechin gallate on cognitive function and mood: a systematic review and meta-analysis}. Typical study doses cluster around 80–100 mg of each, and the classic trial in this line used 97 mg of theanine with 40 mg of caffeine ^{20Reference 20Owen et al. · 2008The combined effects of L-theanine and caffeine on cognitive performance and mood}.

The most useful part is what theanine subtracts: it takes the edge off caffeine’s anxiety and jitter without cancelling the alertness. For the ADORA2A-sensitive people above, that’s the difference between caffeine being usable and not.

This still isn’t “smarter” in the sense of raised intelligence. But attention-switching is a real cognitive operation, and improving it is more than caffeine manages alone. It’s also an argument for the plant over the isolate — tea delivers the combination as a package, the way it evolved.

The Long Game — Decades, Not Hours

There’s one more layer worth separating out, and it’s where the story turns genuinely positive for the plants themselves.

“Caffeine doesn’t make you smarter today” and “coffee drinkers have healthier brains over decades” are different claims with different evidence, and the second one is doing better than the first.

Parkinson’s disease is the strongest signal. A dose-response meta-analysis covering 901,764 participants for coffee, 344,895 for tea and 492,724 for caffeine found a non-linear protective relationship, with maximum protection around 3 cups of coffee per day (smoking-adjusted RR 0.72, 95% CI 0.65–0.81) — roughly a 28% risk reduction. Tea and caffeine showed linear relationships, with risk falling 26% per two cups/day of tea and 17% per 200 mg/day of caffeine ^{21Reference 21Qi et al. · 2014Meta-analysisDose-response meta-analysis on coffee, tea and caffeine consumption with risk of Parkinson’s disease}. A 2020 meta-analysis independently found a hazard ratio of 0.80 for regular caffeine consumers ^{22Reference 22Hong et al. · 2020Meta-analysisThe Effect of Caffeine on the Risk and Progression of Parkinson’s Disease: A Meta-Analysis}. The proposed mechanism runs straight through A2A receptor antagonism in the striatum — the same receptor system that produces the alertness, doing something quite different on a twenty-year timescale. See our Parkinson’s disease page for the wider picture.

Dementia is weaker, and I’d rather say so than oversell it. Dose-response meta-analyses suggest a shallow U-shape with the lowest risk somewhere around 1–3 cups/day, but confidence intervals for coffee versus Alzheimer’s disease specifically often cross 1.0 ^{23Reference 23Larsson et al. · 2018Meta-analysisCoffee Consumption and Risk of Dementia and Alzheimer’s Disease: A Dose-Response Meta-Analysis of Prospective Studies}. Consistent with the Mendelian randomisation result: caffeine isn’t doing much to your cognition either way.

And note that none of the long-game benefit is necessarily caffeine’s. Coffee delivers chlorogenic acids; tea delivers catechins and theanine; both carry substantial polyphenol loads with their own antioxidant and metabolic effects. Decaffeinated coffee retains much of that chemistry. The whole-plant beverage is a more interesting substance than the isolated alkaloid — which is a theme that recurs across almost everything on this site.

Getting the Dose Right

Full dosing, interactions and safety detail live on the caffeine monograph.

The Bottom Line

Caffeine is the most successful drug in human history, and it earned that position honestly. It reverses fatigue, sharpens reaction time, extends endurance, and rescues performance when you’re short on sleep — all reliably, all cheaply, all from plants people have been drinking for a thousand years.

It just doesn’t make you smarter. The trial that named the effect said so in its title, and a decade of better-powered work — a 415,530-person genetic study, a failed replication of the best memory result, a meta-analysis putting the attention effect at a small g = 0.27 — has done nothing but confirm it.

Which is fine. A tool that keeps a tired brain at its normal operating level for a few more hours is genuinely valuable. Just stop expecting it to raise the ceiling, and start using it for what it does: holding the floor.