IIF Certified Assessment
Start IQ Test
IIF Certified Assessment
Caffeine reliably lifts alertness, vigilance and reaction time, most of all in people who are tired. Evidence that it improves reasoning in rested, habitual drinkers is far weaker — and a disputed but substantial share of what it appears to add is a deficit the absence of the last cup created.
Caffeine is generally described as the most widely used psychoactive substance in the world, and it is the one most people reach for before something that matters. It does move performance on cognitive tests, reliably enough that the existence of an effect is not in dispute. What is in dispute is which parts of performance it moves, in whom, and whether the honest word for it is enhancement or repair.
The best-supported findings are about alertness rather than about thinking. Caffeine improves vigilance — the ability to keep responding accurately to a monotonous task minute after minute without your attention going offline — and it shortens reaction time, the gap between a signal appearing and your response to it. It also lifts self-rated alertness and reduces self-rated fatigue. These results appear across a large and old literature — laboratory vigilance tasks, driving studies, shift work — and they are the part nobody seriously argues about.
The conditions matter more than the substance does. The improvement is largest when alertness is already low: after a short night, late in a long shift, in the small hours. In a rested person early in the day the same measures move less, and some of them barely at all. That is the first thing worth noticing about any test-day claim, because a test is usually taken at the time of day and in the state where caffeine has least room to work.
Sorted by how well the evidence supports them, the effects fall into a fairly clear order:
Caffeine improves how impaired you feel faster and further than it improves how impaired you are.
Caffeine does not supply energy. It blocks a signal. Adenosine is a molecule that accumulates in the brain across the hours you are awake and, among other things, promotes sleepiness — it is one of the mechanisms behind the pressure to sleep that builds through a day. Caffeine occupies receptors that adenosine would otherwise bind to, so the signal is still being generated and simply is not being read. Nothing has been added. Something has been muffled.
Occupy those receptors several times a day for years and the system adjusts, which is what tolerance means in practice: the same amount does less than it once did. Stop abruptly and the adjustment is exposed. Caffeine withdrawal is a well-characterised syndrome — headache, fatigue, reduced alertness, low mood, difficulty concentrating — recognised formally enough to appear as a diagnosis in the American Psychiatric Association's manual. It begins within roughly a day of the last dose, and it degrades precisely the functions this article is about.
That creates a methodological trap a great deal of caffeine research has walked into. The standard design recruits habitual consumers, asks them to abstain overnight or longer, gives half of them caffeine and half a placebo — an inert substitute the participant cannot distinguish from the real thing — and compares performance. The problem is the comparison group. Those people are not at their normal baseline; they are in early withdrawal. An improvement measured against them may be a genuine gain above ordinary functioning, or the removal of an impairment the study itself created.
This is the withdrawal-reversal account, argued in the psychopharmacology literature for decades. Studies built to escape the trap — recruiting people who do not use caffeine at all, or extending abstinence until withdrawal resolves — generally report smaller effects than the standard design, which is what the account predicts. But they do not report nothing, particularly on vigilance and particularly in tired participants. The fair summary is that withdrawal reversal accounts for a substantial and genuinely disputed share of the measured benefit in habitual consumers, and that whatever survives on top of it is best supported in exactly the low-alertness conditions where it is least surprising.
For a daily drinker facing a test the practical instruction barely changes: do not skip the usual cup, and do not double it. What changes is the expectation attached to it. You are holding a condition constant, not acquiring an advantage.
This is where the interaction with sleep stops being academic, because caffeine is overwhelmingly what people reach for instead of sleep rather than alongside it. Our note on what sleep loss actually changes on a test sets out the pattern it is being asked to undo: attention, working memory and speed degrade sharply after a short night, while vocabulary and stored knowledge hold up comparatively well. The question here is narrower. How much of that deficit does caffeine take back?
Some of the alertness, less of the performance, and none of it durably. The clearest observation in this corner of the literature is a divergence: subjective alertness recovers more completely than objective performance does. Feeling restored and being restored are separate measurements, usually taken in the same session, and caffeine moves the first further than the second.
It also does not clear the underlying pressure. The adenosine signal is blocked, not removed, so when the caffeine is metabolised the accumulated sleep pressure is still there. An amount taken to get through an afternoon is still measurably present at bedtime for many people, which shortens or fragments the following night. Used as a substitute for sleep across several days it does not stay neutral: it borrows from the account it is paying.
The specific error worth naming is the self-assessment one. Sleep loss degrades your ability to judge your own impairment, and caffeine improves the judgement faster than it improves the performance. The implication is that a tired person who has had coffee is likely to report feeling closer to fine than their performance warrants. Walking into a timed assessment, that is the wrong direction to be wrong in.
The relationship between amount and benefit is not a straight line. Low to moderate amounts do the alerting work. Push beyond that and a second set of effects arrives — tremor, a raised heart rate, restlessness, gastrointestinal discomfort and, in susceptible people, frank anxiety — and every one of them works against a long test. A racing pulse does not help sustained concentration, and anxiety is particularly expensive because worry is thought to occupy the same limited working memory the reasoning items need. At high enough amounts the effect stops being subtle: acute caffeine intoxication is a recognised clinical entity with its own criteria, not a figure of speech.
The shape is an inverted U: benefit rises with the amount, flattens, then falls away as the unwanted effects take over. What matters is not the curve but that it turns down at a different point for every person, which is why a general figure would be useless to any particular reader even if this article were willing to print one. A test day is not the occasion to try more than usual, an unfamiliar product, or an energy drink you have never had before.
Split an assessment into two kinds of demand and the picture resolves. Most cognitive tests contain speeded components, where the items are easy and the limit is how fast you can work accurately through them, and power components, where the items are hard and the limit is whether you can solve them at all. Understanding how the sections of a test are built is worth doing before trying to influence any of them.
Caffeine has a clear route into the first kind and almost none into the second. A speeded section — symbol matching, rapid number work, easy items against a clock — is essentially a vigilance task with arithmetic attached, and vigilance is the thing caffeine most reliably supports. A hard matrix item is not. Whether you see the rule is not obviously a function of how alert you are, once you are alert enough to be looking at it properly. The mechanism runs out before the reasoning starts.
That asymmetry has a familiar shape. It is the same one in what repeated exposure to a test actually changes, where a 2007 meta-analysis in the Journal of Applied Psychology put retest gains in employment settings, averaged across studies, at roughly a quarter of a standard deviation — about four points on a mean-100, standard-deviation-15 scale, the standard deviation being the ordinary spread of scores either side of the average. Four points is not nothing: it lifts a score of 100 to a little over the 60th percentile, the percentile being the share of the reference sample scoring at or below that point. But those gains sit in the speeded sections rather than the reasoning ones. It is the pattern sleep loss produces in reverse. Three unrelated interventions, one result: the parts of a score that respond to your state on the day are the parts nearest the clock.
What follows for a test day:
A test is trying to measure you, not the morning you engineered for it. So sit a properly built assessment in your ordinary state rather than a chemically unusual one: the amount you always have, at the time you always have it, and nothing you are trying out for the first time. A result produced by an unusual morning describes an unusual morning, and it is the least useful version of the number to keep.
Find your IQ score now! →The question in the headline has a two-part answer and most of the internet reports only the first half. Caffeine improves alertness, vigilance and speed of response, most clearly in people who are tired, and that can show up in the speeded parts of a test score. It has not been shown to make anyone better at reasoning, and in a habitual consumer a contested but substantial share of what it appears to do is the undoing of what not having it did. Worth knowing not because the effect is trivial — for a night shift or a long drive it plainly is not — but because the thing most people want from it before a test is the thing it does least well. If you want to know where a score of yours stands, take it under ordinary conditions and put the number through the IQ percentile calculator; for a fresh one, our IQ test is where to get it. Neither improves with a fourth coffee.
It improves some parts of it. The reliable effects are on alertness, vigilance and speed of responding, and they are largest in people who are tired or mildly sleep-deprived — which means they can show up in the speeded sections of a test. Evidence that caffeine improves reasoning or problem-solving in rested, habitual consumers is much weaker and much less consistent, and no reputable review treats it as established.
If you drink it every day, keep the morning ordinary rather than changing it. Skipping your usual amount puts you into early withdrawal during the sitting, and taking substantially more than usual risks tremor, a raised heart rate and anxiety, all of which work against sustained concentration on a long test. If you do not normally use caffeine, a test day is a poor occasion to find out how you react to it.
Only partly, and it masks more than it repairs. Caffeine restores subjective alertness more completely than it restores measured performance, so a tired person who has had it feels closer to normal than they actually are. It also blocks the sleep signal rather than clearing it, so the pressure returns when the caffeine wears off, and a late dose can shorten the following night as well.
Partly, and how large that part is remains genuinely disputed. Most studies test habitual consumers after an enforced overnight abstinence, which means the placebo group is in early withdrawal rather than at a normal baseline. Designs that avoid this — using non-consumers, or longer abstinence — tend to find smaller effects, but not zero, especially on vigilance in tired participants.
Corrections: spotted an error? Email corrections@iqmetrics.org and we will update this story and note the change here.
A bad night does not lower intelligence. It reliably lowers the things a timed test happens to measure most closely — sustained attention, working memory and speed — while leaving vocabulary and general knowledge comparatively intact.
The correlation between them is one of the sturdier findings in cognitive psychology. How strong it is, and what it implies about training either one, is where the agreement stops.
Processing speed peaks in your twenties. Vocabulary and knowledge keep climbing into your sixties. A synthesis published in Intelligence puts overall psychological functioning at its highest between 55 and 60 — and your IQ score cannot tell you where you sit on any of those curves.
Our IIF-certified assessment reports your score with its scale, percentile and confidence range — and a breakdown of the cognitive domains behind it.
Start IQ Test →
