Why I wake up at 3am:
what changes in the second half of the night
You do not wake at 3 any more often than at 1 or at 5. You wake briefly several times a night, at different hours, and you remember almost only the 3am one. The reason is not your liver and it is not a message: it is the way the second half of the night is built, with nearly all your deep sleep already spent beneath you and cortisol on the way up.
You do not wake at 3.
You remember the 3am waking
The first correction is also the most useful: brief wakings are not a fault in the night, they are part of it. You go through four to six sleep cycles, each of about 90 minutes, and the transitions between them are not glued to one another. Most wakings are too short to reach memory. You do not count them because you do not see them.
There is a measurement that puts this very concretely. In 1997, Michael Bonnet and Donna Arand recorded the electroencephalogram and the electrocardiogram of twelve healthy adults, synchronised over five minute epochs, to see what the heart does around wakings. The result is counterintuitive: heart rate starts to rise by at least ten beats before the waking is visible on the EEG trace.
The feeling is “I woke up suddenly, heart pounding, so something must have woken me”. The measurement shows the opposite: the activation was already under way, and the waking came after it. No noise, no dream and no message woke you. You woke because your night had just reached the part where it wakes easily.
From here, the question changes. Not “why do I wake at 3”, but “why do I remember the 3am waking and not the others”. And the second question really does have an answer.
The second half of the night
is built to be thin
A night is not a straight line. It is a sequence that repeats, N1, N2, N3, N2, REM, four or five times, with a full cycle of 90 to 110 minutes. What matters is that the sequence does not repeat identically: the proportions shift from one cycle to the next, always in the same direction.
Deep sleep, stage N3, is the stage you are hardest to wake from. It has the highest arousal threshold of them all, so high that in some people even noises above 100 decibels do not produce a full waking. The trouble is that N3 is spent early. As the night goes on, REM episodes get longer and the time spent in deep sleep falls. The first REM episode of the night lasts around ten minutes. The last one can reach an hour.
Cortisol lies over the top of all this. It is not the stress hormone in the sense advertising uses, it is the hormone that gets you moving in the morning, and its secretion follows a clear daily rhythm: its lowest concentration at night, a climb over the last hours of sleep and a peak around waking. In other words, in the second part of the night your body has already begun getting ready to wake, without asking you.
The model that explains why 3 in particular
Alexander Borbély proposed a two process model in the eighties, reassessed by him and his colleagues in the Journal of Sleep Research in 2016 and used ever since as the skeleton of almost all sleep research. A homeostatic process, S, is the sleep pressure that builds while you are awake and is spent while you sleep. A circadian process, C, is the internal clock signal, independent of how much you have slept.
Lay both of them over one night. At 3am you have already slept four hours, so most of the sleep pressure is spent. The circadian waking signal, meanwhile, has not arrived yet. The result is a window in which nothing pushes you very hard down and nothing pulls you very hard up. A waking that would have lasted forty seconds at midnight finds, at 3, the least to hold on to.
N3 is the stage with the highest arousal threshold of them all. It is also the stage that gets spent first: as the night goes on, the time spent in deep sleep falls. In practice, you spend your strongest protection against wakings in the first two cycles. By 3am you have almost none of it left. This is also where the most notable age related change sits: the progressive fall in time spent in slow wave sleep.
About 75% of sleep is NREM, most of it in stage N2, which lengthens with every cycle until it reaches some 45% of the total. REM goes the other way in episode length: the first is short, around ten minutes, the last can run to an hour. Wakings out of REM come with sympathetic activation, which is exactly that sensation of a heart beating faster. And REM is also where you usually wake from spontaneously in the morning.
Sleep pressure is the only thing on the list that falls because you sleep, not because time passes. By 3am you have spent most of it, and the circadian waking signal has not come yet. It is the window with the least force in either direction. That is why a waking there can stretch out without you having insomnia and without anything at all having happened.
The organ clock,
checked on paper
If you have ever looked the answer up in Romanian, you found it: the liver is said to be at work between 1 and 3 in the morning, so the waking would be a signal. The explanation comes from the organ clock, a scheme in traditional Chinese medicine that splits the day into twelve intervals of two hours each, one organ to each. It is old, it is coherent on its own terms and it deserves to be treated as such, not laughed at.
But an explanation is judged by what it predicts. And here the problem shows: the organ clock gives the same hour to everybody, whatever time they went to bed. Someone who falls asleep at 22:00 and someone who falls asleep at 02:00 should both wake at the same fixed hour. Sleep architecture predicts something else, namely that the window moves along with bedtime, because it depends on how many cycles you have been through, not on what the clock says. The second prediction is the one that matches what happens to people.
What the official list actually says
We downloaded the annex to Regulation (EU) 432/2012 in full, the complete list of health claims authorised in the European Union other than those referring to the reduction of disease risk. Almost 101,000 characters of text. Then we searched it word by word.
| Search in the annex to Regulation 432/2012 | Occurrences | Where |
|---|---|---|
| “sleep” | 1 | in the claim about melatonin and the reduction of time taken to fall asleep |
| “night” | 0 | nowhere |
| “awake” | 0 | nowhere |
| “circadian” | 0 | nowhere |
| “liver” | 1 | in the claim about choline and the maintenance of normal liver function |
The only mention of the liver in the whole list belongs to choline, it refers to the maintenance of normal liver function and it requires at least 82.5 mg of choline per 100 g, per 100 ml or per portion. It has nothing to do with any hour of the night. Which does not prove the organ clock wrong, because a regulation is not an experiment. It does prove something more practical: nobody is allowed to sell you 3am as a symptom, and if somebody does sell it to you, they are doing it outside the rules.
Five boring explanations
that predict more than any clock
None of them is spectacular. All of them can be checked in two weeks, on yourself, without any equipment.
1. Alcohol, the most common and the most overlooked
In 2013, a review published in Alcoholism: Clinical and Experimental Research went through every known study on the effect of alcohol on sleep in healthy volunteers. The pattern is the same at every dose, not just the large ones: alcohol shortens the time it takes you to fall asleep, makes the first half of the night more compact, then increases sleep fragmentation in the second half. The first REM episode is significantly delayed, and slow wave sleep increases in the first part of the night.
Put another way: the glass at dinner buys you the first half of the night and bills you for the second. If you wake at 3 exactly on the nights you have been drinking and not on the others, you need no other explanation. Two weeks with no alcohol after 18:00 will answer the question better than any article.
2. The bladder, the reason nobody wants to write down
A study of 6,000 adults over 40, representative of the population of Poland, published in 2021, found that 73.7% report at least one episode of getting out of bed per night, and 36.1% at least two. Women are affected more often than men, and prevalence rises with age. The interesting part: only between 22 and 29% of those affected have ever sought treatment.
If the trip to the bathroom is the first thing that happens after you wake, the order may be the reverse of how you read it.
3. Age, which changes the architecture without breaking anything
A meta-analysis in Sleep, 2004, pooled 65 studies and 3,577 people between 5 and 102 years old, to describe how sleep changes across a lifetime. In adults, total sleep time, sleep efficiency, the percentage of slow wave sleep, the percentage of REM and REM latency all fall with age, while time awake after falling asleep rises. Nothing has broken. The ceiling has moved. The same waking that went unnoticed at 25 is felt at 45.
4. What you do in the first five minutes after waking
This is the only variable you control directly. A forty second waking and a ninety minute one begin identically. What separates them is what you add: the clock checked, the hours left counted, tomorrow's list. All three work as a reason to worry introduced exactly into the window where sleep pressure is already spent and there is nothing left to pull you back. If this stretch arrived along with a change of rhythm or with going back to work, we have written separately about what happens to sleep when you go back to the office.
5. Breathing
Sleep apnoea produces airway collapse precisely in the deeper stages, which cuts the time spent in N3 and in REM and fragments the night. It is the only cause on this list that habits will not fix and the only one that needs a diagnosis.
Go to a doctor, not to a pharmacy, if the wakings come together with loud snoring, with breathing pauses noticed by someone sleeping next to you, or with daytime sleepiness that affects your work or your driving. The same if you get up several times a night for the bathroom, if you take medication for blood pressure, depression or diabetes, or if you are pregnant or breastfeeding. No food supplement is the answer to any of these situations.
Alcohol in the evening, bedtime, afternoon caffeine, the temperature of the room, the trips to the bathroom. You change one at a time and write it down. If the waking goes away, you have found it.
Sleep apnoea, medication, thyroid conditions, chronic pain, mood disorders. These call for an appointment, and putting them off with supplements is the most expensive way to lose time.
When it is insomnia
and when it is just a night
The figure that circulates in the Romanian press is “one in three people”. It comes from a real place, but it is the broadest of four definitions that do not say the same thing. Maurice Ohayon's review in Sleep Medicine Reviews, 2002, gathered over fifty epidemiological studies and set them side by side.
| The definition used | How much of the population | What it means for you |
|---|---|---|
| At least one insomnia symptom, by the DSM-IV criteria | about a third | Almost everybody, at some point. It is not a diagnosis. |
| Symptoms plus consequences during the day | 9 to 15% | This is where it starts to matter. The day is the test, not the night. |
| Dissatisfaction with your own sleep | 8 to 18% | Subjective, but useful: it is the reason people look for help. |
| A diagnosis of insomnia by DSM-IV | 6% | The small figure, the one that really does call for treatment. |
Prevalence is higher in women at all four definitions. And the line that separates them is not the number of wakings, it is what happens the next day. A broken night you come out of functional is not an illness. Three months of broken nights after which concentration and mood suffer are something else, and they deserve a conversation with a doctor, not a shelf of supplements. If your problem is, in fact, that your bedtime has slipped by two hours, the article on moving your sleep schedule back is more useful than this one.
The idea comes from the organ clock, a scheme in traditional Chinese medicine that splits the day into twelve intervals of two hours each. It has two practical problems. First: it predicts the same hour whatever time you went to bed, so it should wake you just the same whether you fall asleep at 22:00 or at 02:00. Second: it adds nothing on top of sleep architecture, which explains the same thing with numbers. And in the complete list of health claims authorised in the European Union, the only one that mentions the liver is the one about choline, and it says nothing about any hour of the night.
This explanation is repeated in almost every article and it sounds plausible, but we did not find a primary study showing that nocturnal hypoglycaemia is a frequent cause of 3am wakings in healthy people. So we are not selling it to you as truth or as myth. If you have diabetes or take medication that lowers blood sugar, it is a good question for your doctor, not for a blog article.
Half true, and that exact half is the half you pay for. The review published in Alcoholism: Clinical and Experimental Research in 2013 went through every study on healthy volunteers and found the same pattern at every dose: alcohol shortens the time it takes you to fall asleep and makes the first half of the night more compact, then increases fragmentation in the second half. The first REM episode is significantly delayed at every dose. In other words, you buy the first half with the second.
It depends entirely on the definition, and there are four definitions. Maurice Ohayon's review in Sleep Medicine Reviews, 2002, which gathered over fifty general population studies, sets them side by side: about a third of the population has at least one insomnia symptom; between 9 and 15% also have consequences during the day; between 8 and 18% are dissatisfied with their sleep; 6% meet the diagnostic criteria. The one in three figure, the one repeated everywhere, is the broadest of the four and says the least.
Melatonin has exactly two authorised health claims in the European Union. One is about jet lag, the other about the reduction of time taken to fall asleep. Both of them stop at the moment you have fallen asleep. We downloaded the annex to Regulation 432/2012 in full and searched it word by word: “sleep” appears exactly once in the whole list, in the claim about falling asleep, while “night”, “awake” and “circadian” appear zero times.
What melatonin can
and cannot do
We sell a product with melatonin in it, so take this chapter with the suspicion it deserves. We are writing it anyway, because it is the part where the numbers contradict the advertising most often.
In the European Union, melatonin has exactly two authorised health claims. Not three, not twenty, two. And both of them stop the second you have fallen asleep.
| The authorised claim | The exact condition | What it does NOT say |
|---|---|---|
| Melatonin contributes to the reduction of time taken to fall asleep | 1 mg per portion, shortly before bedtime | Nothing about what happens after you have fallen asleep |
| Melatonin contributes to the alleviation of subjective feelings of jet lag | 0.5 mg minimum, on the first day of travel and the few days after | Nothing about ordinary nights, at home |
That is the whole list. There is no authorised claim at all about waking in the night, about “uninterrupted sleep” or about “deep sleep”. If you fall asleep in ten minutes and wake at 3, melatonin taken in the evening is answering a problem you do not have.
And even for what it is allowed to do, the effect is small
The most cited meta-analysis on the subject, published in PLoS ONE in 2013, pooled 19 randomised trials and 1,683 participants with primary sleep disorders. Melatonin shortened the time taken to fall asleep by 7.06 minutes, with a 95% confidence interval between 4.37 and 9.75 minutes. It increased total sleep time by 8.25 minutes, confidence interval between 1.74 and 14.75. Sleep quality improved, with an effect size of 0.22. The authors describe their own result as modest.
Seven minutes. They are real, they are measured, and they are exactly that much. If you have read a double digit percentage about melatonin somewhere, that is not this figure, and this figure is the one that went through peer review.
What is in our box, so you know exactly what you are buying: the portion is two gummies, with 5 mg of melatonin, 200 mg of chamomile extract 25:1 and 200 mg of lavender extract 10:1. For the dosing and the timing, our melatonin guide covers it at length, and the article on what not to take together covers the schedule.
What readers ask us
most often
Main sources: Regulation (EU) No 432/2012, annex, consolidated text. Bonnet MH, Arand DL, Electroencephalography and Clinical Neurophysiology, 1997. Borbély AA, Daan S, Wirz-Justice A, Deboer T, Journal of Sleep Research, 2016. Ebrahim IO, Shapiro CM, Williams AJ, Fenwick PB, Alcoholism: Clinical and Experimental Research, 2013. Ohayon MM, Sleep Medicine Reviews, 2002. Ohayon MM, Carskadon MA, Guilleminault C, Vitiello MV, Sleep, 2004. Przydacz M, Chlosta P, Healthcare, 2021. Ferracioli-Oda E, Qawasmi A, Bloch MH, PLoS ONE, 2013. StatPearls, Physiology, Sleep Stages and Physiology, Cortisol, NCBI Bookshelf.
Comments (0)
There are no comments for this article. Be the first one to leave a message!