Non-24 sleep-wake disorder and alarms
Short answer: in non-24-hour sleep-wake rhythm disorder the internal clock is not being reset by daylight, so it free-runs slightly longer than 24 hours and the sleep window drifts later day after day until it works all the way around and starts again. A fixed daily alarm therefore lands at a different point in your rhythm every morning, which is why some weeks feel effortless and others feel impossible. The management for this is medical, through a sleep specialist, and no alarm clock of any kind changes it.
This is the clearest example on our site of a problem that looks like an alarm problem and is not one. If you are here because a person you support keeps sleeping through everything for a fortnight and then being fine for a fortnight, that cycling pattern is the thing to take to a doctor.
What is actually happening
Nearly everyone's internal circadian period runs slightly off exactly 24 hours when measured in isolation. Under normal conditions this does not matter, because morning light resets the clock every day through light-sensitive cells in the retina, and the small daily error is corrected before it accumulates.
Take that reset away and the error stops being corrected. The clock keeps its own period, so sleep timing shifts a little later each day, night after night, and the drift accumulates. The American Academy of Sleep Medicine's patient material describes the resulting pattern as rotating: sleep moves gradually from the night into the morning, then into the daytime and the afternoon, and eventually back around to the night, with the full rotation taking weeks.
That rotation is the diagnostic signature, and it also explains the most confusing thing about living with it. There are stretches where your internal night happens to coincide with the actual night, and during those stretches you sleep normally and function well. Then the alignment slides, and you spend weeks unable to sleep at night and unable to stay awake in the day. People often describe this as unpredictable insomnia alternating with unpredictable hypersomnia, when it is neither, and the cyclical structure only becomes obvious when it is written down over several weeks.
Who it affects, without overstating it
Non-24 is overwhelmingly a condition of people who are totally blind with no perception of light. The pathway that carries light information to the body clock is separate from the pathway that produces vision, and this matters more than it sounds. Some people with no useful sight still have an intact light-sensing pathway and entrain to daylight perfectly well. Others, particularly after enucleation or where the optic nerves are non-functional, have no light input at all and nothing to entrain to.
How common it is in that group is a question where the published figures do not agree, so we are not going to give you a single number. The American Academy of Sleep Medicine's patient-facing material says more than half of blind people have the disorder. Other clinical sources restrict their estimates to people with no light perception at all, which is a considerably smaller population than "blind people" as the phrase is normally used, and screening studies have produced a range of results depending on how the sample was defined and whether rhythms were measured objectively. What is not in dispute is that it is common among people with no light perception, uncommon among blind people who retain a functioning light-sensitive pathway, and rare in sighted people.
In sighted people it does occur and is described as very rare. The AASM material links it to severely limited light exposure, to some neurological and psychiatric conditions, and notes that it can emerge during treatment for delayed sleep-wake phase disorder. Rare is not the same as impossible, and a sighted person with a documented, steadily drifting sleep log deserves to be taken seriously rather than told the diagnosis does not apply to them.
Why a fixed daily alarm collides with a drifting clock
An alarm is a timer. It fires at the time you set, on the 24-hour clock everyone else uses, with no knowledge of where you are in your own cycle.
When your internal night is roughly aligned with the calendar night, a 7am alarm arrives near the natural end of your sleep and works about as well as it works for anyone. Two weeks later the same alarm arrives an hour or two after you finally fell asleep, right at your circadian low point. Two weeks after that it goes off in the middle of your biological afternoon and you were already awake.
Three things follow from this that are worth naming plainly.
Nothing about the alarm's intensity is the variable. A louder clock, a bed shaker, a strobe or a wrist band all fire at 7am regardless. Escalation changes the stimulus, not the phase you are in when it arrives.
Holding a fixed wake time does not entrain a free-running clock. For most sleep problems, a rigid wake time is the single most useful behavioural intervention, and it is genuinely good advice elsewhere on this site. Here, without a functioning light input, an alarm can force you out of bed at a fixed hour but it does not reset the underlying oscillator, so the drift continues underneath and what accumulates is sleep loss rather than alignment.
The good weeks are misleading. Because the pattern includes stretches of normality, both the person affected and the people around them tend to attribute the bad stretches to stress, effort or discipline. This is one of the main reasons the diagnosis is delayed for years.
This is a clinical problem, and the treatment is not a purchase
Diagnosis is made from the pattern over time rather than from a single night. In practice that means a detailed sleep diary kept over several weeks and actigraphy, a wrist-worn activity recorder, so the drift can be seen objectively. The AASM notes that an overnight sleep study is not required to diagnose it and is used only where another sleep disorder is also suspected. Specialist services may additionally track circadian phase markers over time to confirm a free-running rhythm and to time treatment.
Treatment aims to entrain the clock rather than to force the schedule. For blind adults this centres on timed melatonin, and there is also tasimelteon, a melatonin receptor agonist licensed specifically for non-24. For sighted people with the disorder, timed light exposure is the corresponding approach. In all cases the timing of the intervention relative to your own rhythm is the active ingredient, which is exactly why it needs someone tracking your phase rather than a rule of thumb from a website. We are not going to give you a dose, a timing or a product suggestion for any of this, and availability and funding for the licensed treatments differ considerably between countries.
The practical route is a GP or ophthalmology contact asking for referral to a sleep service, and it is worth starting the sleep diary before the appointment rather than after it, because several weeks of records is the thing that makes the pattern visible and shortens the whole process.