Sleep Regularity Beats Sleep Duration

The framing

The duration narrative is incomplete

For most of the last fifty years, popular guidance about sleep has fixated on a single number. Eight hours. Seven to nine. The right total. That number was useful as a corrective to a culture that had begun treating sleep as optional, and it captured something real. The consensus position of the American Academy of Sleep Medicine and the Sleep Research Society is that adults should sleep seven or more hours on a regular basis to support optimal health, and the data behind that recommendation are solid.^[7] People who chronically sleep less than six hours fare worse than people who sleep seven or eight. People who routinely sleep more than nine hours fare worse on a different axis. The relationship between sleep duration and mortality is real, and it has been replicated across more than a million participants in pooled cohort analyses.^[1]

But notice the phrase the sleep-medicine consensus actually uses: seven or more hours on a regular basis. The second half of that sentence carried almost none of the cultural weight, because for decades the measurement tools required to take it seriously did not exist at population scale. What the duration conversation could not see was the question of when those hours were taken. A person sleeping seven hours from 11 p.m. to 6 a.m. every night of the week was being lumped into the same category as a person sleeping the same seven hours, but starting at 10 p.m. on Monday, midnight on Wednesday, 2 a.m. on Friday, and noon to 7 p.m. on Sunday. By the duration metric, both were getting seven hours. By the newer metric of sleep regularity, the first person was getting the benefits of sleep and the second was paying the metabolic cost of repeated phase shifts that the body never quite catches up with.

This is the distinction the book draws when it insists that sleep health is not one number. As The Health Protocol puts it, the discussion "cannot be reduced to a generic message about getting more hours. Time in bed matters, but it is not the whole question." Sleep health "involves timing, continuity, regularity, darkness, light exposure, symptom burden, and whether the body is given a reliable schedule on which repair can proceed." Duration is the first of those variables and the easiest to count. It is not the whole of them, and for many adults it is no longer the most leveraged one.

The duration narrative, then, is not wrong. It is incomplete. The protocol implication, for most people, is that the highest-leverage sleep improvement is rarely sleeping more. It is sleeping at the same time, more days of the week. This reframe is not a small one. It changes the question. It moves the diagnostic from a number that is easy to brag about, or to feel guilty about, to a pattern that requires looking at a whole week rather than a single night. It also moves the intervention from one that asks for more time, which most adults do not have, to one that asks for more organization, which most adults can produce with some effort. The first ask is often impossible. The second is mostly a matter of decision and design. This is the foundational material the companion article on sleep and restoration develops in full; the present article narrows to the single dimension that recent evidence has pushed to the front of the picture, regularity.

A definition that travels

What sleep regularity actually means

Sleep regularity has a clinical definition. It is not the same as sleep duration, sleep quality, or sleep hygiene, though it touches all three. Regularity is a measure of how consistent the timing of sleep onset and wake onset is from day to day across a window of observation, usually a week. A person who falls asleep at 11 p.m. plus or minus fifteen minutes every night, and wakes at 6:30 a.m. plus or minus fifteen minutes every morning, has high sleep regularity. A person whose sleep onset varies by two or three hours across the week, even if total hours land in a normal range, has low sleep regularity.

The most rigorous version of this metric is the Sleep Regularity Index, or SRI, first formalized in 2017 by a team studying the sleep-wake patterns of college students.^[5] The SRI compares the probability that a person is in the same state, asleep or awake, at the same clock time across consecutive days. It is computed from accelerometry over a multi-day window and ranges from zero, meaning the timing is effectively random from one day to the next, to one hundred, meaning the timing is perfectly identical every day. Most adults score somewhere in the sixties to eighties. The original work found that students with more irregular patterns showed measurably delayed circadian timing, with melatonin onset shifted later in the evening, and poorer academic performance, even when their total sleep duration was no shorter than their more regular peers. That last detail is the whole point of the metric: it isolates a cost that duration cannot see.

The SRI matters because it captures something that step counts and self-reported sleep hours both miss. It captures the degree to which the body's circadian system is receiving a consistent signal from one twenty-four-hour cycle to the next. The internal clock does not simply watch the day go by; in the book's phrasing, "it helps create the daily schedule on which physiology operates." That schedule governs not only sleep but body temperature, cortisol and other hormone release, digestive readiness, insulin sensitivity, immune activity, and alertness across the day. When the input is stable, the system runs predictably. When the input shifts every few days, the downstream systems shift with it, and the cost accumulates.

It is worth pausing on what consistency means at this level of physiology, because the requirement is gentler than most people fear. The body does not need a schedule precise to the minute. It needs a window. The book's own formulation is that "the body repairs best on a schedule it can anticipate," helped by "stable wake times, consistent exposure to morning light, and a reasonably predictable sleep window." A person who sleeps within the same ninety-minute envelope every night, even if the exact onset varies by twenty or thirty minutes, scores well on regularity. A person whose envelope drifts by two or three hours, even occasionally, scores poorly. The system tolerates noise. It does not tolerate phase shifts. As the book observes, irregular patterns "do not only shorten rest. They make the timing system work harder to guess what environment it is in." That distinction is the difference between a working circadian rhythm and a system that is constantly re-anchoring itself to a moving target.

One practical reason regularity has only recently emerged as a dominant predictor of mortality is that self-reports of sleep timing are notoriously unreliable. Most people remember when they tried to go to sleep, not when they actually fell asleep, and they remember the alarm time rather than the moment they truly woke. Accelerometry resolves both. The biology is not new. The measurement is. What changed is that wrist devices worn by tens of thousands of people made it possible, for the first time, to score the regularity of real sleep across a full week and then watch what happened to those people over the years that followed.

The UK Biobank finding

What the data shows

The cleanest evidence comes from the UK Biobank, a long-running prospective cohort of more than five hundred thousand adults in the United Kingdom. A subset of that cohort, roughly eighty-eight thousand participants, wore accelerometers for a week and was then followed for mortality over the next seven years. When researchers scored each person's sleep regularity and related it to all-cause mortality, the pattern was striking. The most irregular sleepers, those with the lowest SRI scores, had roughly thirty to fifty percent higher all-cause mortality risk than the most regular sleepers, even after adjusting for total sleep duration, age, sex, body mass index, socioeconomic status, smoking, alcohol, and underlying disease.^[T1]

The relationship held for cardiovascular mortality and for cancer mortality as well. Irregular sleep timing was independently associated with a higher risk of dying from each of the major causes of death, not because it changed how long people slept but because it changed when they slept and how much that timing varied from day to day. The signal survived adjustment for the variables most likely to confound it. It was not driven by people who slept extremely little or extremely much. It was driven by people whose timing was unstable across the week.

One detail of the Cribb analysis is worth lingering on. The mortality association with irregularity was graded. There was no clean cutoff above which people were safe and below which people were at risk; instead the risk rose steadily as regularity fell, across the entire distribution. A dose-response gradient of this kind is more consistent with a genuine biological relationship than with a statistical artifact, though it is important to be precise about what the study can and cannot establish. This is observational, prospective cohort evidence. It shows a strong, graded, confounder-adjusted association between irregular timing and earlier death; it does not, by itself, prove that irregularity causes the deaths. What the gradient does suggest is that the relationship is unlikely to be an accident of a few outliers, and that even modest improvements in regularity, for a person currently in the middle of the distribution, may move risk in the right direction. The intervention is not all-or-nothing.

An association this consequential invites a request for replication and for evidence from an independent cohort, and both arrived. A separate analysis of the MESA cohort in the United States, a multi-ethnic study of adults free of cardiovascular disease at baseline, used actigraphy to measure how much each person's sleep duration and sleep timing varied from night to night. Greater irregularity in both was associated with a higher rate of incident cardiovascular events over follow-up, independent of average sleep duration and of the traditional cardiovascular risk factors.^[4] Two large cohorts, on two continents, measured by wrist devices rather than questionnaires, pointing the same way: it is not only how long people sleep that tracks with hard outcomes, but how regularly.

Head to head

Regularity beats duration

In 2024, a second analysis of the UK Biobank accelerometry data, this time by Windred and colleagues, asked the question directly. Across the same cohort, which is the stronger predictor of mortality risk, sleep regularity or sleep duration? The answer was unambiguous. Sleep regularity was the stronger predictor.^[T2] When the two were entered into the same statistical model, regularity retained a substantial association with mortality while duration's association weakened. People with high regularity and average duration tended to outperform people with average regularity and ideal duration. The two are not in competition exactly; they are layered. But when forced to choose which to optimize first, the data point to regularity.

This is a meaningful reframe. For two generations, the implicit instruction has been: get more sleep. The instruction the regularity evidence supports, on the back of more rigorous measurement than was previously available, is: get the same sleep, at the same time, more days of the week. The first instruction is harder than it sounds for adults with work, children, and obligations. The second is more tractable than it sounds, because it does not ask people to find new hours. It asks them to organize the hours they already have. The book makes the same point in plainer language when it says "the quality of the schedule matters almost as much as the quantity of the opportunity."

The mortality findings are consistent with what is known about the biology. The body's metabolic machinery is calibrated to anticipate inputs at predictable clock times. Insulin sensitivity is higher in the morning than in the evening, and experimental sleep restriction measurably degrades glucose tolerance and insulin sensitivity within days.^[3] Cortisol release follows a daily curve. Body temperature drops at night to permit deep sleep and rises in the morning to permit wakefulness. Each of these rhythms is timed by the same circadian system that timing-irregular sleep disrupts. The mortality signal is the plausible integrated downstream consequence of a system that never gets to settle into a stable schedule, which is why the association, while not proof of cause, is not biologically surprising either.

It is worth being equally clear about what the regularity finding is not. It is not a claim that duration does not matter. The same UK Biobank data show the familiar J-curve between duration and mortality.^[1] People who sleep less than six hours or more than nine still fare worse than people in the middle of the range. What the regularity finding says is that, within the broad band of normal durations, the timing question is the more powerful one. For a person already sleeping between six and a half and eight hours, the marginal return from making that sleep more regular is larger than the marginal return from squeezing in another half hour. For a person sleeping four or five hours chronically, both regularity and duration need attention, and the duration deficit is acute enough to address first. Regularity is the higher-leverage lever for most adults; it is not the only one.

The multidimensional model

Six dimensions, not one

The clinical sleep-research community has moved, in parallel, toward a framework that captures this directly. The multidimensional sleep-health model, articulated in a 2025 scientific statement from the American Heart Association, treats sleep health as a construct with several dimensions rather than one.^[T3] Duration is one of them. The others are timing, regularity, continuity (how fragmented sleep is across the night), efficiency (the fraction of time in bed actually spent asleep), and daytime functioning (alertness and the absence of excessive daytime sleepiness). The statement links poorer multidimensional sleep health, not duration alone, to a range of cardiometabolic disease and its risk factors, which is exactly the relationship the mortality cohorts surface from the other direction.

The framework matters because it moves the conversation past a single number. A person sleeping seven hours, fragmented across four wakings, at variable times across the week, with twenty percent of bed time spent awake, and nodding off at the desk in the afternoon, is not getting good sleep, even though the headline figure reads seven. A person sleeping seven hours in a single consolidated block, at the same time each night, with ninety percent efficiency, and feeling alert through the day, is. The first person needs to address regularity, continuity, and efficiency before adding hours will help. The second is already there.

The multidimensional model is what allows the regularity finding to make sense. Duration is one of several inputs, and not the most leveraged one for many adults; regularity, timing, and continuity often have larger marginal returns. The protocol implication is not to ignore duration. It is to stop treating duration as a single proxy for sleep health and to organize the conversation around the dimension that is most off for a given person. This is the book's argument that "multidimensional sleep health, not only duration, has become increasingly important in how those relationships are understood," because cardiovascular strain "is influenced by how fully the body enters recovery, not only by how many hours pass."

In clinical practice, the multidimensional model also changes what improvement looks like. Progress is no longer measured by adding minutes to the headline number. It is measured by collapsing the variance across the week, reducing the number of nighttime wakings, raising the percentage of bed time actually spent asleep, and watching daytime functioning recover. Each of these can be observed without any device beyond a paper journal, and each is responsive to the same daily inputs. A person who shifts the dimension that is most off, and leaves the rest alone, usually finds that the others improve in parallel. The dimensions are not independent. They are coupled.

The weekend problem

Why weekends matter

The most common pattern of sleep irregularity in modern adult life is the weekend shift. Monday through Friday, people sleep on a schedule organized around work. Friday and Saturday nights, the schedule slides later, often by two or three hours. Saturday and Sunday mornings, wake time slides later by a similar amount. By Sunday evening, the body's internal clock is misaligned with the alarm clock that will ring on Monday, and the week begins in a hole. The book describes the same mechanism precisely: sleep opportunity "narrows on workdays and expands on weekends in a way that mimics repeated travel across time zones," and "taken together, it teaches the body not to trust its own schedule." The research literature has a name for this self-inflicted jet lag of the calendar; it is usually called social jet lag.

Social jet lag is exactly what it sounds like: the physiological cost of running two schedules, one for work and one for the weekend, and shifting between them every few days. The cost is small in any single week and meaningful across years. Night-to-night variability in sleep timing and duration has been associated, in a systematic review of more than sixty studies, with higher body mass index, weight gain, and a greater likelihood of metabolic syndrome, with the evidence for glucose and cardiovascular endpoints more mixed but pointing in the same direction.^[6] Irregular timing, measured directly by actigraphy, tracks with incident cardiovascular events in the MESA cohort,^[4] and the morning-versus-evening difference in insulin sensitivity gives a clean mechanistic reason why eating and sleeping at shifting clock times exacts a metabolic toll.^[3]

A useful practical heuristic emerges from this work, sometimes stated as a two-hour rule: across any pair of consecutive days, sleep and wake times are best kept within about two hours of each other, weekends included. A person who normally wakes at 6:30 a.m. on weekdays pays little measurable cost for sleeping until 8 a.m. on Saturday. The same person, sleeping until 10:30 a.m. on Saturday and then waking at 6:30 a.m. on Monday, has produced a four-hour phase swing the body then has to absorb, and circadian re-alignment after a shift of that size tends to take the better part of the workweek. The exact numbers are approximate and vary from person to person; the direction is not in doubt. The adjustment is invisible. The cost is not.

This is the most actionable single change for most adults pursuing better sleep. It does not require sleeping more. It requires sleeping at the same time, more days of the week, including the days when the schedule is technically free.

The cultural pressure to use weekends to recover sleep debt deserves a direct comment, because the book addresses it head-on. The notion that a person can bank four extra hours on Saturday and Sunday to repay a chronically short workweek is, by the regularity evidence, almost exactly backwards. As The Health Protocol puts it, "catch-up sleep does not fully erase the costs of irregularity when the underlying schedule remains unstable"; a run of shortened nights followed by late recovery mornings "can preserve a sense of survival while still leaving circadian timing less anchored and next day function less predictable." Borrowing from sleep is real, the book allows, "but repayment is imperfect when the borrowing becomes routine." People who feel wrecked on Monday morning often blame the weekend for being too short. The more accurate diagnosis is usually that the weekend was too misaligned. A weekend that stays within an hour or two of the weekday schedule, even at the cost of an earlier Friday and Saturday bedtime, almost always produces a better Monday than a weekend that maximizes total sleep at the expense of phase consistency.

The inputs that anchor it

What daily life can do

Sleep regularity is downstream of the daily inputs that organize the circadian system. The same practices that move the cortisol curve, the glucose curve, and the body-temperature curve also stabilize sleep timing, which is why the book treats regularity not as a discipline to be willed into place but as the natural output of a coherent day. The list is short and unsurprising, and the order matters as much as the contents.

• Morning sunlight within the first hour of waking. Five to fifteen minutes of outdoor light, ideally without sunglasses, anchors the circadian phase to actual local time. Natural morning light entrains the human clock far more reliably than indoor lighting and advances the cortisol curve toward the start of the day.^[2] The book is blunt about why this works: "darkness is biological information," and morning light is the strongest external cue the timing system receives. This is the single highest-leverage input for sleep regularity in most adults.
• A consistent wake time, including weekends. Wake time is more powerful than bedtime for anchoring the circadian system. Hold the wake time steady and a consistent sleep onset tends to follow within a few weeks, even without forcing the bedtime. This is the behavioral core of the two-hour rule.
• Dimmer light in the evening, particularly in the two hours before sleep. Bright artificial light, especially the blue-enriched spectrum of screens and overhead LEDs, delays melatonin onset and pushes sleep timing later. Lowering evening light intensity pulls sleep timing earlier and stabilizes it.
• An earlier last meal. Eating large meals close to sleep onset keeps metabolic and digestive work active into the night and blunts the temperature drop that permits deep sleep. A two- to three-hour gap between the last meal and sleep is a useful target, and it doubles as a regularity input because meal timing is one of the clock's secondary synchronizers.
• Caffeine cutoff in the early afternoon. Caffeine has a half-life of five to seven hours in most people, with effects on sleep architecture that persist longer still. Afternoon doses delay sleep onset and thin slow-wave sleep in many people who do not notice it subjectively.
• Reduced evening alcohol. Alcohol sedates at first but fragments the second half of the night and suppresses REM, which is the book's distinction between sedation and restoration made concrete. People who cut evening alcohol often see substantially more consistent sleep timing within weeks.
• Daily movement, preferably outdoors and earlier in the day. Physical activity helps anchor the circadian system, particularly in the morning or early afternoon. Late-evening high-intensity exercise can delay sleep onset in some people.

None of these requires a sleep tracker. None requires a supplement. Applied consistently for a few weeks, they tend to produce measurable improvements in sleep regularity, often well before any of the more glamorous interventions in the sleep market. The book makes the same point against the consumer framing of sleep: "the deepest determinants of restorative sleep remain more structural than commercial," and "better sleep is rarely built by one object. It is usually built by a more coherent set of conditions."

It is worth stating the order of operations plainly, because inverting it is the most common way people waste months. Begin with morning light and a consistent wake time. Give those two inputs three to four weeks. Most adults find that bedtime drifts earlier on its own, that sleep onset becomes faster, and that weekend timing stabilizes without further effort. If sleep is still inconsistent after that, move next to evening light and the last-meal window. Caffeine and alcohol are useful adjustments after that. Sleep trackers, supplements, and more elaborate interventions belong, if anywhere, after the basic anchors are in place. The cost of inverting that order is months of optimization on the wrong layer.

There is one more observation worth registering, and the book insists on it. For people who feel that consistent sleep timing is impossible because of shift work, caregiving, infants, travel, or other genuinely binding constraints, the goal is not perfection. It is harm reduction. The right response to a hard schedule "is not shame. It is nuance." Even partial regularity, on the days where it is achievable, reduces the cumulative cost of the days where it is not. The biology rewards what consistency it can get. It does not require an unattainable standard to begin paying back.

Where this lives in The Health Protocol

Mapped to the book

Sleep regularity is part of the larger material on sleep, light, and repair developed in Chapter VIII of The Health Protocol, and it is the central operational lever of Module 4 of the seminar (Sleep and Biological Restoration). Where the companion article on sleep and restoration develops what the night is for and what makes it restorative, this article isolates the single dimension the recent evidence has pushed to the front, the regularity of timing, and follows it from the UK Biobank and MESA cohorts down to the daily inputs that move it. The seminar walks through those inputs in narrated form and pairs them with the workbook practices that build them into a sustainable rhythm. For most people, the path to better sleep is not more hours. It is a more anchored schedule, and the protocol is built around making that schedule possible inside a working adult life. The downstream gain in longevity is not a side effect. It is the point.

The book treats sleep as one of the few interventions whose evidence base is now too strong to relegate to lifestyle advice, while being careful not to overstate what that evidence proves. The cohorts are observational; they establish association, not certainty of cause, and the protocol does not promise a cure. What it offers is a high-leverage, low-cost change in a direction the evidence consistently supports. The reframe from duration to regularity is not a small change in instruction. It is a change in what the protocol is asking the body to do. The body is not being asked to spend more time asleep. It is being asked to spend the same time asleep, predictably enough that the rest of its machinery can plan around it. That predictability is what restoration depends on. The hours are the input. The pattern is the work.

Time in bed is not the same as continuity. Exhaustion is not the same as recovery.

The Health Protocol · Chapter VIII · p. 149