School start times research: biology versus the bell.

School systems set their clocks around buses, sports and parental commutes. Adolescents set theirs around a gland. Those two schedules drifted apart across the twentieth century, until American high schools were commonly ringing their first bell at 7:30 in the morning — some earlier. The students kept nodding off. The adults kept blaming the students.

This article walks through the research that reversed the blame. It runs from the sleep laboratory in the 1990s, through a series of district-level natural experiments, to a 2018 study that put activity monitors on teenagers’ wrists before and after a city moved its bell. It ends with the policy statements and the economics. Few questions in education policy have an evidence trail this tidy. It is worth seeing whole.

Two clocks in conflict

Start with the biology, because everything else hangs on it. Human sleep timing is governed by a circadian clock — an internal, roughly 24-hour rhythm that tells the body when to release melatonin, the hormone that opens the gate to sleep. In childhood that clock runs early. At puberty it shifts. Melatonin onset moves later by an hour or more, and the homeostatic pressure to sleep builds more slowly across the day (Carskadon, Wolfson, Acebo, Tzischinsky & Seifer, 1998).

The result is a teenager who genuinely cannot fall asleep at 9:30 p.m., and whose biological night now stretches toward 8 in the morning. This is not attitude. Mary Carskadon’s laboratory established the shift with hormone assays and sleep recordings, across cultures and even across other mammal species. Parents can enforce a bedtime. They cannot enforce sleep onset.

The landmark demonstration tracked students crossing a real schedule cliff: the move from a junior high that started at 8:25 to a high school that started at 7:20 (Carskadon, Wolfson, Acebo, Tzischinsky & Seifer, 1998). The students’ circadian phase did not budge to meet the new bell. Their sleep simply shrank — most were getting seven hours or less. On morning laboratory nap tests, students fell asleep with a speed normally seen in clinical sleep disorders, and roughly half of one group dropped straight into REM sleep at 8:30 a.m. — a signature of severe sleep pressure. These were ordinary students at an ordinary school. The schedule alone produced a lab result that would alarm any clinician.

Professional bodies now put the adolescent sleep need at roughly 8.5 to 9.5 hours a night (American Academy of Pediatrics, 2014). Hold that number against a 7:20 bell, a 6:15 alarm and a body that cannot initiate sleep before 11 p.m. The arithmetic does not close. That is the whole policy problem in one sentence.

What sleepy teenagers lose

The second foundation study asked what this chronic shortfall costs. Wolfson and Carskadon surveyed about 3,000 Rhode Island high-school students on their sleep habits, moods and grades (Wolfson & Carskadon, 1998). The pattern was blunt. Students reporting C grades and below also reported markedly less sleep — on the order of 25 minutes less per school night — and bedtimes roughly 40 minutes later than their A- and B-grade peers. Short and irregular sleep travelled with more depressed mood and more daytime sleepiness.

This was a survey, and the authors said so plainly. Self-reported sleep is noisy, and correlation runs both ways — struggling students may sleep badly because they are struggling. What the study established was the shape of the problem and its scale: by their own account, most adolescents were getting far less than the recommended amount, and the shortfall clustered exactly where school performance sagged.

The mechanism is no mystery. Sleep is when the brain consolidates the day’s learning into durable memory; the adult evidence on that process is reviewed in our corporate research library on sleep and learning. For adolescents the question was narrower and more practical: if the bell forces the shortfall, does moving the bell reverse it? Surveys could not answer that. Districts that changed their clocks could.

The first natural experiments

In 1996 and 1997, two Minnesota districts did something rare: they acted on the biology. Edina moved its high-school start from 7:20 to 8:30. Minneapolis followed, from 7:15 to 8:40, across seven high schools and tens of thousands of students. Kyla Wahlstrom’s team at the University of Minnesota tracked the change for years (Wahlstrom, 2002).

The central fear was that teenagers would simply stay up later and bank no extra sleep. That is not what happened. Bedtimes stayed essentially put; wake times moved later; students gained close to an hour of sleep on school nights relative to peers in early-start districts (Wahlstrom, 2002). Attendance improved. Students reported less sleepiness in class. Grades drifted upward, though the academic gains were modest and harder to pin down than the sleep gains. The predicted collapse of after-school sports and jobs never materialized.

A decade later the natural experiments multiplied. A three-state study followed more than 9,000 students across eight high schools with bells ranging from 7:30 to 8:55 (Wahlstrom et al., 2014). The dose-response pattern was striking: the later the bell, the larger the share of students reaching eight hours of sleep — roughly a third at 7:30 schools, and about two-thirds at the 8:55 school. A small controlled delay study ran the same direction. When one school moved its bell just 30 minutes, from 8:00 to 8:30, average school-night sleep rose by about 45 minutes, and the share sleeping eight hours or more jumped from 16% to over half (Owens, Belon & Moss, 2010). Students reported better mood; the school kept the change.

Seattle puts it to the test

Every study so far shared a weakness: sleep was self-reported. Teenagers estimating their own sleep are not precision instruments, and a district that just spent political capital moving its bell is primed to hear good news. The Seattle study closed that gap.

For the 2016–17 school year, Seattle Public Schools moved its secondary bell from 7:50 to 8:45. Researchers at the University of Washington put wrist activity monitors — actigraphs, devices that record movement to measure sleep objectively — on sophomores at two high schools, one cohort before the change and one after (Dunster et al., 2018). No diaries, no recall. The monitors reported a median gain of 34 minutes of nightly sleep, from about 6 hours 50 minutes to about 7 hours 24 minutes.

Three details make the result unusually persuasive. First, bedtimes did not drift later — the gained time was real sleep, exactly as the Minnesota work had predicted (Wahlstrom, 2002). Second, the students’ light-based circadian markers aligned better with their schedules, and their daytime sleepiness dropped. Third, performance moved: final grades in the biology course hosting the study were about 4.5% higher in the later-start cohort — a small gain, honestly sized, in one course at two schools (Dunster et al., 2018).

The equity finding may matter most for policy. At the school serving more economically disadvantaged students, first-period lateness and absences fell after the change; at the wealthier school they were low throughout (Dunster et al., 2018). An early bell taxes most heavily the students with long commutes, jobs and noisy households. A later bell partially refunds them.

Crashes, moods and public health

The case for later bells was never only academic. Chronic adolescent sleep loss travels with depressed mood, weight gain, heavy caffeine use and — most concretely — drowsy driving. When the American Academy of Pediatrics reviewed this evidence in 2014, it issued a formal policy statement: middle and high schools should not start before 8:30 a.m. (American Academy of Pediatrics, 2014). At the time, the large majority of American high schools started earlier than that.

The road-safety signal is the starkest. Sixteen- and seventeen-year-olds are novice drivers running on six and a half hours of sleep, often driving at the drowsiest point of their circadian day. In the three-state study, the Wyoming district that moved its bell to 8:55 saw teen crash rates fall by roughly 70% in the following year — a small district and small counts, so the exact figure deserves caution, but the direction has appeared in district after district (Wahlstrom et al., 2014).

Mood runs through the same channel. In the Rhode Island delay study, the share of students rating themselves at least somewhat unhappy or depressed fell visibly during the later-start term, and health-centre visits declined (Owens, Belon & Moss, 2010). For school leaders weighing a bell change, this is the frame the pediatricians settled on: sleep is not an amenity to be traded against bus routes. It is a public-health input with a schedule lever attached.

The economics of a later bell

Moving a bell is not free. Bus fleets run in tiers; athletics need daylight; some families need teenagers home to watch siblings. For years these costs — real, immediate, concentrated — outweighed benefits that seemed diffuse. Then economists priced the benefits.

A RAND Corporation analysis modelled a nationwide move of American middle and high schools to 8:30, feeding the sleep-to-attainment and sleep-to-crash evidence into a standard macroeconomic model (Hafner, Stepanek & Troxel, 2017). The projected gains: roughly $8.6 billion to the U.S. economy within two years, around $83 billion within a decade, and on the order of $140 billion after fifteen years. The channels are unglamorous — slightly higher graduation and attainment feeding lifetime earnings, plus fewer crash deaths among young drivers.

Against that, the model priced the costs of re-tiering buses at a few hundred dollars per student per year in the most expensive scenarios. On those numbers the reform pays for itself within a handful of years, and keeps paying (Hafner, Stepanek & Troxel, 2017). These are projections from a model, not measurements — a distinction the next section takes seriously. But they reframed the question for school boards. The expensive option is not the later bell. It is the status quo.

What the evidence doesn’t show

The start-times literature is unusually consistent, which makes it doubly important to say what it has not established. Six boundaries matter.

• Not an achievement moonshot. The reliably measured academic gains are small — about 4.5% in one Seattle course, modest and uneven grade effects in the district studies (Dunster et al., 2018). Later bells buy sleep, attendance and safety first; test scores move less, and more slowly.
• No randomized trial exists. Districts self-select into reform, and before-after designs compare different student cohorts. The Seattle study measured different sophomores in each year, not the same students twice (Dunster et al., 2018).
• Much of the base is self-report. Outside the actigraphy work, sleep and grades are usually reported by the students themselves, with the noise and flattery that implies (Wolfson & Carskadon, 1998).
• The bell cannot fix bedtimes. Biology sets the floor, but evening light, phones and workload still erode sleep after a delay. A later start is necessary for adequate teen sleep in most districts; it is not sufficient (Carskadon, Wolfson, Acebo, Tzischinsky & Seifer, 1998).
• Elementary effects are understudied. Many districts fund teen delays by starting younger children earlier. Young children’s clocks do run earlier, but the consequences of that swap have been measured far less carefully than the high-school half.
• The economics are projections. The billion-dollar figures come from a simulation fed by the attainment and crash literatures — a reasonable model, not an observed ledger (Hafner, Stepanek & Troxel, 2017).

Where the evidence stops

1. 1Not an achievement moonshot
2. 2No randomized trial exists
3. 3Much of the base is self-report
4. 4The bell cannot fix bedtimes
5. 5Elementary effects are understudied
6. 6The economics are projections

© 2026 FUTURE PROOF™

Moving the bell by the evidence

For a school board, a ministry, or a parent group pushing for change, the literature reads as a short operating manual.

Aim at 8:30 or later, and say why. The pediatric recommendation is specific, and it is anchored in physiology rather than preference (American Academy of Pediatrics, 2014). Framing the move as a health measure — like vaccination schedules or seatbelts on buses — changes the politics. It is not one interest group’s convenience against another’s.

Promise the right outcomes. Lead with sleep, attendance, mood and road safety, where effects are largest and fastest (Wahlstrom et al., 2014). Treat grade gains as a welcome secondary effect on the order of a few percent (Dunster et al., 2018). Boards that promise transformed test scores are writing cheques the literature never signed.

Handle the bedtime objection with data. The recurring fear — teens will just stay up later — has now failed to materialize in Minnesota, Rhode Island and Seattle alike (Wahlstrom, 2002). Bedtimes are pinned by biology and evening routine; wake times are pinned by the bell. Move the bell and the sleep window widens.

Budget the logistics honestly. Bus re-tiering, sports lighting and childcare shifts are the real costs, and they are bounded — the modelling prices them well below the projected benefits (Hafner, Stepanek & Troxel, 2017). Districts that failed usually lost on logistics and communication, not on the science.

Measure the change like Seattle did. Decide before the switch what counts as success: attendance, first-period lateness, crash counts, sleep from a surveyed sample — ideally with objective measurement in a subsample (Dunster et al., 2018). A bell change is a rare, clean policy experiment. It deserves a pre-registered scorecard, and the disadvantaged schools should be watched most closely, because that is where the attendance gains concentrated.