Split View: 6시간 수면 2주는 밤샘 이틀과 같다 — 수면 부채의 도스-반응 실험

Two Weeks on Six Hours Equals Two All-Nighters — The Dose-Response Experiment on Sleep Debt

Introduction — The Belief That "I Am a Short Sleeper"

There is always one around you: the person who says "six hours is plenty for me." Maybe it is you. And they genuinely feel fine. They may get sleepy during the day, but coffee gets them through, and work more or less runs along.

There is an experiment that tested this belief head-on. The eighth installment of Psychology, Straight from the Papers is one of the most-cited experiments in sleep science: the 2003 dose-response study by Hans Van Dongen and David Dinges. To give away the conclusion up front, the feeling that "six hours is enough" is not a fact but a symptom.

The Study Design — Dosing Sleep Like a Drug

What makes this study special is its design. It is neither a survey nor a correlation, but a controlled laboratory experiment.

Forty-eight healthy adults are isolated in the lab for 14 days. No caffeine, no alcohol, no naps. Sleep is randomly assigned like a drug dose — groups of 8 hours, 6 hours, and 4 hours a day, plus a 3-day total sleep deprivation (0 hours) group as a benchmark. Even time in bed was controlled and verified with polysomnography.

And while awake, cognition is measured every two hours. The main gauge is the psychomotor vigilance task (PVT), a simple reaction task in which you press a button as fast as you can when a number appears on the screen — boring, but powerful for exactly that reason. It does not improve with practice, it cannot be defended by motivation, and it counts precisely the moments when attention breaks off for more than 0.5 seconds (lapses). At the same time, participants were continually asked for their subjective sleepiness (the Stanford Sleepiness Scale).

The Results — Damage Is a Straight Line, Awareness a Flat One

The 14-day results come down to three sentences.

First, only the 8-hour group stayed stable. Over the whole two weeks, PVT lapses barely increased.

Second, the 6-hour and 4-hour groups worsened in a straight line, with no recovery. Contrary to the common belief that things stabilize after a day or two of adaptation, the damage piled up day after day, accumulating linearly. By day 14, the 6-hour group's lapse level was similar to someone who had stayed up an entire night, and on some measures reached the level of two nights without sleep. The 4-hour group was steeper still. In the paper's famous conversion, two weeks on six hours of sleep produces the same cognitive impairment as two consecutive all-nighters.

Third — the real discovery of this paper — subjective sleepiness stopped after only a few days. While performance worsened every day, the sleepiness participants reported rose for the first few days and then flattened out. In other words, on day 14 they were wrecked to the level of two all-nighters while feeling merely "a little tired." Chronic sleep deprivation dulls, along with performance, the very ability to detect the impairment. That is why "I am fine on six hours" is a symptom.

Understanding It Through Code — Drawing the Debt Curve

Let us reproduce the paper's key figure (lapse-accumulation curves by amount of sleep) with a simplified model. These are not the actual coefficients but a sketch meant for understanding the pattern.

# stylized sketch of Van Dongen et al. (2003), not the fitted model
DAYS = 14
BASELINE_LAPSES = 2          # well-rested PVT lapses per session

# roughly: damage accumulates in proportion to nightly sleep shortfall
DAILY_HIT = {8: 0.05, 6: 0.55, 4: 1.15}   # lapses added per day
TOTAL_DEPRIVATION_1N = 9     # ~1 night awake, in lapses over baseline
TOTAL_DEPRIVATION_2N = 14    # ~2 nights awake

for hours, hit in DAILY_HIT.items():
    lapses = [round(BASELINE_LAPSES + hit * d, 1) for d in range(1, DAYS + 1)]
    day_cross_1n = next((d + 1 for d, v in enumerate(lapses)
                         if v - BASELINE_LAPSES >= TOTAL_DEPRIVATION_1N), None)
    print(f"{hours}h sleep: day14={lapses[-1]:>5}  "
          f"reaches 1-night-awake level on day {day_cross_1n}")

# 8h sleep: day14=  2.7  reaches 1-night-awake level on day None
# 6h sleep: day14=  9.7  reaches 1-night-awake level on day None->~day 17
# 4h sleep: day14= 18.1  reaches 1-night-awake level on day 8

Look at the structure rather than the numbers. The slope for 8 hours is close to 0, while 6 hours and 4 hours are straight lines with a slope. Having a slope means time does not fix it. Debt piles up as daily shortfall times the number of days, and without repayment it never reaches equilibrium. And nowhere on this curve is your own "feeling" a reliable gauge.

Limits and Follow-ups — Reading It Fairly

Let us also note the limits of this experiment. The sample is 48 young, healthy adults, and the duration is 14 days. There are real individual differences in how much sleep people need — but genuine, genetically verified short sleepers (the DEC2 variant and the like) are estimated to be a tiny fraction of the population, an order of magnitude apart from the number of people who believe "I can get by on little sleep." Also, what this study measured is cognitive performance. The effects on emotion regulation, metabolism, and immunity have their own literature, and the direction is the same (the amygdala study covered in the body-first post is one example).

Follow-up studies also showed an asymmetry of recovery: one big catch-up sleep after several days of debt does not fully restore performance, and recovery after chronic deficit in particular is slow. Debt is easy to build and hard to repay.

What We Take Away

1. Sleep duration is not a negotiable item but a fixed line in the budget. Since this experiment, the answer to "how many hours is enough" converges, for most adults, on 7-9 hours. Six hours is not an amount you can get by on but an amount that slowly puts you into debt.

2. Do not trust your self-assessment; trust structure. The "feeling fine" is the very gauge this experiment demolished. Structures like a bedtime alarm, a caffeine cutoff line, and a fixed wake-up time are more accurate than feelings. It is the principle of systems beat goals applied to sleep.

3. Sleep before an important performance is part of the training. Pulling an all-nighter before a test, a presentation, or a deploy is a double loss — both because cramming is bad for memory and because of your cognitive performance on the day. Just as athletes manage their sleep before a match, a knowledge worker's conditioning starts in bed.

Reading Guide

• The original paper: Van Dongen, H. P. A., Maislin, G., Mullington, J. M., & Dinges, D. F. (2003). The cumulative cost of additional wakefulness: Dose-response effects on neurobehavioral functions and sleep physiology from chronic sleep restriction and total sleep deprivation. Sleep, 26(2), 117-126.
• A contemporaneous twin study: Belenky, G., et al. (2003). Patterns of performance degradation and restoration during sleep restriction and subsequent recovery: A sleep dose-response study. Journal of Sleep Research, 12(1), 1-12. — It reaches the same conclusion with 3, 5, 7, and 9-hour conditions.

Reading tip: if you look at just one thing in the original paper, make it Figure 2. The contrast between the 14-day curves of PVT lapses by condition and, beside them, the subjective-sleepiness curve that flattens out tells the whole story of the paper. The next installment is the clinical trial on exercise versus depression.