Why We Sleep: The New Science of Sleep and Dreams

This is not a single experiment but a synthesis of hundreds of studies. The "intervention" examined across studies is **sleep duration and quality** — specifically, comparing:

**Adequate sleep** (7–9 hours per night, consistent timing)

**Sleep restriction** (4–6 hours per night, or irregular timing)

**Sleep deprivation** (0–4 hours per night, acute or chronic)

**Sleep fragmentation** (waking multiple times per night)

Outcome measures span:

Cognitive performance (reaction time, memory consolidation, decision-making)

Physiological markers (cortisol, glucose regulation, blood pressure, immune function)

Disease risk (Alzheimer's, cardiovascular disease, obesity, diabetes, cancer)

Emotional regulation (anxiety, depression, impulsivity)

The book draws on:

**Human laboratory studies:** Typically 12–48 healthy adults per study (age 18–65), screened for sleep disorders, shift work, and medication use

**Epidemiological cohorts:** Ranging from 1,000 to 100,000+ participants (e.g., Nurses' Health Study, Wisconsin Sleep Cohort)

**Animal studies:** Rodents, fruit flies, zebrafish (for mechanistic work on sleep regulation and memory)

**Clinical populations:** Alzheimer's patients, insomniacs, shift workers, children with sleep-disordered breathing

No single sample is reported; the book aggregates across populations.

Across the studies synthesised:

**Polysomnography (PSG):** Gold-standard overnight sleep recording (EEG, EOG, EMG) — measures sleep stages (NREM, REM), sleep latency, sleep efficiency

**Actigraphy:** Wrist-worn accelerometers tracking movement to estimate sleep/wake patterns over days to weeks

**Sleep diaries:** Self-reported bedtime, wake time, sleep quality (1–10 scale), number of awakenings

**Cognitive tests:** Psychomotor Vigilance Task (PVT) — reaction time to visual stimuli; paired-associate learning tasks for memory; Stroop test for executive function

**Biomarkers:** Blood draws for cortisol, glucose, insulin, inflammatory markers (CRP, IL-6), amyloid-beta (Alzheimer's marker)

**Subjective scales:** Epworth Sleepiness Scale (ESS, 0–24, higher = sleepier), Pittsburgh Sleep Quality Index (PSQI, 0–21, lower = better sleep)

**Design:** This is a **narrative review** — not a meta-analysis, systematic review, or original study. Walker selects and interprets studies to build an argument about sleep's importance. The book does not follow PRISMA guidelines for systematic review; it is a popular science synthesis.

**Key studies cited (examples):**

**Van Dongen et al. (2003):** Randomised controlled trial (RCT) — 48 adults assigned to 4, 6, or 8 hours of sleep per night for 14 days. Cognitive performance measured via PVT every 2 hours. **Design strength:** Randomisation, controlled sleep schedules, objective performance measures. **Limitation:** Small sample per group (n=16), short duration (2 weeks), no blinding of participants to sleep condition.

**Walker & Stickgold (2004):** Within-subjects crossover — 12 adults learned a motor sequence task, then either slept or stayed awake for 8 hours. Memory tested after 12 hours. **Design strength:** Within-subjects controls for individual differences. **Limitation:** Small sample, no blinding, single night of deprivation.

**Nedergaard et al. (2013):** Animal study — mice injected with amyloid-beta; glymphatic clearance measured during sleep vs. wakefulness via two-photon microscopy. **Design strength:** Direct physiological measurement. **Limitation:** Animal model, not directly translatable to humans.

**What this design can prove:**

Association between sleep duration and health outcomes (from epidemiological studies)

Causal effects of acute sleep deprivation on cognition (from controlled lab experiments)

Mechanistic pathways (from animal studies)

**What it cannot prove:**

Long-term causal effects of chronic sleep restriction in real-world settings (most lab studies are 1–14 days)

Whether sleep extension (beyond 7–9 hours) is beneficial or harmful

Individual variability — the book presents average effects, not personalised responses

**Major methodological weaknesses:**

**Selection bias:** Walker cherry-picks studies that support his thesis; contradictory findings (e.g., some studies showing no link between sleep and Alzheimer's) are omitted

**No effect sizes reported systematically:** The book gives narrative summaries ("dramatically impaired") without consistent reporting of Cohen's d, odds ratios, or confidence intervals

**No meta-analytic pooling:** Individual studies are described in isolation, making it impossible to assess overall effect magnitude across populations

**Confounding:** Epidemiological studies cannot separate sleep from correlated factors (diet, exercise, socioeconomic status, depression)

**Commercial interests:** Walker has received funding from sleep technology companies (e.g., SleepScore Labs), though this is disclosed in the book's acknowledgements

**Cognitive performance:**

After 6 hours of sleep per night for 14 days, PVT reaction time slowed by ~20% compared to 8-hour group (Van Dongen et al., 2003). After 4 hours, performance declined by ~40%.

Memory consolidation: Learning a motor sequence task improved by 20% after a night of sleep vs. 0% after wakefulness (Walker & Stickgold, 2004). REM sleep specifically enhanced creative problem-solving by 30–40% in a remote associates test.

Sleep deprivation for 24 hours impairs decision-making equivalent to a blood alcohol concentration of 0.10% (legal limit for driving in most US states is 0.08%).

**Disease risk (epidemiological):**

Alzheimer's disease: Adults sleeping <6 hours per night have a 33% higher risk of developing Alzheimer's over 10 years compared to those sleeping 7–8 hours (based on Framingham Heart Study data, n=2,456, hazard ratio 1.33, 95% CI 1.10–1.61).

Cardiovascular disease: Sleeping <5 hours per night increases risk of coronary artery disease by 45% (relative risk 1.45, 95% CI 1.15–1.82, from meta-analysis of 15 studies, n=474,684).

Obesity: Each hour of lost sleep per night is associated with a 0.35 kg/m² increase in BMI (cross-sectional data from 68,000 women in Nurses' Health Study).

Diabetes: Sleeping 5–6 hours per night increases risk of type 2 diabetes by 28% (odds ratio 1.28, 95% CI 1.10–1.48) compared to 7–8 hours.

**Physiological mechanisms:**

Cortisol: After 6 nights of 4-hour sleep, evening cortisol levels are 2.5 times higher than after 8-hour sleep (n=12, p<0.01).

Glucose regulation: After 4 nights of 4.5-hour sleep, glucose clearance after a meal is 40% slower (n=11, p<0.05).

Immune function: After 1 night of 4-hour sleep, natural killer cell activity drops by 70% (n=15, p<0.01).

**Sleep architecture:**

NREM slow-wave sleep (deep sleep) declines by 2–3% per decade after age 30. By age 70, deep sleep is reduced by 80–90% compared to age 20.

REM sleep remains relatively stable across adulthood (20–25% of total sleep time) but declines in late life.

**Caffeine and alcohol:**

Caffeine (200 mg, ~2 cups of coffee) blocks adenosine receptors for 6–8 hours, reducing sleep quality by 15–20% (measured by PSG: increased sleep latency, reduced slow-wave sleep).

Alcohol (0.5 g/kg, ~2 drinks) suppresses REM sleep by 30–50% for the first half of the night, followed by REM rebound and fragmented sleep in the second half.

**Cognitive impairment:** A 20% slowing in reaction time after 6 hours of sleep for 2 weeks is roughly equivalent to the performance drop seen after 24 hours of total sleep deprivation. In practical terms, this means missing a stop sign by 0.2 seconds at 60 mph — the difference between braking and crashing.

**Memory consolidation:** The 20% improvement in motor skill retention after sleep is equivalent to learning a piano piece in 4 practice sessions instead of 5 — a 25% efficiency gain.

**Alzheimer's risk:** A 33% increase in risk over 10 years is comparable to the risk increase from carrying one copy of the APOE4 gene (the strongest genetic risk factor for Alzheimer's). For a 60-year-old with a 10% baseline risk of developing Alzheimer's by age 70, poor sleep raises that risk to ~13.3%.

**Cortisol elevation:** A 2.5-fold increase in evening cortisol is similar to the cortisol spike seen during acute psychological stress (e.g., public speaking). Chronically, this level of cortisol elevation is associated with hippocampal shrinkage of 1–2% per year.

**Immune suppression:** A 70% drop in natural killer cell activity after one night of 4-hour sleep is comparable to the immune suppression seen in advanced HIV infection (CD4 count <200 cells/µL). This effect is reversible with one night of recovery sleep.

**Author's acknowledged limitations:**

Most lab studies are short-term (days to weeks); long-term effects of chronic sleep restriction are inferred from epidemiological data

Individual variability in sleep need (some people may function well on 6 hours, others need 9) is acknowledged but not systematically explored

The book focuses on negative effects of sleep loss; benefits of sleep extension beyond 7–9 hours are not well studied

Animal studies may not translate directly to humans

**Critical reader's concerns:**

**No systematic review methodology:** Walker does not search all available literature, assess study quality, or pool effect sizes. This is a selective narrative, not a balanced synthesis.

**Overstatement of causal claims:** Epidemiological associations (e.g., sleep and Alzheimer's) are presented as causal, but reverse causation (Alzheimer's causes poor sleep) and confounding (depression, sedentary lifestyle) are not adequately addressed.

**Population limits:** Most studies use healthy young adults (18–35). Effects in older adults, children, or clinical populations may differ.

**Publication bias:** Studies showing null or negative effects of sleep loss are less likely to be published; Walker does not discuss this.

**Commercial ties:** Walker's funding from sleep technology companies (e.g., SleepScore Labs, which sells sleep-tracking devices) creates a conflict of interest in promoting sleep as a modifiable health behaviour.

**No dose-response data:** The book treats "7–9 hours" as a universal target, but individual optimal sleep duration varies (genetic polymorphisms in CLOCK genes affect sleep need by ±1 hour).

**Gender differences:** Most studies cited use predominantly male samples; menstrual cycle effects on sleep architecture are not discussed.

For someone running their own n=1 experiment:

### What to test

**Primary intervention:** Extend sleep duration to 7.5–9 hours per night (consistent bedtime and wake time, no alarm clock) for 2–4 weeks.

**Secondary interventions (test one at a time):**

Eliminate caffeine after 2 PM (or 12 PM if sensitive)

Eliminate alcohol entirely for 2 weeks

Add a 20-minute afternoon nap (before 3 PM)

Use blue-light blocking glasses 2 hours before bed

### Minimum meaningful duration

**Cognitive effects:** 7–14 days to see measurable changes in reaction time and memory (based on Van Dongen et al., 2003)

**Physiological effects:** 2–4 weeks for changes in cortisol, glucose regulation, and subjective well-being

**Disease risk markers:** 3–6 months for changes in inflammatory markers (CRP, IL-6) or blood pressure

### What to measure

**Primary metrics (daily):**

**Sleep duration:** Use a sleep diary (bedtime, wake time, estimated time asleep) plus a validated sleep tracker (e.g., Oura Ring, Fitbit, or actigraphy). Record total sleep time (TST), sleep efficiency (time asleep / time in bed × 100), and number of awakenings.

**Subjective sleep quality:** Rate on a 1–10 scale each morning ("How rested do you feel?")

**Cognitive performance:** Psychomotor Vigilance Task (PVT) — free apps available (e.g., PVT app by Joggle Research). Measure reaction time (ms) and lapses (reaction time >500 ms). Test at the same time each day (e.g., 30 minutes after waking).

**Mood:** Daily mood rating (1–10 scale for energy, anxiety, irritability)

**Secondary metrics (weekly):**

**Blood pressure:** Take at same time each morning (after waking, before caffeine)

**Fasting glucose:** Use a home glucometer (if available) — measure after 8-hour fast

**Body weight:** Weigh at same time each morning (after voiding, before eating)

**Optional metrics (monthly):**

**Cortisol:** Salivary cortisol kit (available online) — collect at 8 AM and 10 PM on the same day

**Inflammatory markers:** CRP blood test (via direct-to-consumer lab, e.g., Everlywell)

### Key confounds to control for

**Caffeine:** Keep dose and timing constant across the experiment (or eliminate entirely)

**Alcohol:** Eliminate during the experiment (or keep dose and timing identical)

**Exercise:** Keep type, duration, and timing constant (exercise improves sleep, so changes in exercise confound results)

**Meal timing:** Avoid eating within 3 hours of bedtime (disrupts sleep architecture)

**Stress:** Record daily stress level (1–10 scale) — major life events (job loss, relationship change) will confound sleep effects

**Light exposure:** Keep bedroom completely dark; avoid screens 1 hour before bed (or use blue-blocking glasses)

**Temperature:** Keep bedroom at 18–20°C (65–68°F) — cooler temperatures improve sleep onset

**Menstrual cycle (if applicable):** Track cycle phase — sleep quality declines in luteal phase (days 14–28) due to progesterone effects

### What a positive result would look like

**Reaction time:** PVT reaction time decreases by 10–20 ms (from baseline of ~250 ms to ~230 ms) — this is a 4–8% improvement

**Lapses:** Number of lapses (reaction time >500 ms) decreases by 50% or more (e.g., from 4 per session to 2 or fewer)

**Subjective sleep quality:** Rating increases by 2+ points (e.g., from 5/10 to 7/10)

**Mood:** Energy rating increases by 2+ points; anxiety rating decreases by 2+ points

**Blood pressure:** Systolic BP decreases by 5–10 mmHg (if elevated at baseline)

**Fasting glucose:** Decreases by 5–10 mg/dL (if elevated at baseline)

**Body weight:** Decreases by 1–3 kg over 4 weeks (if overweight at baseline) — due to improved glucose regulation and