The case is unusually clean: one well-controlled trial showed over an hour of total sleep lost to caffeine taken six hours before bed, and a 24-trial meta-analysis put the safe gap at closer to nine hours for a normal coffee. Free, reversible night to night, no side effects, no purchases. The catch is the first week — afternoon coffee is a real habit — but the sleep dividend lands fast, and the morning grogginess the cup was treating turns out to be partly what the cup was causing.
A tiredness chemical called adenosine builds up in your brain the longer you're awake. It binds to receptors that read sleep pressure and tell you it's time to wind down. Caffeine fits into those same receptors and blocks them — it doesn't add energy, it hides the tiredness Reichert et al. 2022. The brain keeps building pressure underneath; you just can't feel it. About thirty minutes after the cup, the block is on. The question is when it comes off.
For most adults, half the caffeine is still in the bloodstream five hours later, a quarter at ten hours, and caffeine's main breakdown product does the same blocking job IOM 2001; Reichert et al. 2022. So a two-cup coffee at 3 pm still has ~100 mg working at 8 pm and ~50 mg at 1 am — enough to keep some of the receptors occupied, which is enough to flatten the deep sleep that delivers most of the night's restoration. Even morning caffeine measurably reduces deep-sleep brain-wave activity that night Landolt et al. 1995; the afternoon dose is just louder.
The hour you didn't know you were losing
The cleanest test of this is the one most people haven't heard of. Twelve healthy sleepers, each given 400 mg of caffeine (about two strong cups) at bedtime, three hours before, and six hours before — versus placebo, double-blind, with brain-wave recording overnight. The six-hour-before-bed dose cost over an hour of total sleep. And the participants did not report sleeping badly. The disruption was on the machine; the felt experience was "fine."
The 2023 meta-analysis pulled together 24 controlled trials and found the same pattern at scale: caffeine takes about 45 minutes off total sleep, drops sleep efficiency by about 7%, and — the one that matters most — cuts deep sleep by about 11 minutes a night. Deep sleep is where the brain consolidates memory and the body does its physical repair; you lose it first when caffeine is still on board.
And tolerance doesn't save you. A controlled 10-day study of habitual coffee drinkers found that regular caffeine intake delayed the brain's REM-sleep timing and dulled overall sleep quality even when participants felt normal Weibel et al. 2021. You adapt to feeling awake. You don't adapt to sleeping well.
Why the morning grogginess keeps getting worse
The mechanism that makes this entry matter is a feedback loop. Late caffeine cuts deep sleep. Less deep sleep means more adenosine pressure unloaded onto tomorrow. Tomorrow you wake up flat, and the first coffee feels like rescue — because it is, partly, masking the pressure your last late coffee created. By 3 pm the morning hit has worn off; you reach for the second cup, which costs you tonight's deep sleep, which makes tomorrow morning worse. The dose creeps. The morning gets harder. You start needing caffeine to feel like a person Roehrs & Roth 2008.
What gives this stakes section its weight is how invisible it is. Six months in, you don't notice — your mornings have always been like this. Two years in, your partner has stopped asking why you're so tired because it's just who you are now. You catch every cold going around. You can't read at night because you fall asleep on the page, but then you wake at 3 am and can't get back under. The version of this that's a sleep problem and the version of this that's a caffeine problem are the same version. The diagnostic is whether moving the cutoff back fixes the morning.
The cutoff
Pick a target bedtime. Count back at least eight hours. That's your cutoff. For an 11 pm bedtime, no caffeine after 3 pm. For a 10 pm bedtime, no caffeine after 2 pm. The eight-hour number is the cleanest fit to the meta-analytic data for a normal-sized coffee Gardiner et al. 2023; if you're drinking strong coffee or an energy drink or a pre-workout serving, stretch to ten or twelve.
How do you know if you metabolize caffeine slowly? Without a genetic test, the felt-experience signal is reliable: if a 2 pm coffee leaves you wired at 10 pm, you metabolize slowly and your cutoff is earlier than the average person's. If you barely notice the cup at 4 pm, you metabolize fast — but the sleep architecture damage is still there, on the polysomnograph, whether you feel it or not Drake 2013.
The first week is the hard part. The 2 pm urge is real. Bridges that work for most people: a brisk walk outside, a glass of cold water, a 20-minute nap if the schedule allows, switching to decaf so the cup ritual stays. After about a week, the urge softens — because the morning grogginess the cup was masking starts to lift, and the cycle breaks.
"I sleep fine after my afternoon coffee"
The most-repeated objection is also the best-studied. In the Drake trial, participants drank caffeine six hours before bed, lost over an hour of measured sleep, and rated their sleep quality as fine Drake 2013. The disruption is invisible. Felt sleep quality and actual sleep architecture are different signals; habitual late-caffeine users learn to ignore the second one and trust the first. The way you find out the cup at 4 pm is hurting you is by removing it for two weeks and seeing what mornings start to feel like — not by introspecting on whether tonight's sleep felt okay.
"I've built up tolerance." Tolerance develops to the alertness effect — that's why the morning cup doesn't feel like much anymore. Tolerance does not develop to the sleep-disruption effect; chronic daily caffeine continues to suppress deep sleep and delay REM in habitual users at the same dose Weibel et al. 2021; Clark & Landolt 2017. You stop feeling the wakefulness; the receptors stay occupied.
"Morning coffee is consequence-free." Mostly true, with an asterisk. Even 200 mg taken at 7 am produced a measurable reduction in deep-sleep brain-wave activity that night Landolt et al. 1995. The effect is small enough that most people don't need to worry about morning caffeine — but it's why "drink less overall" beats "drink it all in the morning" for the sensitive sleeper.
Where this goes wrong in practice
- The clock-anchored cutoff that ignores bedtime. A 3 pm cutoff is fine for an 11 pm bed; it's useless if you go to bed at 9 pm during the week and 1 am on weekends. Anchor on bedtime minus eight hours, not on a fixed clock time.
- Forgetting the hidden caffeine. Black and green tea, cola, dark chocolate, some "decaf" coffee (7–15 mg per cup), pre-workout supplements, "energy" gum, certain headache medications. A "no coffee after 3" rule that lets in two iced teas at 5 pm is the same dose at the same time.
- Doing the cutoff without doing the dose. Front-loading 400 mg in the morning to "make the cutoff work" still leaves enough caffeine on board at midnight to flatten deep sleep. The cutoff and the total daily dose are two separate dials.
- Treating metabolizer status as fixed. If you quit smoking, your clearance halves and your old cutoff is no longer aggressive enough. Pregnancy roughly doubles half-life. Oral contraceptives extend it. Recalibrate when the life situation changes.
- Giving up at day three. The first three days of cutting afternoon caffeine feel worse, not better — withdrawal symptoms peak then resolve Roehrs & Roth 2008. The sleep benefit shows up in week one. Quitting on day three is quitting before the data comes in.
What changes — and when
Night one. The brain-wave measurements rebound fast; deep sleep starts returning the first night you clear the receptors before bed Landolt et al. 1995. You won't feel it as "good sleep" yet — the withdrawal headache and the afternoon flatness mask the win.
Week one. The 2 pm urge softens. Mornings stop hurting the way they did. The thing you used to call "I'm not a morning person" turns out to have been "I'm drinking caffeine too late." People start telling you you look less tired before you notice it yourself.
Month one. The 3 am wake-up — the one you'd accepted as a feature of getting older — doesn't show up most weeks. The evening between dinner and bed becomes long again; you have hours back you didn't know you were spending dragging yourself across the afternoon. The hobby you'd been meaning to start is suddenly thinkable, because the bandwidth is there.
Year one. The morning brain is the sharpest brain. The first two hours of the workday — the ones you used to wait out — start producing your best work. The version of you that walks into hard conversations is rested; the rationing you'd quietly accepted of the afternoon, the evening, and the morning is over.
If a hard cutoff is hard
Two softer moves are still useful. Switch the late dose to a lower-caffeine form. An afternoon black tea (~50 mg) is not equivalent to an afternoon coffee (~95–200 mg); the pharmacokinetic floor scales with dose Gardiner et al. 2023. Keep the ritual, lose the drug. Decaf, herbal tea, hot water with lemon — the social and emotional function of the 3 pm cup is real, and you don't have to give it up to give up the caffeine. The one combination that doesn't help: caffeine + L-theanine. It modulates how the caffeine feels; it doesn't change the half-life or the receptor occupancy.
For people whose work demands afternoon alertness, the caffeine nap — a small caffeine dose taken right before a 20-minute nap, so it peaks on waking — gives a sharper afternoon than the standard cup and doesn't require giving anything up. The curfew rule still applies to the evening; the nap is just a better afternoon tool than a late espresso.
Adjacent
How much caffeine total is too much — the dose question, separate from the timing one — is a different entry. So is morning sunlight, which competes with caffeine for the wake-up job and works on a different system (circadian, not adenosine). For the underlying sleep architecture this entry protects, see entries on slow-wave sleep, sleep timing, and the alertness-mood trade in chronic caffeine use. If the morning grogginess persists after a clean curfew, the next thing to rule out is a breathing disorder during sleep; it's the most common explanation for "I slept eight hours and feel destroyed."
Substance + claimed effects
A caffeine curfew is a self-imposed daily cutoff time after which no caffeine is consumed, with the goal of letting the drug clear before sleep onset. The conventional anchor is at least 6 hours before bed, derived directly from Drake et al. 2013; recent meta-analytic work pushes that to ~8.8 hours for coffee and ~13.2 hours for pre-workout-dose caffeine Gardiner et al. 2023. The claimed effects are entirely downstream of clearing caffeine before sleep: shorter sleep latency, more deep sleep, fewer nighttime awakenings, less morning grogginess, and a knock-on lift to next-day energy, focus, and mood once the sleep debt the curfew was hiding stops accumulating. The entry covers all of these holistically; the substance is the cutoff, the consequences are the sleep architecture and its second-order effects.
Evidence by addressing question
mechanism
Caffeine works by blocking adenosine A1 and A2A receptors in the central nervous system, roughly 30 minutes after ingestion Reichert et al. 2022. Adenosine accumulates extracellularly during waking and is the principal homeostatic accumulator of sleep pressure; caffeine occupies its receptors so the brain stops registering the pressure it has built. The effect on sleep architecture is mechanistic: blocking A2A receptors in the nucleus accumbens shell drives caffeine's wake-promoting action Reichert et al. 2022. Importantly, caffeine's main metabolite, paraxanthine, has similar receptor affinity and is itself sleep-disrupting, which extends the wake-promoting window beyond what parent-compound half-life alone would predict.
Pharmacokinetics. Mean elimination half-life in adults is ~5 hours (range 1.5–9.5 h), with primary metabolism via hepatic CYP1A2 IOM 2001. The CYP1A2 −163 C>A polymorphism (rs762551) splits the population into fast metabolizers (AA, ~50%), intermediate (CA, ~42%), and slow metabolizers (CC, ~8%) Cornelis et al. 2006. Half-life can extend to 6–10+ hours in slow metabolizers and to ~15 hours in late pregnancy or on hormonal contraceptives; smoking induces CYP1A2 and roughly doubles clearance. With a 5-hour half-life, a 200 mg afternoon dose still leaves ~100 mg circulating at 5 hours and ~50 mg at 10 hours — enough to occupy receptors and blunt slow-wave activity. Caffeine taken in the morning still measurably alters that night's EEG, with reduced spectral power in the delta (deep-sleep) band Landolt et al. 1995.
evidence
The anchor trial is Drake et al. 2013: 12 healthy normal sleepers received 400 mg caffeine (≈ two to three cups of coffee) at 0, 3, and 6 hours before habitual bedtime, double-blind versus placebo, with polysomnographic recording. The 6-hour-before-bed dose produced a >1 hour reduction in objective total sleep time compared to placebo — and crucially, subjective sleep quality did not flag the disruption: participants slept less but did not report sleeping worse. That dissociation is the practical case for a curfew that doesn't rely on felt experience.
The recent meta-analysis covering 24 trials in adults Gardiner et al. 2023 found caffeine consumption reduces total sleep time by ~45 minutes, sleep efficiency by ~7%, increases sleep onset latency by ~9 minutes and wake-after-sleep-onset by ~12 minutes, increases light (N1) sleep duration by ~6 minutes, and decreases deep (N3) sleep duration by ~11 minutes. Modeling against time-to-bed, the authors derived ~8.8 hours as the minimum gap for a standard coffee (~107 mg) and ~13.2 hours for a standard pre-workout serving (~217.5 mg).
Habitual users do not escape the disruption. Weibel et al. 2021, a 10-day controlled trial in healthy men, found that regular daily caffeine intake delayed REM sleep promotion and attenuated overall sleep quality, with EEG sigma activity reductions consistent with reduced sleep depth. The systematic review by Clark & Landolt 2017 reaches the same conclusion across heterogeneous trials: tolerance to alertness effects develops, tolerance to sleep effects does not.
protocol
Operational rules that fall out of the evidence:
- Default cutoff: at least 8 hours before target sleep time for one to two cups of coffee. For an 11 pm bedtime that means no caffeine after 3 pm.
- Stretch to 10–12 hours for higher doses (large coffees, pre-workout supplements, multiple energy drinks). For a 16 oz energy drink (~160 mg) or a double espresso afternoon stack, the curfew effectively becomes noon.
- Slow metabolizers (CC genotype, late pregnancy, oral contraceptive use): add 2–4 hours to the cutoff. Without genetic testing, the heuristic is felt experience — if caffeine in the early afternoon still keeps you wired at 10 pm, you metabolize slowly.
- The dose matters as much as the timing. 50 mg of black tea at 4 pm is not equivalent to 200 mg of coffee at 4 pm. The Gardiner cutoffs scale with dose.
- Decaf (~2–4 mg per cup) is functionally exempt.
misconceptions
"I sleep fine after my afternoon coffee." The Drake polysomnography work shows objective sleep disruption without matching subjective complaint — habitual late-caffeine users have lost the ability to feel the disruption directly Drake et al. 2013. "I'm tolerant." Tolerance develops to the alertness effect, not to the sleep-disruption effect; chronic caffeine attenuates sleep quality across the entire night, every night Weibel et al. 2021; Clark & Landolt 2017. "Morning coffee is consequence-free." Landolt et al. 1995 showed 200 mg at 7 am still reduced slow-wave delta activity that night — the effect is small but measurable. "The morning grogginess means I need caffeine." The morning grogginess is partly a withdrawal symptom; the curfew breaks the cycle Roehrs & Roth 2008.
failure-modes
The common screw-ups: (1) anchoring the curfew on clock time without recalibrating when bedtime moves — a 3 pm cutoff is wrong if you're going to bed at 9 pm; (2) forgetting hidden caffeine sources (tea, chocolate, cola, pre-workout, "Diet" sodas, certain medications including some headache OTC); (3) "decaf" coffee actually containing 7–15 mg per cup, which adds up; (4) doing the curfew but not the dose discipline — a 400 mg morning load with a 2 pm cutoff still produces evening receptor occupancy; (5) treating slow-metabolizer status as a binary — if you smoke and then quit, your metabolism changes and your old curfew becomes inadequate.
stakes
The curfew exists because the consequences of late caffeine compound. Each night's deep-sleep deficit (~11 min N3 reduction) Gardiner et al. 2023 drives a next-day adenosine load that the morning coffee masks — the user reads the masking as "my normal" and increases dose to maintain it Roehrs & Roth 2008. The sleep architecture changes (delayed REM, attenuated SWA) operate independently of perceived sleep quality. Over months and years this is the metabolic and mood profile of chronic mild sleep restriction.
payoff
The change shows up within nights, not weeks: in trials that switched habitual users to placebo or earlier cutoffs, the slow-wave delta activity rebounds quickly Landolt et al. 1995, and total sleep time and sleep efficiency normalize Gardiner et al. 2023. The lived consequence is that the felt-tired-but-still-can't-sleep paradox dissolves, morning awakenings shift away from 3–5 am, and the 2 pm "I need a coffee or I'll die" urge softens because the receptor-down-regulation pressure has eased Roehrs & Roth 2008.
alternatives
Switching to lower-dose caffeine forms (tea over coffee), spreading the dose earlier in the day, or substituting decaf after the curfew preserve the ritual without the receptor occupancy. L-theanine pairing is sometimes cited but doesn't change caffeine pharmacokinetics — it modulates the subjective experience, not the half-life. For people whose work demands afternoon alertness, a 20-minute nap with a small pre-nap caffeine dose (so it peaks on wake) is an alternative strategy, but the underlying curfew rule still applies to the evening tail.
The credibility range
Optimist case. The mechanistic case for a curfew is tight: caffeine is a competitive adenosine antagonist with a measurable half-life, and adenosine signaling is the homeostatic substrate of sleep pressure. The trial evidence is consistent across decades — Landolt 1995, Drake 2013, Weibel 2021, Gardiner 2023 — with effect sizes (45 min less TST, 11 min less deep sleep, 7% less efficiency) that easily clear felt-experience thresholds. The intervention is free, side-effect-free, and reversible night to night. There is no realistic basis for a person with a sleep complaint to not try moving their caffeine cutoff earlier.
Skeptic case. Most trials are small (n=8–24 in the canonical studies), in young healthy men, with acute dosing — not chronic habituated coffee drinkers whose sleep may already be in steady-state adaptation. Effect sizes vary substantially across individuals; the meta-analysis confidence intervals are wide. Some habitual heavy users self-report no curfew benefit. The cutoff-time recommendations (8.8 h, 13.2 h) are model-derived single-point estimates that pool widely varying responses. And the curfew alone, without sleep timing or sleep hygiene attention, may produce no felt change because other variables dominate.
Author's call. The mechanism is solid, the direction of effect is settled, and the practical curfew rule (≥6 h, better 8 h, longer for slow metabolizers or high doses) is one of the highest-yield, lowest-cost sleep interventions in the catalogue. Where the data are softer is the precision of the cutoff time, not its existence; individual calibration to felt experience is appropriate. Score evidence high (4), controversy low (1).
Stakeholder + incentive map
Pushing the curfew: sleep clinicians, AASM sleep hygiene guidance, behavioural sleep medicine practitioners, the broader CBT-I framework. No commercial stakeholder profits from a caffeine cutoff. Pushing against: the coffee shop afternoon-trade economy and the energy drink industry (whose 3–6 pm marketing window depends on people drinking caffeine then), pre-workout supplement makers who position late-afternoon dosing as standard. The "biohacker" online culture is mixed — some advocate strict early cutoffs; others promote nootropic stacks that ignore the pharmacokinetic floor. The skeptic counter-position is informal — there isn't an organised dissent, just the inertia of the existing afternoon coffee ritual.
Population variability
Variation runs along several axes: (1) CYP1A2 genotype — slow metabolizers (CC, ~8% of Europeans) need 2–4 extra hours of cutoff; fast metabolizers (AA, ~50%) can sometimes tolerate later cutoffs but still get sleep architecture effects Cornelis et al. 2006. (2) Age — elderly adults have reduced caffeine clearance and increased sensitivity; the cutoff should be earlier. (3) Pregnancy and oral contraceptives — half-life extends to ~10–15 h; effective cutoff is mid-morning. (4) Smoking status — smokers metabolize ~50–70% faster, but former smokers regress quickly. (5) Baseline sleep status — sleep-deprived individuals show more alertness benefit from caffeine but the underlying sleep architecture damage is still present Roehrs & Roth 2008. (6) Adenosine receptor polymorphisms (ADORA2A) — some variants increase subjective caffeine-induced anxiety and may amplify sleep disruption.
Knowledge gaps
Most acute-dose trials use young, healthy, male participants; chronic-habituated and female and older-adult data are thinner. Field studies using wearable EEG in real-life caffeine schedules are nascent. The interaction between curfew and chronotype (extreme evening vs morning types) is under-studied. The dose-response curve below 100 mg is poorly characterised — the cutoff-time models extrapolate from higher doses. Long-term cumulative effects of habitual late-caffeine on disease endpoints (cardiovascular, metabolic) via the sleep-loss pathway are inferred, not directly trialled.
Scope. The brief listed sleep latency, sleep depth/architecture, nighttime awakenings, and next-day alertness. All four are covered in evidence, mechanism, and payoff. No narrowing.
Hard call: dose-vs-timing. The brief framed this as a timing entry. The literature shows dose and timing co-determine sleep impact (Gardiner 2023's cutoffs scale with dose). I kept the entry centered on the curfew but included the "total daily dose is a separate dial" caveat in failure-modes. A full caffeine-dosing entry is a separate item — flagged below.
Rating difficulty: energy and focus. Both are indirect: the curfew restores sleep, sleep delivers energy and focus. Scored at 3 (clear effect) rather than 4 because the user experiences the lift through the sleep mechanism, not directly from the cutoff. The Roehrs & Roth feedback-loop framing is the load-bearing claim — if it doesn't break, the energy/focus scores would fall to 2.
Rating difficulty: sleep at 4 not 5. The 5-anchor is "dominant effect on sleep architecture or duration." Caffeine curfew restores ~45 min TST and ~11 min N3 on the meta-analysis — substantial, but not the level of an apnea diagnosis or a circadian re-anchor. A 4 is the honest call.
Dropped beauty_cumulative and longevity to 0. Initially scored both at 1 on the "chronic mild sleep restriction is bad for everything" logic, but the curfew's marginal contribution above general sleep hygiene is too indirect to claim honestly, and the article doesn't and shouldn't dwell on either. Better a clean zero than a defensible one. Stakes for those long-horizon consequences live in upstream entries (sleep duration, sleep apnea, etc.) rather than here.
Applicability lift to 4. Current daily caffeine consumers are ~80% of adults; the decision audience (consumers reconsidering their cutoff) is broader. Used the meta.md §6 avoid/decision-audience principle to round up rather than splitting hairs on the prevalence number.
Overall score and dream tier. Weighted overall came out to ~43, above the 40-threshold for a mandatory dream narrative. Lever picked was relief (you get the afternoon, evening, and morning back), not aspiration — the entry's honest hook is what you've been losing, not what you could become. The dek and tagline are written from the narrative; opening paragraph carries it too.
Excluded:
- Total daily caffeine dose ceiling — its own entry. Different decision, different evidence base (FDA ≤ 400 mg/day, pregnancy ≤ 200 mg/day).
- Caffeine for sleep deprivation / shift work — distinct use case where the cost-benefit flips. Future entry candidate.
- CYP1A2 genetic testing — mentioned as a calibration cue but the testing decision is its own entry under screening or supplements.
- Coffee's cardiovascular and metabolic effects — entirely separate substance discussion (the bean, not the molecule's sleep impact). Out of scope.
- Anxiety / panic interactions — the ADORA2A polymorphism literature is real but doesn't bear on the cutoff decision; out of scope here, flag for a "caffeine and anxiety" entry.
Future links: sleep apnea (out-of-scope already points to it), slow-wave sleep entry, morning sunlight entry (mechanistic competitor for the wake-up function), the eventual total-dose entry, the caffeine-nap protocol.
Citation hygiene. Library held no caffeine citations at start — built nine from scratch. All article cites resolve to refs that appear in the research dossier. No pre-prints used.
Caffeine Curfew
Direct, measurable, dominant. The anchor RCT shows 400 mg of caffeine 6 hours before bed cuts objective total sleep time by over an hour (Drake et al. 2013); the 2023 meta-analysis confirms +9 min sleep latency, +12 min WASO, -11 min deep sleep across 24 trials (Gardiner et al. 2023). Effect persists in habitual users (Weibel et al. 2021).
Behavioral, daily, and culturally embedded — afternoon coffee is a social and emotional ritual for many adults. Skipping it requires sustained willpower for the first 1-2 weeks; once the sleep dividend lands, the urge softens.
One well-controlled placebo RCT with polysomnography (Drake et al. 2013), a 2023 meta-analysis across 24 trials (Gardiner et al. 2023), EEG mechanism studies (Landolt et al. 1995), and a chronic-dosing trial (Weibel et al. 2021). Mechanism (adenosine A1/A2A antagonism) is settled (Reichert et al. 2022). Backed by AASM sleep hygiene guidance.
Reversing chronic mild sleep restriction lifts daily function within days: fewer 3 am awakenings, less morning grogginess, lower resting heart rate. Driven by the +45 min total sleep time and +11 min deep sleep restored when caffeine is cleared (Gardiner et al. 2023).
The afternoon coffee that wrecks sleep is what created the morning need for it. Cutting caffeine 8+ hours before bed restores deep sleep (Drake et al. 2013) and breaks the masking cycle (Roehrs and Roth 2008), so daytime energy stops depending on the next dose.
Sleep depth (slow-wave activity) is the substrate of next-day cognitive performance; caffeine taken within 6 hours of bed measurably suppresses it (Landolt et al. 1995). Restoring it lifts working memory, attention, and deep-work capacity at no caffeine cost.
Mood stability is sensitive to sleep continuity; reducing nighttime awakenings and restoring REM (delayed by caffeine, per Weibel et al. 2021) produces a small but real improvement in irritability and emotional regulation.