Health × Learning and the Ceiling Problem

Two things the system gets wrong about the learner: the body it sits in, and the limit it imposes

education-atlas — deep series, doc 04. A companion to `EDUCATIONPROBLEMS.md`, which establishes the measured global crisis (learning poverty 48.3%, 51M+ out-of-school children, financing below the 4%-of-GDP floor). This doc turns from the measured system to two largely unmeasured levers the founding problem statement flagged but could not quantify: the biology of the learner (§A) and the ceiling the system imposes on the learners who could go furthest (§B)._

A note on evidence discipline, up front. This document grades every claim into tiers and labels speculation as speculation. Part A in particular touches a literature — circadian biology, light, water, mitochondria — that has a rigorous, established core and a fringe that overclaims. Bucket Foundation's whole proposition is that primary research can be citeable and trustworthy. A document that laundered fringe claims as settled science would betray exactly that proposition. So the rule here is strict: STRONG, EMERGING, and FRINGE are marked explicitly, effect sizes are given where they exist, and the one genuinely fringe source (the Jack Kruse corpus) is presented as a provocation to investigate, never as evidence of fact.

PART A — Health × Learning: the biology the system ignores

A.0 The thesis

Modern schooling treats the brain as a disembodied information processor: a thing that takes in symbols at 8:00 a.m. under a fluorescent tube, sits motionless for six hours, runs on a vending-machine breakfast, and is expected to encode and retain. Every one of those design choices runs against a body of physiology that is, in its core findings, not controversial. Sleep, daylight, movement, and nutrition are not wellness garnish on the side of "real" academics. They are upstream determinants of whether encoding, consolidation, and retrieval happen at all. The atlas's §5 said plainly that the indicators "miss the problems that may matter most" — pedagogy, relevance, and, we add here, the physiological state of the learner. No global dataset measures whether the child was rested, fed, moved, and lit. The evidence below says it should.

The honest shape of this literature: a large established core (sleep, exercise, gross nutritional deficiency, neurotoxic exposures, sensory acuity), a plausible and growing middle (circadian timing, light environment, screens), and a fringe edge that takes the real circadian/light science and extrapolates it into claims that are not established. We take them in that order, because the order is the credibility.

A.1 STRONG, established evidence — lead with these

These findings rest on multiple studies, meta-analyses, and/or the formal position statements of medical bodies. They are the ground a reform argument should stand on.

A.1.1 Sleep and school start times

The single best-evidenced, most policy-actionable health-learning finding in education. Adolescent circadian biology shifts later at puberty — the natural sleep-onset and wake times move back by roughly two hours — yet most middle and high schools start at or before 8:00 a.m., guaranteeing chronic sleep restriction in exactly the population whose clocks have moved.

• The American Academy of Sleep Medicine (AASM) position statement (Watson et

al., J Clin Sleep Med 2017) asserts that middle and high schools should start no earlier than 8:30 a.m., and that later start times improve daytime alertness, attendance, tardiness, mood, and academic performance. The American Academy of Pediatrics, the American Medical Association, and the CDC all hold the same position.

• A *2022 meta-analysis in Pediatrics*** (Yip et al., "School Start Times,

Sleep, and Youth Outcomes") found a robust result: middle schools starting at 8:30 a.m. or later showed the best outcomes for sleep duration and the greatest improvements in behavioral health and academic performance.

This is as close to a free lunch as education policy offers: changing a number on a bell schedule, at near-zero cost, returns sleep to a sleep-deprived population. That schools overwhelmingly still start early is the cleanest single illustration of the system ignoring the body.

A.1.2 Sleep and memory consolidation

Independent of when school starts, sleep is mechanistically load-bearing for learning. The consolidation of newly encoded memory — moving it from fragile short-term traces to durable storage — happens substantially during sleep, via slow-wave activity and sleep spindles.

• A systematic review and meta-analysis of sleep and novel word learning

(Schreiner & Rasch, and the broader literature synthesized in Psychon Bull Rev 2021) found sleep beneficial versus equivalent wakefulness at g ≈ 0.50 overall (recall g ≈ 0.57; recognition g ≈ 0.52) — a moderate, robust effect.

• Meta-analyses spanning five decades show that **total sleep deprivation, both

before and after learning, impairs memory* for newly learned material (Newbury et al., Neurosci Biobehav Rev* 2021).

• The mechanism has converging support: sleep-spindle activity (53-study

meta-analysis) and slow-oscillation–spindle coupling are associated with consolidation, and targeted memory reactivation during sleep reliably improves retention (Hedges' g ≈ 0.29; Hu et al., 2020).

The applied translation: a student who studies and then loses sleep has paid the encoding cost and forfeited the consolidation return. The school day that ends in homework that ends in a short night is structurally anti-retention.

A.1.3 Physical activity, fitness, and cognition

The "sit still to learn" model is contradicted by the exercise-cognition literature.

• The landmark consensus review is Donnelly, Hillman, et al. (2016),

Medicine & Science in Sports & Exercise — 64 studies on cognition/brain, 73 on academic achievement — concluding that physical fitness, single bouts of activity, and activity interventions benefit children's cognitive function, with the strongest and most consistent effects on cognitive control / executive function (attention, inhibition, working memory) — precisely the faculties a classroom depends on.

• A 2020 systematic review and meta-analysis (de Greeff et al. / Álvarez-Bueno

lineage; PMC7372103) confirmed positive associations between physical activity and both academic performance and cognition in children.

• A 2023 multi-level meta-analysis of childhood physical-activity interventions

(Educational Psychology Review) again found positive causal effects on cognition and academic achievement.

Effect sizes here are typically small-to-moderate and heterogeneous — this is not a claim that gym class beats reading instruction — but the direction is settled and the cost of inactivity is real. Recess and movement are not time stolen from learning; suppressing them taxes the executive function learning runs on.

A.1.4 Nutrition: gross deficiency is a direct IQ lever

The strongest nutrition findings are not about superfoods; they are about the catastrophic cost of specific deficiencies, especially in early life.

• Iodine deficiency is the world's leading preventable cause of intellectual

impairment. Meta-analytic estimates put the IQ cost of deficiency at roughly 13.5 points on average; controlled comparisons find iodine-deficient children scoring 6.9–10.2 points lower, with supplementation improving cognition in mildly deficient children.

• Iron deficiency / anemia is associated with poorer cognition and school

achievement; supplementation in school-age children improves cognitive outcomes (meta-analysis, PMC10298800). A grave caveat the literature is honest about: iron deficiency in very early life can cause irreversible damage that later supplementation does not undo. Timing is not negotiable.

• Breakfast: eating breakfast is associated with better cognition and school

performance (Adolphus et al., Front Hum Neurosci 2013), plausibly partly via correcting micronutrient gaps. The breakfast literature is more confounded than the deficiency literature (hungry-vs-fed is entangled with poverty), so we rate the deficiency findings as strong and the breakfast-program findings as strong-but-confounded.

The systemic indictment: a child cannot be iodine-corrected by a curriculum. A vending-machine food environment and unaddressed nutritional deficiency are a learning intervention the system is making by default — in the wrong direction.

A.1.5 The boring basics: sensory acuity, lead, and air quality

The least glamorous and among the best-evidenced.

• Vision and hearing. Uncorrected refractive error and undetected hearing loss

directly degrade the input channel for instruction. Randomized provision of eyeglasses to children with refractive error improves academic outcomes; this is one of the cheapest, highest-certainty interventions in the entire literature.

• Lead exposure is a direct, dose-dependent neurotoxin. A 2022 systematic

review and meta-analysis (Reuben-lineage; Systematic Reviews 2022) found IQ deficits scaling with exposure duration (≈3.5 points under 4.5 years of exposure; far larger with chronic exposure). Population-level analyses attribute multi-point average IQ losses to historical leaded gasoline. There is no established safe threshold.

• Air quality. Traffic-related and indoor air pollution near and in schools is

associated with slower cognitive development (Sunyer et al., PLoS Med 2015 — Barcelona BREATHE cohort: reduced growth in working memory in high-pollution schools) and lower reading/math/science performance in U.S. schoolchildren (Environmental Research 2019).

These are environmental-justice findings as much as education findings: the neurotoxic load falls hardest on the poorest schools, compounding the equity gradient the atlas documents.

Summary of Tier 1. Where a child sleeps, moves, eats, sees, hears, and breathes is causally upstream of whether they learn. None of it appears in the global education indicators. This is the under-used lever.

A.2 EMERGING / plausible — promising, not yet settled

Real signal, weaker or thinner evidence than Tier 1. Worth acting on cautiously; not worth overclaiming.

A.2.1 Chronotype-vs-school-time mismatch (beyond just "start later")

Distinct from total sleep loss: even rested students can be mismatched. Late chronotypes tested in the morning underperform their own afternoon performance (Goldin et al., Sci Rep / Vollmer et al., 2015: exam timing affects early vs. late chronotypes differently). A 2023 study (npj Science of Learning) found better chronotype–school-timing alignment associated with lower grade retention, with the effect concentrated in morning schedules. The implication — that when you test, not only whether the student slept, moves scores — is plausible and mechanistically coherent, but rests on a smaller, more observational base than the start-times literature.

A.2.2 Daylight and the classroom light environment

The famous result is the Heschong Mahone Group (1999) California study: ~21,000 students, 2,000+ classrooms, multivariate models — students in the daylightiest classrooms progressed 20% faster in math and 26% faster in reading over a year than those in the least-daylit rooms. The effect sizes are large and the study is widely cited; it is also observational and a later reanalysis moderated some claims while preserving a positive daylight association. We rate it emerging-strong: a real, replicated-in-spirit signal that the windowless fluorescent box is not a neutral container. Mechanistically it plausibly braids sensory/visual quality with circadian light input (next item).

A.2.3 Circadian light environment and alertness

The established half: bright, blue-enriched light in the morning advances and entrains the circadian clock and acutely raises alertness; light's effect on the melanopsin/ipRGC → suprachiasmatic-nucleus pathway and on melatonin is textbook photobiology. The emerging half is the dose-response in real classrooms — how much the indoor light environment of an actual school day moves learning outcomes (as opposed to lab alertness). The biology is solid; the field quantification in education is still thin. Schools as "fluorescent boxes" with little daylight and circadian-flat lighting are plausibly leaving alertness and entrainment on the table.

A.2.4 Screens and sleep

Evening use of light-emitting screens is associated with shorter, worse sleep, plausibly via melatonin suppression and arousal/displacement (the displacement of sleep time may matter as much as the photopic effect). Strong as a correlation and mechanistically reasonable; causal magnitude in adolescents is still debated and confounded. Rate: emerging, lean toward real.

A.3 FRINGE / one-source, NOT established science — explicitly flagged

Read this framing before the claims. What follows is drawn from the Jack Kruse corpus (~/agfarms/kruse-corpus/, ~460 articles) held in Bucket's biophysics intake as one partial source, explicitly not the centre of the biophysics branch. Kruse is a neurosurgeon and prolific blogger who has built an elaborate framework around light, water, and mitochondria. None of the specific claims below are established science. They are presented here for one reason: to draw the line, in public, between the real circadian/light science above (Tier 1–2) and the speculative extrapolations that share its vocabulary. To a casual reader the two can sound identical. They are not. Citing the corpus as the source of a provocation is legitimate; presenting its conclusions as fact would be the exact failure mode this document exists to avoid.

What the corpus claims (sourced to the corpus, asserted as its position, not as truth):

• Circadian/light "quantum biology" of cognition. Kruse frames diseases like

Alzheimer's as light-driven "proteopathies" — e.g., Can't Remember? Is Your Protein Bent? argues AD is caused by electron loss and blue-light-mediated protein misfolding via the retina, invoking microtubule "gamma coherence." (Source: kruse-corpus cant-remember-is-your-protein-bent.md.) Status: the claimed causal chain — blue light → retinal DHA electron loss → protein bending → AD — is not established; it stitches real components (melanopsin signaling, DHA in retina, tau pathology) into a mechanism mainstream neuroscience does not endorse.

• EZ / structured ("fourth phase") water. The corpus leans heavily on Gerald

Pollack's exclusion-zone water and on "aquaphotomics" (Aquaphotomics 101: Battery Creation) to claim cells summon ATP from light-and-water beyond what food can supply. Status: the EZ phenomenon is real and replicated (see Pollack in Bucket's 05-biophysics.md); the biological scope Kruse assigns it — water as a primary cellular "battery" powering cognition — is disputed and not established.

• Deuterium-depletion. The corpus treats deuterium as a metabolic toxin and

"UCP2 as the deuterium carburetor" (qt-18), implying deuterium-depleted water / diet improves mitochondrial and cognitive function. Status: deuterium kinetic isotope effects are real chemistry; clinical deuterium-depletion for cognition or health is not established and the human evidence is minimal.

• Mitochondrial-light cognition / cold thermogenesis. A large body of corpus

posts (the Cold Thermogenesis and Hypoxia series, CPC-15: blood as a magnetohydrodynamic fluid linking sun to mitochondria) claim sunlight, cold, and magnetism directly tune mitochondrial energy and therefore mental performance. Status: mitochondrial bioenergetics is real (see Mitchell, Wallace, Lane in Bucket's canon); the specific sun/cold/EMF → cognition dosing claims are not established and are largely unsupported by controlled human trials.

The honest distinction, stated once more. Established science says: morning daylight entrains the clock and raises alertness (A.2.3); sleep consolidates memory (A.1.2); the EZ water phenomenon exists at interfaces (Pollack). Kruse's framework starts from these real findings and extrapolates to a unified light-water- mitochondria theory of cognition and disease that the evidence does not support. The value of holding the corpus in intake is as a hypothesis generator and a stress test — some of its provocations (the classroom light environment matters; circadian disruption has cognitive cost) point at real, fundable questions that the mainstream A.2 literature is independently beginning to confirm. The value is not in adopting its conclusions. Anyone citing this document should be able to repeat that sentence.

A.4 What Part A means for reform

The atlas measures seats, money, and reading scores. It cannot measure whether the learner arrived rested, fed, moved, and well-lit. The Tier 1 evidence says those states are causally upstream of learning, cheaply improvable, and steepest exactly where the equity gradient is worst (the poorest schools carry the most lead, the worst air, the least daylight, the most food insecurity). The cheapest reforms in education may not be pedagogical at all: start school later, let children move, fix vision and hearing, remove neurotoxins, feed them, and stop running classrooms as windowless boxes. None of it requires believing anything fringe. All of it is ignored at scale.

PART B — The Ceiling Problem: how systems cap the top, and what letting excellence super-excel would take

B.0 The thesis

Part A is about the floor the system ignores under every learner. Part B is about the ceiling it presses down on the few who could go furthest. The industrial model batches students by birth year and paces them to the middle of the batch. That design quietly inflicts on high-ability students a continuous, invisible loss: the cost of years spent at a fraction of their possible rate. The gifted-education research is unusually clear and unusually ignored — clearer, in fact, than most of education research — and it says the cap is real, the remedies work, and the remedies do not harm anyone else.

This connects directly to Bucket Foundation's thesis: the canon exists for "the small number of people who can do genius work with AI — people who can take a model and an axiom and reach a layer of reality nobody has reached before." A system that caps its top students is, in Bucket's terms, throttling exactly the people the frontier depends on. But — and Part B holds this honestly — equity and excellence are a real tension, not a rhetorical one, and pretending otherwise would be its own dishonesty.

B.1 STRONG evidence: ability predicts, the cap is real, acceleration works

B.1.1 Ability measured early predicts real outcomes — across the whole range, including the very top

The Study of Mathematically Precocious Youth (SMPY) — founded by Julian Stanley at Johns Hopkins in 1971, now a 50-year longitudinal study at Vanderbilt under David Lubinski and Camilla Benbow, tracking 5,000+ intellectually talented individuals identified before age 13 — is the deepest dataset on the gifted.

Its central, repeatedly replicated findings:

• *Ability differences within the gifted band still matter.* Among the top 1%,

those in the top quartile (Q4) substantially outperform the bottom quartile (Q1) of that same elite group on doctorates, patents, publications, tenure, and income (Lubinski & Benbow, 2006; Kell, Lubinski & Benbow, 2013). There is no plateau where "smart enough" stops predicting. The right tail keeps paying out.

• The profoundly gifted (top 1 in 10,000) pursued doctorates at **>50× base

rate* and produced notable creative/scientific work strikingly young (Lubinski, Webb, Morelock & Benbow, 2001, J Appl Psych*).

• The specific shape of ability predicts the shape of the career — quantitative

vs. verbal tilt, plus spatial ability, forecasts who becomes a STEM innovator (the "aptitude complex": math + spatial + investigative interests + theoretical values).

• Crucially for Part B: 95% of this population used some form of acceleration,

and they strongly preferred educational placement matched to their rate of learning.

The takeaway is uncomfortable for an egalitarian-leveling instinct but is what the data shows: at the top, capability is not fungible, the differences keep mattering, and the people themselves want — and benefit from — being unleashed, not held to the median pace.

B.1.2 The cap is structural, and the remedy (acceleration) is the best-evidenced intervention in gifted ed

Acceleration — grade-skipping, subject-acceleration, early entrance, dual enrollment — is, by the weight of evidence, the single most effective and most cost-effective intervention for gifted students, and the most resisted.

• A Nation Deceived (Colangelo, Assouline & Gross, 2004) and its successor **A

Nation Empowered (Assouline, Colangelo, VanTassel-Baska, 2015) synthesize the acceleration literature. Headline conclusions: acceleration is the most effective curriculum intervention for gifted children; it has long-term academic and social benefits; and grade-skipping is virtually cost-free**. Accelerated students consistently and significantly outperform matched non-accelerated peers, in high school and college.

• The most authoritative quantitative synthesis is **Steenbergen-Hu, Makel &

Olszewski-Kubilius (2016)*, Review of Educational Research* — two second-order meta-analyses covering ~100 years of research:

• Accelerated students outperform same-age non-accelerated peers at **g ≈

0.70 (a large effect), and — tellingly — perform on par with older peers** they were accelerated to join (g ≈ 0.09, i.e., no deficit). Acceleration moves them up without breaking them.

• Special grouping for the gifted: g ≈ 0.37; within-class grouping g ≈

0.19–0.30; cross-grade subject grouping g ≈ 0.26.

B.1.3 The decisive equity finding: grouping/acceleration help the top without harming the rest

This is the finding that dissolves the most common objection. In Steenbergen-Hu et al. (2016), the effects of within-class and cross-grade grouping did not vary across high-, medium-, and low-ability students — grouping did not come at the expense of the lower groups. The line that should be quoted: a well-implemented ability-grouped program for gifted students "can be of great value with minimal impact on other learners in the same cohort." (Note: between-class tracking showed near-zero benefit, g ≈ 0.04–0.06 — so this is not a blanket endorsement of rigid tracking; the evidence favors flexible, subject-specific grouping and acceleration, not permanent sorting.)

B.1.4 Bloom's 2-sigma: the size of the prize, honestly stated

Benjamin Bloom (1984) reported that students tutored one-to-one with mastery learning performed ~2 standard deviations better than conventional-classroom students — the average tutored student above the 98th percentile of the classroom. The "2 sigma problem" he posed: find group methods as effective as one-to-one tutoring.

The honest update is important and credibility-protecting: the 2-sigma figure was likely inflated by the specific conditions of Bloom's graduate students' small studies. Later systematic reviews put mastery learning alone closer to ~0.5 sigma, and human one-to-one tutoring's average effect at roughly d ≈ 0.79 — not 2.0, but still one of the largest effects in education. So: the exact number is contested; the direction and magnitude — that individualized, mastery-paced instruction is dramatically more effective than batch instruction — is robust. This matters for the AI-tutor argument below: the claim should be "tutoring is a very large effect and AI may scale it," not "AI tutors deliver a guaranteed 2 sigma."

B.1.5 How systems actually cap the top — the mechanisms

Synthesizing the above against the industrial model (and the atlas's "teach to the middle/test" critique):

1. Age-batched pacing. Progress is yoked to birth year, not mastery. The fast

student waits.

1. Ceiling effects in assessment. Grade-level tests cannot measure above

grade level; a student who maxes the test is invisible — the system literally cannot see how far ahead they are, so it cannot respond.

1. Egalitarian leveling / detracking. Pressure to keep cohorts together can

remove acceleration and grouping — well-intentioned, but the evidence says it costs the top without the claimed benefit to the rest (the harm-to-others fear is largely unsupported for flexible grouping).

1. Teach-to-the-middle and teach-to-the-test. Accountability regimes reward

moving the median and the bubble-near-proficiency students; they create no incentive to extend the already-proficient. The atlas notes this as "teach to the middle/test"; the gifted-ed data quantifies its cost.

1. No frontier access. Even an unleashed top student eventually hits the limit

of the local curriculum, the local teacher's expertise, and the building's walls. The frontier of a field is not in the building.

B.2 What letting excellence super-excel would take

If the cap is real and the remedies are evidence-backed, the constructive question is what an un-capped path looks like. Four components, the first three evidence-grounded, the fourth promising-but-unproven:

1. Acceleration as default, not exception (STRONG). Make subject-acceleration,

grade-skipping, early entrance, and dual/early college routine and frictionless for students whose mastery warrants it. This is the best-evidenced, cheapest lever and the most under-used.

1. Above-level assessment (STRONG, methodological). You cannot serve what you

cannot measure. Out-of-level / adaptive testing (the SMPY method — give a 12-year-old the SAT) reveals the true ceiling so the system can respond to it. Ceiling-bounded grade-level tests structurally hide the top.

1. Mentorship and talent development (STRONG). SMPY and the talent-development

tradition (Bloom's Developing Talent in Young People, 1985) both show that eminence requires sustained, individualized mentorship into a real domain — not just acceleration through a curriculum, but apprenticeship at the edge of a field.

1. AI tutors with no ceiling (EMERGING / PROMISING — do not overclaim). The

structural appeal: a tutor that never runs out of grade level, never has to teach to the median, and can pace to mastery for one student is, in principle, the Bloom one-to-one condition at scale. But the evidence is early. Use the honest framing: tutoring is a very large effect (d ≈ 0.79); mastery pacing is real (~0.5σ); AI may scale these — it has not yet demonstrated Bloom's 2 sigma in rigorous trials, and the 2-sigma number itself is contested (B.1.4). The promise is real and the proof is pending. An honest Bucket position claims the opportunity, flags the evidence gap, and refuses the hype number.

B.3 The equity-vs-excellence tension — held, not dismissed

This document would be dishonest if it pretended the tension away. It is real:

• Finite teacher attention, time, and funding are rival goods; what goes to the top

can, under scarcity, come from elsewhere.

• The atlas's own central finding is an equity catastrophe — 48% learning

poverty, a tenfold gap by region, conflict states uncounted. For most of the world's children the binding problem is the floor, not the ceiling. A reform agenda that led with the gifted would be morally and empirically misaimed.

• Gifted programs have a documented history of access inequity — under-identifying

poor and minority students — so "serve the top" can entrench privilege unless identification is deliberately equitable (above-level testing, universal screening, and local norms are the evidence-based mitigations).

The reconciling facts, also honest: the strongest evidence (Steenbergen-Hu 2016) says flexible grouping and acceleration help the top without harming the rest — so this is not strictly zero-sum, and the most common objection is empirically weaker than assumed. The floor (Part A, plus the atlas's access/learning/financing crises) is the first priority by sheer scale. The ceiling is a real, separate, cheaply-addressable problem that the system caps for reasons that are mostly ideological and administrative, not evidentiary. Both can be true. Raise the floor for nearly everyone; remove the ceiling for the few who would otherwise be held at a fraction of their rate. The two agendas use largely different levers and need not compete for the same dollar.

B.4 The tie to Bucket's thesis — stated carefully

Bucket's canon thesis is that "AI + foundations + a small number of brilliant humans = the next layer of reality," and that the foundation exists for the few who can "take a model and an axiom and reach a layer of reality nobody has reached before." Part B is the education-policy face of that thesis: the system most reliably fails the people Bucket is built for. It batches them, paces them to the median, ceilings their assessment, and walls them off from the frontier of their field. Letting excellence super-excel — open access to the full frontier of knowledge, mentorship into a real domain, acceleration as default, and AI tutors that never run out of room — is the same move the canon makes for research: remove the gate between a capable mind and the boundary of what is known.

The discipline Bucket must keep here is the discipline of this whole document: the acceleration/grouping/SMPY evidence is strong and should be led with; the AI-tutor-no-ceiling promise is emerging and must be sold as opportunity, not as a settled 2-sigma guarantee; and the equity tension is real and must be held, not waved away. The frontier-access argument is most credible precisely when it refuses to overclaim — which is, not coincidentally, the same reason anyone should trust Bucket's canon.

Appendix — Evidence ledger (established vs. speculative, at a glance)

PART A

PART B

Key sources

Part A. Watson et al., AASM position statement on school start times, J Clin Sleep Med 2017 (PMC5359340); Yip et al., "School Start Times, Sleep, and Youth Outcomes: A Meta-analysis," Pediatrics 2022; sleep & novel-word-learning meta-analysis, Psychon Bull Rev 2021; sleep-deprivation memory meta-analyses, Neurosci Biobehav Rev 2021; targeted memory reactivation meta-analysis (Hu et al. 2020); Donnelly, Hillman et al., Med Sci Sports Exerc 2016; physical-activity & cognition meta-analysis (PMC7372103); childhood PA intervention meta-analysis, Educ Psychol Rev 2023; iodine & cognition (≈13.5 IQ pts); iron supplementation meta-analysis (PMC10298800); breakfast & cognition, Front Hum Neurosci 2013; lead & IQ meta-analysis, Systematic Reviews 2022 (PMC9150353); BREATHE air-pollution cohort, PLoS Med 2015 (PMC4348510); Heschong Mahone Group, Daylighting in Schools 1999 (ERIC ED444337); chronotype–school-timing alignment, npj Science of Learning 2023 (PMC10284813); exam-timing × chronotype (PubMed 25537752). Fringe source, explicitly flagged: Jack Kruse corpus, ~/agfarms/kruse-corpus/ (~460 articles; e.g., cant-remember-is-your-protein-bent.md, aquaphotomics-101-battery-creation.md, qt-18...ucp2-is-the-deuterium-carburetor.md) — one partial source, not established science.

Part B. Lubinski & Benbow, "SMPY After 35 Years," Perspectives on Psychological Science 2006; Lubinski, Webb, Morelock & Benbow, "Top 1 in 10,000," J Appl Psych 2001; Kell, Lubinski & Benbow 2013; SMPY 50-year program (Vanderbilt); Colangelo, Assouline & Gross, A Nation Deceived 2004; Assouline, Colangelo & VanTassel-Baska, A Nation Empowered 2015 (Acceleration Institute); Steenbergen-Hu, Makel & Olszewski-Kubilius, "What One Hundred Years of Research Says…," Review of Educational Research 2016 (ERIC EJ1121483); Bloom, "The 2 Sigma Problem," Educational Researcher 1984; Bloom, Developing Talent in Young People 1985; modern tutoring/mastery effect-size reviews (Education Next; Nintil synthesis).

Bucket canon cross-reference. bucket-foundation/canon-figures/05-biophysics.md (Pollack EZ water, Ling structured water, Mitchell/Wallace/Lane bioenergetics, Burr/Becker/Levin bioelectrics) — for the established-vs-disputed line within biophysics that Part A's fringe section relies on.

Evidence discipline statement: Every STRONG claim above rests on ≥1 meta-analysis or formal medical/scientific consensus body. Every EMERGING claim is labeled as plausible-not-settled. The one FRINGE source (Kruse corpus) is cited as a source of provocations, never as evidence of fact, and the established circadian/light science is held distinct from his extrapolations throughout. If any single sentence in this document is later quoted, it should be quotable without correction. That is the standard Bucket's canon claims to hold; this document holds it.