Moderate Cross-Pillar

Hormesis and adaptive stress: when a stressor builds capacity and when it just does damage

Summary

Some stressors make you more capable afterwards than you were before, but only when they are acute rather than continuous, matched to what your body can currently handle, and followed by real recovery — and the fact that this is true for training, heat, cold and fasting does not make it true for smoke, alcohol, radiation, sleep loss or unremitting life stress.

Why Moderate

This entry sits at Moderate.

Why not the tier above (Strong). The entry's central organising claims are syntheses rather than replicated findings. The three-condition framework is Realised's articulation with no single supporting study and two of its three legs unoperationalised. The composite "the acute panel reads like pathology" claim is assembled from separate literatures with no study having drawn a full panel mid-exercise. The mechanistic keystone — that removing the stress signal removes the adaptation — rests on a single small trial with roughly ten people per arm and inconsistent subsequent replication. And the hormesis principle itself, at the level of generality this entry uses it, is contested in exactly the domain where the contest has consequences. Any of those alone would cap the entry below Strong.

Why not the tier below (Emerging). The component acute-response findings are individually robust and independently replicated: the myokine account of exercise interleukin-6, the troponin conversion meta-analysis across 939 participants, the immune-redistribution reading, and the allostatic-load framework are all well established. The exposure-specific guards — smoking, radiation, particulates — rest on Foundational-tier institutional assessments. The lever map routes to entries several of which carry stronger evidence than this hub does. This is not a speculative entry; it is a solid entry whose organising frame is softer than its parts.

Per-sub-area split. The acute-panel component findings are Strong individually, while their composite framing is a synthesis and is presented as such. The signal-is-the-mechanism claim is Moderate on the human trial and Emerging on the mitohormesis mechanism, with an explicit species caveat. The hormesis dose-response principle is Moderate on curve existence and openly contested on generality. The three conditions are Moderate as a practical frame and weak as a falsifiable hypothesis. Overtraining syndrome is Moderate as a concept with a major diagnostic caveat. Allostatic load is Strong as a framework with softer quantitative footing than its ubiquity implies. Supercompensation is Emerging and explicitly labelled a model. The no-hormetic-dose guards are Foundational for smoking, particulates and radiation, Strong for alcohol and sleep restriction, and Moderate for chronic psychological stress.

Citation audit state (July 2026 adversarial pass). Seven citations were re-verified against primary bibliographic records: Regwan 2010 (all four statistics confirmed exactly), Ristow 2009 (sample composition and doses confirmed exactly), Shrader-Frechette 2010 in Synthese (authorship established, replacing a previous unverified flag), Van Dongen 2003 (confirmed, flag retired), Seery 2010 (confirmed, and a design mischaracterisation corrected — the outcomes were longitudinal, only the exposure was recalled), Biddinger 2022 (confirmed, and an overstatement of the genetic result corrected), and the NCRP linear-no-threshold update (previously misattributed to the wrong author; corrected to Boice 2017). One citation remains not independently verified in this pass — the 2021 World Health Organization air quality guidelines — though its no-threshold position is uncontroversial. The remaining sources are standard institutional and review works whose bibliographic details were not re-pulled in this pass.

Evidence detail

Why This Entry Exists

There is a measurement problem sitting under most of health and fitness, and almost nobody states it plainly. If you draw blood from someone in the middle of a hard training block, the panel reads like illness. Interleukin-6, a signalling molecule that also rises in infection, can climb enormously. Cardiac troponin, the protein used to diagnose heart attacks, crosses the reference limit in a large share of marathon finishers — the pooled estimate across sixteen studies is 51%, but the confidence interval runs from 33% to 69% and the disagreement between studies is extreme, so the honest statement is "a large and very uncertain fraction," not "half." Circulating lymphocyte counts collapse in the hour or two afterwards. Markers of muscle damage and oxidative stress rise. Read cold, that panel describes an infected, immunosuppressed person with cardiac injury. Read as a process — with a second and third measurement days later — it describes someone who is about to be fitter than they were.

That gap between the snapshot and the process is what this entry is about. The scientific name for the underlying pattern is hormesis: a dose-response shape in which a low or moderate dose of something stimulates while a high dose inhibits or harms. It is the reason a training session and an injury are the same event at different doses, and it is also the single most commercially abused idea in the wellness space, because "the dose makes the medicine" can be stretched to sell almost any unpleasant experience.

So this entry does two jobs that pull against each other, and it has to do both honestly. It defends the principle — the acute stress signal is not damage to be suppressed, and blunting it can blunt the adaptation. And it fences the principle — biphasic dose-response curves existing in toxicology does not license seeking out small doses of harmful things, and that inference has a documented history of being funded by parties who wanted low-dose exposure limits relaxed.

What bad advice this protects against, in all directions:
• "Your inflammatory markers are elevated after training, so training is inflaming you — take antioxidants to fix it." This gets the direction of causation backwards. The acute rise is the signal that drives the adaptation. In a randomised human trial, high-dose vitamin C and E supplementation across four weeks of training abolished the improvement in insulin sensitivity that the non-supplemented groups got, and prevented the upregulation of the body's own antioxidant enzymes. The exercise was identical; only the signal was removed. That trial was small and later replication has been mixed, so it is not settled — but the direction of the concern is real, and "suppress the inflammation" is not the obvious good idea it sounds like.
• "Hormesis means a small dose of anything harmful is beneficial." This is the load-bearing error. Biphasic curves occur for some agents at some endpoints; that is a statement about curve shapes, not a general licence. There is no risk-free level of tobacco smoke, no identified threshold for particulate air pollution, and radiation hormesis is explicitly not accepted by the bodies that set radiation protection standards. An agent can stimulate one endpoint at low dose while harming the organism overall, and organism-level net benefit is a separate question requiring separate evidence every single time.
• "All stress is damage — minimise it." The mirror error, and the one that quietly pathologises the exact acute signals adaptation requires. The post-exercise lymphocyte drop was read for three decades as an immunosuppressive "open window" that needed supplementing shut. The current reading, which fits the mechanism better, is that those cells redistribute into peripheral tissues — heightened surveillance, not suppression. Treating every acute perturbation as harm is how people end up avoiding the stressors that would have restored their capacity.
• "What doesn't kill you makes you stronger." The adversity literature does show a U-shape in some samples — moderate lifetime adversity associated with better later stress regulation than either none or a great deal. But the exposure in that work is recalled rather than measured, replication is mixed, and survivorship problems are obvious, and it is one sentence away from being used to justify traumatic exposure. Aggregate patterns are not prescriptions, and this one must never be applied to trauma.
• "You need to nail the supercompensation window — adaptation peaks at 48 to 72 hours." The supercompensation curve is a coaching model, not a measured physiological law. Glycogen supercompensation after depletion plus carbohydrate loading is genuinely demonstrated; the generalisation of the same tidy curve to strength, endurance and "fitness" is a drawing convention. Specific timings quoted from it have no general empirical basis.
• "Sleeping less is training your sleep need down." Sleep restriction produces cumulative, dose-dependent deficits with no demonstrated overshoot on resumption. Worse, subjective sleepiness stops tracking the damage while objective performance keeps declining, so the person under chronic restriction is the last to know.

What this entry OWNS: the general principle of adaptive stress; the distinction between an acute, resolving stressor and a chronic, unremitting one; the measurement-timing error that makes an adaptive process look like pathology on a single panel; the three conditions that separate adaptation from damage; the routing map of which stressors have which strength of evidence; and the guard against generalising hormesis into "a little bit of anything bad is good."

What this entry DEFERS: every individual lever's protocol, dose, evidence quality and contraindications belongs to its own entry. This hub says "heat is a real hormetic stressor with reasonable human evidence, go here"; it does not re-argue sauna temperatures, cold water durations, fasting windows, zone 2 prescriptions or polyphenol doses. Mitochondrial detail defers to mitochondrial_health. Recovery programming defers to deload_and_recovery_management and overtraining_recovery_management. The psychological-stress version of this argument defers to stress_response_is_adaptive_but_context_dependent.

Evidence

Read the tiers, not the thesis. The thesis of this entry is a synthesis and sits at Moderate confidence overall. Individual components under it range from Foundational to Emerging, and the difference matters more than the overall label.

Sub-area A: The acute panel reads like pathology (Strong, with an important structural caveat)

1. [Strong] Plasma interleukin-6 can rise up to roughly 100-fold during exercise, produced by contracting skeletal muscle itself rather than by the classic inflammatory cascade, in proportion to duration, intensity and muscle mass recruited. The exercise IL-6 response is not preceded by a rise in tumour necrosis factor alpha, which is what distinguishes it from the sepsis pattern. It peaks at or shortly after cessation and returns to baseline rapidly. The 100-fold figure is the ceiling for marathon-scale events, not a typical session — ordinary training produces far smaller rises, and the qualifier is routinely dropped when this number is quoted. Note also that IL-6 is genuinely dual: an acute spike is adaptive signalling, while a chronically elevated floor is a cardiometabolic risk marker. Same molecule, opposite meanings, distinguished by time course. Pedersen & Febbraio 2008, Physiological Reviews. Cui bono: the exercise-physiology field gained a major funding narrative from the muscle-as-endocrine-organ framing, and Pedersen's group has been criticised for promoting the concept strongly. Pulling the other way, companies developing IL-6-blocking drugs benefit from IL-6 reading as uniformly pathological, and supplement sellers benefit from post-exercise inflammation reading as damage requiring their product.

2. [Strong] Cardiac troponin, the standard marker used to diagnose myocardial infarction, converts from a normal pre-race value to an elevated post-race value in a substantial share of endurance-event finishers. A systematic review and meta-analysis of 16 studies covering 939 marathon participants found a pooled prevalence of 51%, with a confidence interval spanning 33% to 69%. Heterogeneity across studies was extreme (I² = 98%), so the pooled figure is a weak central estimate — individual study results span a very wide range, and assay generation, draw timing and cut-off choice drive most of the spread, with the highest conversion rates coming from high-sensitivity assays. Quote the range and the heterogeneity, not the midpoint. The elevation typically resolves within one to three days. Reduced renal clearance during and after prolonged exertion has been proposed as a partial contributor rather than the elevation being purely cardiac in origin — that proposal was not independently verified in our pass and is not from the meta-analysis above, so treat it as a hypothesis rather than a finding. Important honesty point: "transient and benign" is the mainstream reading, but it is inferred from the fact that values resolve, not from long-term outcome data in these cohorts. It is not proven harmless. Regwan et al. 2010, Journal of Interventional Cardiology 23(5):443-450 — bibliographic details and all four figures re-verified against the primary record in this audit. Cui bono: endurance-sport culture benefits from "your troponin is meaningless, keep racing"; cardiology screening services and exercise-is-bad-for-your-heart contrarian media benefit from treating every post-race troponin as occult damage. Both readings outrun the data.

3. [Strong] The sharp fall in circulating lymphocyte number and function one to two hours after vigorous exercise reflects transient, time-dependent redistribution of immune cells into peripheral tissues rather than suppression. A review of the "open window" hypothesis concluded that there is limited reliable evidence that vigorous exercise raises opportunistic-infection risk, that post-exercise salivary immunoglobulin A changes do not mark immune suppression, and that labelling acute exercise immunosuppressive is a misconception. This is the cleanest worked example of the entry's thesis: the identical measurement supported the opposite conclusion for three decades, and what changed was reading the process rather than the snapshot. Honest limitation: this is a narrative review arguing a revisionist position, not a meta-analysis. The field is not unanimous, self-reported upper-respiratory symptoms after mass-participation events are real (though frequently not confirmed as infection when swabbed), and the "no increased infection risk" claim rests substantially on the weakness of the opposing evidence rather than on strong positive evidence of no effect. Campbell & Turner 2018, Frontiers in Immunology. Cui bono: the athlete immune-support supplement market — vitamin C, zinc, glutamine, colostrum — depends entirely on the open window being real and dangerous, which makes it the single clearest financial stake in this question. Pulling the other way, endurance-sport promotion and caveat-free "exercise more" messaging benefit from the revisionist reading.

4. Structural caveat, and it constrains how this entry may be cited. The composite claim — "the whole panel looks unwell" — is a synthesis across separate literatures with different designs, populations and timepoints. No single study drew a full panel mid-workout. It must be built from the components above and never attributed to one paper. The populations sampled were mostly trained or recreationally active, mostly male, mostly endurance athletes, and most measurements were immediately post-exercise or hours later rather than during. The acute response is benign in health, which is precisely the population these studies enrolled. Untrained, older or cardiovascularly compromised people cannot be assumed to show the same benign trajectory.

Sub-area B: The stress signal is the mechanism, not a side effect (Moderate, on a small trial)

5. [Moderate] In a randomised human trial, 39 healthy young men (19 previously untrained, 20 pretrained) completed four weeks of exercise training with or without vitamin C at 1000 mg/day plus vitamin E at 400 IU/day. Insulin sensitivity measured by hyperinsulinaemic-euglycaemic clamp, plus plasma adiponectin, improved only in the groups not taking antioxidants — in both the untrained and pretrained strata, at P<0.001. Antioxidant supplementation also prevented the exercise-induced upregulation of the body's own defence enzymes (superoxide dismutases 1 and 2, glutathione peroxidase 1) and of the PGC1-alpha, PGC1-beta and PPAR-gamma transcriptional programme. The design is the argument: hold the exercise constant, remove the oxidative-stress signal, lose the adaptation. Limitations are serious and must travel with the finding — roughly ten people per arm after a four-way split, single centre, young healthy men only, four weeks, no clinical endpoints. Subsequent human trials have been inconsistent: some found blunting of specific adaptations, particularly mitochondrial biogenesis markers and some endurance measures, others found no effect on performance or hypertrophy. The defensible statement is that high-dose antioxidant supplementation around training may blunt some adaptations, that the mechanism is plausible and partly demonstrated, and that the 2009 result has not been cleanly replicated at scale. These were supraphysiological supplement doses; this does not indict antioxidants from food. Ristow et al. 2009, PNAS — sample composition and doses re-verified against the primary record in this audit. Cui bono: the vitamin supplement industry is directly threatened by this finding, so expect motivated criticism from that direction — some of it methodologically fair given the sample size. Pulling the other way, exercise purism and sports-nutrition brands now selling "timing-aware" or low-dose antioxidant products both benefit from the finding standing.

6. [Emerging] Mitohormesis — the proposal that a modest, transient rise in mitochondrial reactive oxygen species triggers a compensatory adaptive programme leaving the organism more resilient than before — is the mechanistic core linking hormesis to exercise. Species overreach warning, stated explicitly: the strongest mitohormesis evidence, including lifespan extension from mild mitochondrial stress, comes from C. elegans and rodent work. Human evidence exists for the signalling step (see claim 5), but human lifespan or healthspan claims from mitohormesis are not established and must never be framed as human proof. The literature is also concentrated in one lab lineage, its originator's. Ristow & Zarse 2010, Experimental Gerontology. Cui bono: longevity supplements and the whole hormetic-stressor commerce — sauna, cold plunge, fasting products — lean on mitohormesis as their mechanism story. Pulling the other way, the antioxidant supplement industry's entire premise is inverted by it. Detail defers to mitochondrial_health.

Sub-area C: Hormesis as a dose-response principle (Moderate, and contested where it matters)

7. [Moderate] A large assembled database of dose-response relationships shows a biphasic pattern — modest stimulation, typically around 30% to 60% above control, over a limited dose range below the toxic threshold, with inhibition above it. Its principal proponent argues this is a common and perhaps the most fundamental dose-response form in toxicology and pharmacology. The concept has a long history, having lost to the threshold model in the early twentieth century. The existence of biphasic curves is not seriously disputed. What is disputed is their generality and, far more consequentially, their regulatory implications. The database was assembled by an advocate applying his own inclusion criteria, which is a real selection-bias concern, and demonstrating a biphasic curve in vitro or at a single endpoint does not establish net organism-level benefit. Calabrese 2008, Environmental Toxicology and Chemistry. Cui bono: see claim 8 — this is the live conflict. Pulling the other way, environmental and public-health advocacy benefit from precautionary threshold-free models, and regulators have institutional investment in existing risk frameworks that hormesis adoption would force them to revise at cost.

8. [Moderate] The conflict-of-interest critique is documented in the peer-reviewed literature, not merely alleged in commentary. A philosophy-of-science analysis by Kristin Shrader-Frechette, published in Synthese in 2010, examined hormesis as a case of what she calls special-interest science — science with welfare consequences funded by interested parties to establish a predetermined conclusion — described its leading advocate as chemical-industry-funded, and argued that the chemical industry funds hormesis-supporting work because it seeks deregulation of toxic emissions and avoidance of costly cleanups. Her paper's narrower technical charge is that the advocacy conflates three distinct concepts of hormesis. Separately, the same advocate has argued that the linear-no-threshold radiation model was adopted on the basis of a historical mistake and, in later work, misconduct by early radiation geneticists — claims that have drawn substantial published rebuttal, though we did not verify specific rebuttal papers in this pass. Two things must be held together here. A funding critique is not a refutation; the curves stand or fall on their own evidence, and disclosure of industry funding does not prove distortion in any specific paper. Shrader-Frechette is herself a prominent environmental-justice philosopher arguing a position, not a neutral referee. But the dispute is live, it is about regulatory consequence rather than only curve shape, and pretending otherwise would be dishonest. Shrader-Frechette K (2010), Synthese 177(3):449-469, doi:10.1007/s11229-010-9792-5 — authorship and venue re-verified in this audit; the previous "authorship unverified" flag is retired. Cui bono, both ways: chemical manufacturers, the nuclear industry and any party facing low-dose exposure limits would benefit materially from hormesis becoming a regulatory principle. Plaintiff litigation, environmental organisations and the agencies whose authority rests on precautionary models benefit from it not becoming one. Both stakes are real and neither settles the science.

Sub-area D: The three conditions and what happens when one is removed (Moderate)

9. [Moderate] Three conditions separate adaptive stress from damage: the stressor is acute or intermittent rather than continuous, it is matched to current capacity, and it is followed by adequate recovery. Each leg has independent support — hormesis curves are defined by dose; the exercise, heat, cold and fasting literatures all describe adaptation to repeated bouts with return to baseline between them; training prescription is universally relative to current capacity because the same absolute load is trivial for one person and injurious for another; and removing recovery produces the overtraining pattern rather than the adaptation pattern. But the triad is Realised's articulation, not a quoted result from any one paper, and no single source should be cited for it. Its honest weakness is that "adequate recovery" has no validated objective threshold and is defined after the fact by whether adaptation occurred, which makes the frame partly circular and hard to falsify. "Matched to capacity" is likewise not operationalised by any accepted metric. It is a good decision aid and a poor scientific hypothesis, and this entry presents it as the former. Synthesis; support from McEwen & Stellar 1993 on allostatic load and from the overtraining consensus literature. Cui bono: coaching, recovery technology, heart-rate-variability wearables and the entire readiness-monitoring market monetise the recovery leg specifically and have an interest in it being both essential and measurable. Pulling the other way, high-volume training culture and "more is more" fitness commerce benefit from the recovery leg being soft.

10. [Moderate] Overtraining syndrome is what this pattern looks like with the recovery leg removed — the clean negative control. It is defined as exhaustion and persistent fatigue arising from an imbalance between training strain and the athlete's tolerance, where load continues without adequate rest and the athlete enters prolonged maladaptation with adverse changes in performance and in biological, neurochemical and hormonal regulation. It is framed in the consensus literature as an unsuccessful attempt to cope with the total load of training plus life. Major caveat: overtraining syndrome has no validated diagnostic biomarker or test. It is a diagnosis of exclusion requiring anaemia, thyroid disease, infection, low energy availability and depression to be ruled out, and it overlaps heavily with all of them. Prevalence figures are unreliable. The distinction between functional overreaching (adaptive, resolves in days), non-functional overreaching (weeks) and true overtraining syndrome (months) is defined retrospectively by how long recovery took, so the categories cannot be applied prospectively. Use it as an illustration of the principle, not as a crisply diagnosable condition. Meeusen et al. 2013, joint consensus statement of the European College of Sport Science and the American College of Sports Medicine. Detail defers to overtraining_recovery_management.

11. [Strong] Allostatic load names the cumulative cost of repeated or chronic activation of the stress response — the mechanism by which unremitting stress converts adaptation into damage. The core insight fits this entry precisely: the same mediators that are protective when mobilised acutely and then switched off (cortisol, catecholamines, inflammatory cytokines) become damaging when the response is too frequent, too prolonged, or fails to switch off. Composite allostatic-load indices predict mortality and functional decline in cohort studies. Softer than its ubiquity suggests, though: there is no standard index, so studies use different marker panels and scoring rules and effect sizes are not comparable across them; the evidence is observational and heavily confounded by socioeconomic position, which drives both the chronic stress exposure and the outcomes; and directionality is hard to establish. A strong organising framework with softer quantitative footing than it appears to have. McEwen & Stellar 1993, Archives of Internal Medicine; elaborated McEwen 1998, New England Journal of Medicine. Cui bono: stress-management, corporate wellness and biomarker-panel commerce benefit from load framing, as do researchers arguing for structural determinants of health. Pulling the other way, employers and institutions whose working conditions constitute the chronic stressor have an interest in individualising the problem as a resilience deficit rather than accepting the load account.

12. [Emerging] Supercompensation — the model in which post-stressor recovery overshoots baseline before decaying back — is a coaching heuristic, not a validated physiological law, and this entry marks it as such. Glycogen supercompensation after depletion plus carbohydrate loading is genuinely demonstrated. The generalisation of the same curve to strength, endurance and "fitness" as a whole is a modelling convention, and direct empirical validation of the literal curve's timing, magnitude and applicability beyond muscle glycogen is weak. Specific timings quoted from it — the ubiquitous 48-to-72-hour peak — have no general empirical basis and vary enormously by stimulus, tissue and individual. Modern training science has largely moved to fitness-fatigue and dose-response models that fit data better. No citation offered for the general law, because none supports it; the glycogen-specific finding traces to the Bergström and Hultman work of the 1960s. Cui bono: coaching services and periodisation software sell schedule optimisation predicated on the curve being real and predictable. Almost nobody has a commercial interest in debunking it, which is part of why it persists unexamined.

Sub-area E: Where there is no worthwhile hormetic dose (Foundational for the specific exposures)

13. [Foundational] Tobacco smoke: no hormetic dose. Surgeon General's reporting states there is no risk-free level of exposure and that even brief or low-level exposure causes immediate cardiovascular harm. The cardiovascular dose-response is notably supralinear at the low end — the first few cigarettes per day carry a disproportionate share of coronary risk, which is the opposite of a hormetic curve. Light and intermittent smoking carries substantial excess mortality. The genuine anomaly repeatedly cited by hormesis advocates is the replicated inverse association between smoking and Parkinson's disease and ulcerative colitis; these associations are real, partly explained by nicotine pharmacology and possibly by reverse causation, and in no case produce net benefit — the mortality arithmetic is overwhelming. That is the honest exception that demonstrates the framework rather than breaking it: an endpoint-specific inverse association is not organism-level hormesis. US Surgeon General's Reports, 2010 and 2014. Cui bono: the tobacco industry has an extensively documented history of funding dose-response and threshold research to argue low exposures are harmless. That history is the historical template for industry misuse of hormesis, and this entry says so plainly. Pulling the other way, tobacco control funding and litigation depend on the no-safe-dose position.

14. [Strong] Alcohol: the protective J-curve has been substantially undermined and there is no established hormetic dose worth chasing. Two independent lines converge. On method, the observational J-curve is substantially attributable to abstainer misclassification and sick-quitter bias, because "non-drinker" reference groups contain former drinkers who quit because of illness alongside lifelong abstainers who differ systematically in health and socioeconomic position; when studies apply strict measures to remove abstainer bias, the apparent low-volume protection largely disappears. On genetics, Mendelian randomisation studies using variants affecting alcohol metabolism as exposure proxies — and therefore not subject to the same behavioural confounding — have generally found no cardiovascular benefit at low intake and monotonically rising harm. The largest such analysis, in 371,463 UK Biobank participants of mean age 57, found that light-to-moderate drinkers had systematically healthier lifestyle characteristics and that adjusting for these attenuated the apparent cardioprotective association; its non-linear genetic analysis found risk increasing at every level of intake — only minimally across the light range, then rising exponentially at higher intake — with no protective nadir anywhere on the curve. State that shape precisely: the finding is "no nadir," not "steeply harmful at one drink." Limits both ways: Mendelian randomisation is not confound-free (pleiotropy; gene-environment correlation), the analysis above is a European-ancestry cohort while the strongest metabolic instruments are concentrated in East Asian populations, so generalisability cuts in both directions, and some such studies do show a small nadir. The defensible statement is that the cardioprotective J-curve is no longer a safe assumption and that no health authority now recommends starting to drink for health — not that one drink is proven harmful in a given individual. Cancer risk, unlike cardiovascular, shows no threshold at any level, and overall-mortality and cardiovascular-specific curves behave differently and are frequently conflated. Biddinger et al. 2022, JAMA Network Open — sample, design and the non-linear result re-verified against the primary record in this audit. Cui bono: the alcohol industry, extensively — the NIH moderate-drinking trial was terminated in 2018 after investigation found NIH staff had solicited industry funding, the clearest documented case of commercial interest shaping this evidence base. Wine-region economies and hospitality benefit too. Pulling the other way, temperance-aligned advocacy benefits from the revisionist position, as do researchers whose careers now rest on it, and the reassessment has itself been criticised as overcorrection.

15. [Strong] Sleep deprivation: no worthwhile hormetic dose, and deliberate sleep restriction is not a training stimulus. Restriction produces cumulative, dose-dependent deficits in cognition, glucose tolerance, appetite regulation, immune function and cardiovascular risk, with no demonstrated adaptation or overshoot on resumption — recovery sleep restores function toward baseline rather than above it, and some deficits recover incompletely over short recovery windows. The canonical dose-response experiment restricted healthy adults to four or six hours in bed nightly for fourteen days and found cumulative, dose-dependent deficits across every cognitive task tested, with the six-hour condition producing impairment equivalent to up to two nights of total sleep deprivation. The most practically important finding is the dissociation: subjective sleepiness rose acutely and then only slightly further, and did not distinguish the six-hour from the four-hour condition, while objective performance kept declining — so the person under chronic restriction cannot perceive their own deficit. Honest exception: acute total sleep deprivation produces a rapid but transient antidepressant response in a substantial fraction of people with major depression, relapsing after recovery sleep. That is a clinical chronotherapeutic, studied as such, and it constitutes no hormetic case for sleep restriction in healthy people. Van Dongen, Maislin, Mullington & Dinges 2003, Sleep 26(2):117-126 — re-verified against the primary record in this audit; the previous "unverified" flag is retired. Cui bono: hustle culture, shift-work employers and military or performance contexts benefit from sleep need being framed as trainable. Pulling the other way, the very large sleep-product and sleep-tracking market benefits from maximal alarm about any sleep loss, which can overstate the harm of one bad night. Training-specific detail defers to sleep_for_training_recovery_and_adaptation.

16. [Foundational] Ambient air pollution: no hormetic dose and no identified threshold below which particulate exposure is harmless. World Health Organization air quality guidance lowered recommended limits and states that adverse effects are observed below previous guideline levels with no evidence of a no-damage threshold. Large cohort studies consistently find mortality associations extending below existing regulatory limits, and the low-concentration end of the exposure-response curve is, if anything, steeper than linear for cardiovascular endpoints. Honest qualification: at the very lowest concentrations, exposure measurement error and residual confounding by socioeconomic position are substantial, so "no threshold identified" partly reflects the resolving power of epidemiology rather than proof of harm at any dose. WHO Global Air Quality Guidelines, 2021 — this is the one citation in the entry whose bibliographic details were not independently verified in our audit pass, though the no-threshold position itself is uncontroversial. Cui bono: industrial emitters and vehicle manufacturers have a direct, heavily documented interest in threshold models, and low-dose hormesis arguments have been deployed in exactly this regulatory arena — which is the concrete link between an abstract dose-response dispute and real-world harm. Pulling the other way, regulators, environmental advocacy and the air-purifier market benefit from no-threshold framing.

17. [Foundational] Ionising radiation: radiation hormesis is not accepted by the bodies that set radiation protection standards. The US National Academies' BEIR VII report concluded that the preponderance of available experimental information does not support the contention that low levels of ionising radiation have a beneficial effect, and that assuming stimulatory effects would produce net benefit exceeding detriment at the same dose is unwarranted. A subsequent update on how the National Council on Radiation Protection uses the linear-no-threshold model reviewed the recent epidemiology and concluded that linear-no-threshold remains the most appropriate practical basis for radiation protection — the best of the available models rather than a proven one — and the main international bodies continue to use it for protection purposes; petitions to replace it with a hormesis model have been repeatedly rejected. Symmetry that must be stated: linear-no-threshold is itself a precautionary regulatory extrapolation rather than a directly demonstrated dose-response at very low doses, because epidemiology cannot resolve effects near background. A minority of researchers continue to argue for a threshold or for hormesis. The accurate statement is that radiation hormesis is not accepted and is not the basis of any protection standard, that linear-no-threshold is the operating model, and that the very-low-dose region is genuinely uncertain. Not "hormesis is disproven." National Research Council, BEIR VII Phase 2, 2006; Boice JD Jr 2017, International Journal of Radiation Biology 93(10):1079-1092 — the NCRP-update citation was previously misattributed to a different author and has been corrected against the primary record in this audit. Cui bono: the nuclear industry and radiological waste interests would face dramatically lower compliance costs if low-dose limits were relaxed — a large, direct stake, and hormesis advocacy in radiation has been funded from that direction. Pulling the other way, linear-no-threshold sustains regulatory apparatus, remediation contracting and the radiation-protection professions, and anti-nuclear advocacy benefits from maximal low-dose risk estimates.

18. [Moderate] Chronic psychological stress: no hormetic dose — but acute, controlled psychological challenge is genuinely adaptive, and conflating the two is the common error in both directions. Chronic, uncontrollable, unremitting stress is the paradigm case of allostatic load and is associated with cardiovascular, metabolic, immune and psychiatric harm. Acute controllable stress with recovery behaves oppositely: the stress-inoculation and challenge-versus-threat literatures find that moderate, manageable, time-limited adversity followed by adequate recovery is associated with better subsequent stress regulation than either no adversity or high cumulative adversity. The most-cited support is a multiyear longitudinal study of a US national sample, which found U-shaped relationships in which some but non-zero lifetime adversity predicted lower distress, lower self-rated impairment, fewer post-traumatic stress symptoms and higher life satisfaction over time, compared with both high adversity and none. The variables that flip the sign are controllability, duration and recovery — exactly the three conditions, applied to psychological load rather than physical. Serious limits, stated precisely: the outcomes were tracked prospectively but the adversity exposure was recalled, so this is a longitudinal study on a retrospectively reported exposure — recall bias is real, and survivorship is a genuine problem because people destroyed by adversity are less likely to be in the sample — and the U-shape has been debated with mixed replication. This must never become "what doesn't kill you makes you stronger" as a general claim, and it must never be applied to trauma. Seery, Holman & Silver 2010, Journal of Personality and Social Psychology 99(6):1025-1041 — re-verified against the primary record in this audit, including the design correction above; the previous "unverified" flag is retired. This remains the most easily misused item in the entry. Cui bono: resilience-training and corporate-wellness commerce use the adaptive-stress framing to locate responsibility in individuals rather than in the conditions producing the chronic stress. Pulling the other way, over-pathologising all stress serves therapy and anxiety-product markets, and pathologising the ordinary acute stress response is precisely the failure mode this entry names. Detail defers to stress_response_is_adaptive_but_context_dependent.

Mechanism

The signal is the point, not the byproduct. The intuitive model of training is that the stress is a cost you pay for the benefit — you damage tissue, then you repair it, and the repair happens to overshoot. The better-supported model is that the perturbation itself is what the cell reads. Contracting muscle releases interleukin-6 as a signalling molecule; mitochondria under load emit reactive oxygen species that act as second messengers; heat raises the temperature at which proteins begin to misfold, and the cell responds by manufacturing chaperones. In each case the "stress marker" is upstream of the adaptation, not downstream of it. This is why the antioxidant-blunting result matters so much despite its small size: it is the one human design that separates the exercise from the signal and shows that removing the signal removes the gain.

Acute and chronic are different signals carried by the same molecules. Cortisol mobilised in the morning and shut off by evening is a functioning system. Cortisol elevated flat across weeks is allostatic load. Interleukin-6 spiking and returning to baseline in hours is myokine signalling. Interleukin-6 elevated as a chronic floor is a cardiometabolic risk marker. The molecule does not carry the meaning; the time course does. Nearly every "is this hormone good or bad" argument in wellness content is a failure to ask over what interval.

Adaptation is specific to the stressor and to the tissue that received it. Heat produces heat-shock proteins and plasma volume expansion. Cold produces brown adipose activation and vasomotor changes. Fasting engages nutrient-sensing and autophagy pathways. Loading produces remodelling in the loaded tissue. There is some cross-talk — the cellular stress-response machinery overlaps — but the idea that any hormetic stressor produces general robustness is an extrapolation, and it is the extrapolation used to justify buying a stressor you have no specific reason to want. Specificity detail belongs to specificity_of_adaptation.

Recovery is not passive downtime; it is where the adaptation is constructed. Protein synthesis, mitochondrial biogenesis, immune resolution and central nervous system restoration all happen in the interval between exposures, and most of them are sleep-dependent. This is why removing recovery does not merely slow progress — it inverts the sign. The same load that was building capacity begins consuming it, which is the overtraining pattern.

Why the evolutionary filter is a reasonable prior and a bad proof. The exposures with good human hormetic evidence — muscular loading, heat, cold, fasting, hypoxia, controlled cognitive and psychological challenge — are precisely those with dedicated evolved recovery machinery: heat-shock proteins, cold-induced thermogenesis, autophagy and nutrient sensing, muscle remodelling, hypoxia-inducible signalling, hypothalamic-pituitary-adrenal regulation. The exposures with no worthwhile hormetic dose are those with no dedicated recovery programme, or where damage accumulates faster than repair, or where population evidence simply shows monotonic harm. That is a useful filter for raising or lowering prior probability. It is not a test that settles cases: ethanol, ultraviolet light and several plant compounds sit awkwardly inside it, having evolved exposure and partial adaptive machinery but net harm at relevant human doses. And the same evolutionary framing is used to sell polyphenols, which is exactly why it must be labelled as a heuristic rather than evidence. Every candidate stressor still needs its own human outcome data.

PRACTICAL APPLICATION — the lever map

This hub routes; it does not prescribe. Each lever below gets one line on what it does and how strong the human evidence is, and then a pointer. Dose, protocol, contraindications and the arguments for and against belong to the linked entry, and nothing here should be acted on without reading it.

Before the map, three questions that apply to every lever. Is the exposure acute and resolving, or continuous? Is the dose matched to what you can currently recover from, rather than to what someone fitter is doing? Is there real recovery after it? If any answer is no, the lever is not currently hormetic for you regardless of what the research says about it in general.

Training load — the best-evidenced hormetic stressor in existence, and the one where the acute panel most dramatically resembles pathology.
• Aerobic base work: builds mitochondrial density and metabolic flexibility at a load low enough to be repeated frequently. Human evidence is strong for the adaptation, moderate for any specific intensity prescription. Go to physical_zone2_cardio.
• High-intensity work and cardiorespiratory fitness: the fitness metric with the most robust association with mortality across the whole literature. Human evidence is strong for the association, weaker for any specific protocol and unresolved on reverse causation. Go to vo2max_and_longevity.
• The stimulus-to-tissue matching problem — why the adaptation you get is the one you trained for. Human evidence is strong. Go to specificity_of_adaptation.
• The muscle-damage signal specifically, and whether soreness means anything. Human evidence is moderate, and mostly negative about soreness as a signal. Go to muscle_soreness_recovery_doms.
• Planned load reduction as part of the stimulus rather than a break from it. Human evidence is moderate and largely from athletic populations. Go to deload_and_recovery_management.
• What it looks like when the recovery leg is removed. Moderate as a concept, with no validated diagnostic test. Go to overtraining_recovery_management.

Heat — repeated passive heat exposure raises core temperature enough to trigger heat-shock protein production and plasma volume expansion. Human evidence is moderate, with the cardiovascular cohort data stronger than the mechanistic claims usually made about it, and the observational studies carrying real confounding. Go to sauna_heat_exposure.

Cold — deliberate cold exposure produces a large acute catecholamine and sympathetic response plus adaptation in thermogenic tissue. Human evidence is moderate for the acute physiology and considerably weaker for the downstream health claims attached to it commercially. There is also a genuine interference question about cold applied immediately after resistance training blunting the adaptation signal — which is the same mechanism as the antioxidant finding above. Go to deliberate_cold_exposure_stress.

Fasting and energy restriction — extended fasting engages nutrient-sensing pathways and autophagy, with the strongest mechanistic data in animals and human evidence that is mixed and often confounded with the effects of weight loss itself. Human evidence is moderate at best for the fasting-specific claims. Go to diet_extended_fasting.

Plant defence compounds — the xenohormesis argument holds that plant stress compounds act as mild stressors in us and trigger adaptive responses. This is the weakest lever on the map and the one most aggressively marketed. Human evidence is emerging at best, and the evolutionary story is much more persuasive than the outcome data. Go to plant_defence_compounds_and_xenohormesis; for two specific worked cases, capsaicin_and_spicy_food_evidence and cruciferous_vegetables_and_goitrogens.

Hypoxia and altitude — intermittent hypoxic exposure triggers hypoxia-inducible signalling and haematological adaptation, and this is a real evolved stress-response pathway. Honest tier, since we route nowhere: human evidence is moderate for altitude and hypoxic training in athletes, where the haematological adaptation is well described but performance transfer is inconsistent, and emerging at best for the consumer intermittent-hypoxia device market, which is far ahead of its evidence. Realised does not currently hold a dedicated entry on this lever, so treat what you read elsewhere with the usual scepticism rather than following a pointer we do not have.

Psychological challenge — acute, controllable, time-limited challenge followed by recovery is genuinely adaptive; chronic uncontrollable stress is the paradigm of harm, and the two get conflated constantly in both directions. Human evidence is moderate and largely observational. Go to stress_response_is_adaptive_but_context_dependent; for the measurement side, hrv_training_for_stress_resilience; for what the failure state looks like, burnout_physiology_and_recovery.

The recovery substrate under all of them — sleep is where most adaptation is constructed, and it is not itself a hormetic lever. Human evidence that restriction produces cumulative dose-dependent harm with no overshoot is strong; there is no useful small dose of sleep deprivation. Go to sleep_for_training_recovery_and_adaptation.

The mechanism layer under most of them — mitochondrial_health owns the reactive-oxygen-species-as-signal detail and the mitohormesis evidence, which is emerging in humans and must carry its species caveat.

And the levers with no hormetic dose, which is the half of the map most content omits: tobacco smoke, ambient particulate pollution, avoidable ionising radiation, alcohol, sleep restriction, and chronic uncontrollable stress. There is no small dose of these to seek out. If a product or protocol is being sold to you on the grounds that "the dose makes the poison," check whether the exposure in question is one your body has machinery to recover from — and then check whether there is human outcome evidence, because the machinery argument alone proves nothing.

Risks And Contraindications

The single largest risk is using this entry to dismiss a real abnormal result. "Your labs lie, trust the process" is a genuinely dangerous framing, and this entry can be misread into it. The acute-response literature was built in healthy, mostly trained, mostly younger, mostly male populations. An out-of-range troponin in someone with chest pain, cardiovascular risk factors, or no recent hard exertion is a clinical finding, not an artefact of the timing of the draw. If a marker is abnormal and you cannot point to the specific recent stressor that explains it and watch it resolve on a repeat, it needs a clinician, not a reframe.

Capacity is individual and the same absolute dose is not the same relative dose. Every lever on the map has a population in whom the ordinary dose is not benign: heat exposure in pregnancy and in cardiovascular disease, cold exposure in cardiac and arrhythmic conditions, fasting in diabetes on glucose-lowering medication, in eating-disorder history, in pregnancy and in the underweight, and high-intensity training in undiagnosed cardiac disease. This hub does not carry those contraindications; each spoke does, and the spoke must be read before the lever is used.

The recovery leg fails silently. There is no biomarker that tells you recovery was adequate. Overtraining syndrome is a diagnosis of exclusion with no confirmatory test and heavy overlap with anaemia, thyroid disease, infection, low energy availability and depression — which means the correct response to persistent unexplained fatigue is medical assessment, not a longer deload.

Stacking stressors is not additive in the way people assume. Training, heat, cold, fasting and life stress draw on overlapping recovery capacity. Adding a second hormetic lever to a schedule that is already at the limit of what recovery supports converts both into load. The commercial incentive runs entirely the other way, because every lever is sold separately.

Do not apply the psychological version to trauma. The adversity findings are aggregate, rest on recalled exposure, and are survivorship-affected. Nothing in this entry supports the idea that a traumatic exposure was beneficial, and using it that way is both unsupported and harmful.

A small dose of a genuine toxicant is not a lever. This is stated again here because it is the failure mode with the worst consequences: no reader should come away seeking low-dose exposure to smoke, alcohol, radiation, or particulates on hormetic grounds.

Controversy

Nature of the dispute. There are actually two arguments here that get fused, and separating them is most of the work. The narrow argument is empirical: do biphasic dose-response curves occur, and how commonly? That one is largely settled in the affirmative and is not very interesting. The broad argument is about consequence: should hormesis replace threshold and linear-no-threshold models as the default assumption in regulatory toxicology, and does the existence of biphasic curves license individuals to seek out low doses of harmful exposures? That argument is live, is heavily entangled with industry funding on one side and regulatory institutional interest on the other, and is where the real disagreement lives.

Position A: hormesis is a fundamental, under-recognised dose-response principle whose neglect distorts both toxicology and health advice. Best evidence for it: the assembled database of biphasic relationships is large and the pattern is reproducible across endpoints and agents; the mechanistic account (a modest perturbation triggering an overcompensating adaptive response) is well characterised in the cellular stress-response literature; and the exercise, heat, cold and fasting evidence in humans is exactly the shape this position predicts. The antioxidant-blunting trial is a genuine point in its favour, because it shows the stimulation is causal rather than incidental. Where it overreaches: the database was assembled by an advocate with his own inclusion criteria; a biphasic curve at one endpoint does not establish net organism benefit, and the leap from "biphasic curves are common" to "low doses of harmful agents are beneficial" is invalid on its face; and the regulatory version of the claim would have enormous financial consequences for the parties who have funded some of the work, which is a reason for elevated rather than ordinary scrutiny.

Position B: hormesis is a real but narrow curve phenomenon that should not touch regulatory policy or individual behaviour around toxicants. Best evidence for it: for the exposures where the argument actually matters — tobacco smoke, particulates, ionising radiation — the bodies that assess the evidence have examined hormesis and declined to adopt it, with BEIR VII stating that the available experimental information does not support net benefit from low-dose radiation; the smoking dose-response at the low end is supralinear, the opposite shape; and the alcohol J-curve, long the strongest popular example of a hormetic exposure, has substantially dissolved under abstainer-bias correction and Mendelian randomisation. Where it overreaches: linear-no-threshold is itself a precautionary extrapolation rather than a demonstrated low-dose response — the most recent institutional restatement defends it as the best available practical model, not as a measured fact — and defending it as though it were proven is not honest; the very-low-dose region is genuinely unresolved in both directions; and the strong form of this position slides easily into treating all stress as damage, which is the error that produces antioxidant supplementation around training, immune supplements for a redistribution phenomenon, and general avoidance of the acute perturbations that adaptation requires.

The funding and bias dimension — cui bono, both ways. On the pro-hormesis side the stakes are unusually concrete. A peer-reviewed philosophy-of-science analysis by Kristin Shrader-Frechette characterised the field's leading advocate as chemical-industry-funded and argued the industry supports this work because it seeks deregulation of emissions and avoidance of cleanup costs — an analysis that is itself argued from a position rather than written as a neutral survey. The nuclear industry would face materially lower compliance costs under a hormetic or threshold model. And in the consumer market, every seller of a stressor — cold plunges, saunas, fasting programmes, polyphenol supplements — benefits from the unbounded version of the idea, because it converts "unpleasant" into "therapeutic" without requiring outcome evidence. On the anti-hormesis side the stakes are equally real and less often named. Regulatory agencies have institutional investment in existing risk models and would face costly revision. Remediation contracting, radiation-protection professions and plaintiff litigation all depend on precautionary low-dose estimates. The antioxidant and immune-supplement industries depend specifically on acute exercise stress being read as damage. And the sleep-tracking and recovery-monitoring markets benefit from maximal alarm about any perturbation. Neither set of interests settles anything, and the appropriate response to both is to read the exposure-specific evidence rather than the framework argument.

Realised Position: Hormesis is real as a mechanism and as a description of how training, heat, cold, fasting and controlled challenge work, and the acute stress signal in those contexts should not be suppressed on the assumption that it is damage. We hold that at Moderate confidence for the framework and higher for individual well-evidenced levers. We reject the generalised form completely: the existence of biphasic curves does not license seeking low doses of harmful exposures, and we treat tobacco smoke, ambient particulates, avoidable ionising radiation, alcohol, sleep restriction and chronic uncontrollable stress as having no worthwhile hormetic dose. On the regulatory dispute we take no position beyond reporting it accurately, because it is outside what a health platform should be adjudicating — but we do report the funding on both sides, because a reader who does not know that chemical and nuclear interests have funded hormesis advocacy, and does not know that linear-no-threshold is itself precautionary rather than demonstrated, cannot evaluate what they read elsewhere. The practical rule we teach is the three conditions plus the evolutionary filter, offered explicitly as a prior-setting heuristic rather than a test, with human outcome evidence required for every individual lever regardless.

Cross-Pillar Connections

• mitochondrial_health — owns the reactive-oxygen-species-as-signal mechanism, mitochondrial biogenesis, and the mitohormesis evidence including its species limits. This entry states the principle that a modest transient stress signal drives adaptation and defers all mitochondrial detail there.
• deliberate_cold_exposure_stress — owns cold protocols, doses, the acute catecholamine response, the post-training interference question and cardiac contraindications. This entry places cold on the lever map and defers all of it.
• sauna_heat_exposure — owns heat protocols, heat-shock protein and plasma-volume adaptation, the cardiovascular cohort evidence and its confounding. This entry places heat on the map and defers.
• diet_extended_fasting — owns fasting windows, autophagy and nutrient-sensing evidence, the weight-loss confound and the contraindications. This entry places fasting on the map and defers.
• physical_zone2_cardio — owns aerobic base prescription and the mitochondrial-density case for low-intensity volume. This entry names it as the highest-frequency hormetic dose and defers.
• vo2max_and_longevity — owns the cardiorespiratory-fitness-to-mortality evidence and its reverse-causation problems. This entry defers the entire longevity argument there.
• deload_and_recovery_management — owns planned load reduction, deload structure and periodisation. This entry owns why recovery is a condition of adaptation; the programming is there.
• overtraining_recovery_management — owns overtraining and overreaching identification, the diagnosis-of-exclusion problem and management. This entry uses it as the illustration of the recovery leg failing and defers everything clinical.
• muscle_soreness_recovery_doms — owns delayed-onset soreness, what it does and does not indicate, and the interventions marketed at it. This entry defers the whole question of whether soreness signals anything.
• specificity_of_adaptation — owns the principle that adaptation matches the stimulus and the tissue. This entry names specificity as the reason general-robustness claims for hormetic stressors are an extrapolation, and defers the detail.
• stress_response_is_adaptive_but_context_dependent — owns the psychological stress response, controllability and appraisal. This entry owns the acute-versus-chronic distinction as a general principle and defers the psychological application there.
• hrv_training_for_stress_resilience — owns heart-rate variability as a measure, what it tracks and its limits. This entry notes that no validated readiness measure closes the recovery-leg gap, and defers the measurement argument.
• burnout_physiology_and_recovery — owns burnout as the chronic-stress endpoint and its physiology. This entry uses allostatic load as the mechanism and defers the clinical picture.
• sleep_for_training_recovery_and_adaptation — owns sleep's role in constructing adaptation and the training-specific evidence. This entry owns the guard that sleep restriction has no hormetic dose and defers everything about how sleep supports adaptation.
• plant_defence_compounds_and_xenohormesis — owns the xenohormesis argument, the polyphenol evidence and its weakness. This entry places it on the map as the weakest lever and explicitly declines to re-argue it.
• capsaicin_and_spicy_food_evidence — owns capsaicin as a worked case of a plant defence compound with human data. This entry defers entirely.
• cruciferous_vegetables_and_goitrogens — owns the glucosinolate case, including where the defence-compound story cuts against benefit. This entry defers entirely.

What would change our mind

Falsifiability: explicit upgrade/downgrade criteria from source

On the framework. Large, well-powered replication of the antioxidant-blunting design — a multi-centre trial with adequate arms, mixed sex, older participants and clinical rather than surrogate endpoints — would move the "the signal is the mechanism" claim from Moderate toward Strong if it confirmed, and would substantially weaken it if a properly powered trial found no blunting of any meaningful adaptation. That is the single most decisive available experiment.

On the three conditions. If someone operationalised "adequate recovery" or "matched to capacity" with a validated, prospectively predictive measure — one that identified maladaptation before performance decline rather than diagnosing it retrospectively — the triad would move from a decision aid to a testable framework. Conversely, if prospective work showed that adaptation proceeds normally across a wide range of recovery conditions and that the recovery leg is largely a coaching folk belief, we would have to demote it. Current wearable readiness scores are not that evidence and should not be mistaken for it.

On mitohormesis. Human intervention data showing that mild mitochondrial stress produces durable healthspan benefit through the proposed signalling route, from a lab lineage other than the concept's originators, would move it from Emerging. Absence of that within a reasonable further window, or human trials that find the signalling step without any downstream benefit, would keep it where it is.

On the exposure-specific guards. For alcohol, a well-conducted Mendelian randomisation study in a population of different ancestry from the existing European-cohort analyses that recovered a clear protective nadir for overall mortality would force us to soften the current framing. For radiation, adoption of a threshold or hormetic model by the main protection bodies after formal re-evaluation would change what we report, though we would report their reasoning rather than endorse it. For smoking and particulates we cannot currently imagine evidence that would establish a beneficial dose, and we say so rather than pretending to an open mind we do not have.

What would NOT move us. A new tranche of biphasic dose-response curves added to the existing database, because generality of the curve shape is not the disputed question. Any single mechanistic or in-vitro demonstration that a toxicant stimulates some endpoint at low dose, because endpoint stimulation is not organism-level benefit and this is precisely the inference we refuse. Invertebrate or rodent lifespan extension from any hormetic stressor, because the species gap is the specific thing being papered over when those results are quoted at humans. Commercial claims that a product is "hormetic," which is a mechanism assertion carrying no outcome evidence. Disclosure that a hormesis researcher received industry funding, which is a reason for scrutiny but not a refutation — we would need the specific work shown to be wrong. And equally, an institutional body restating its existing position without new evidence, on either side, since restatement is not evidence.

Industry bias note

Structural incentives the evidence base may reflect

The pro-stressor end. The commercial structure here is unusual and worth naming precisely: hormesis is the only framework that lets a seller charge money for an unpleasant experience without producing outcome evidence, because the unpleasantness is the product claim. Cold plunge equipment, sauna installation, fasting programmes and supplements, breathwork and hypoxia devices, and polyphenol and "activator" supplements all lean on some version of mitohormesis or xenohormesis as their mechanism story. The tell is a mechanism argument standing in for a human outcome argument — "this triggers heat-shock proteins" or "this activates autophagy" rather than "in humans this produced this result." A second tell is stressor stacking, since every lever is sold separately by a party with no interest in your total recoverable load. And at the industrial end, the deployment of hormesis arguments against low-dose exposure regulation is documented rather than speculative, with the chemical and nuclear sectors having direct, large financial stakes in threshold models. That is the same idea being used at two very different scales.

The anti-stressor end. Less discussed and equally real. The athlete immune-support supplement market — vitamin C, zinc, glutamine, colostrum — exists almost entirely because the post-exercise "open window" was read as immunosuppression, which is a reading the mechanism does not support. The antioxidant supplement industry depends on exercise-induced oxidative stress being damage requiring correction, which is the interpretation the blunting trial inverts. Recovery technology, heart-rate-variability wearables and readiness scoring monetise the recovery leg specifically, and have an interest in that leg being both essential and precisely measurable — the first is defensible, the second is not currently supported. Sleep-tracking commerce benefits from maximal alarm about any perturbation. Regulatory agencies, remediation contracting and radiation-protection professions have institutional investment in precautionary models. And corporate wellness sits awkwardly on both ends at once, selling resilience training that uses adaptive-stress framing to individualise a problem that the load framework would locate in working conditions.

The clean signal. The findings with the least commercial contamination are the ones nobody profits from. The measurement-timing observation itself sells nothing — it argues against buying supplements to correct exercise-induced markers, and against buying reassurance that abnormal markers are meaningless. The immune-redistribution reading directly costs the supplement market and benefits no product. The three-condition framework licenses only the free stressors — training, heat if you have access, cold, fasting, sleep, controlled challenge — and rules out most of what is sold. The no-hormetic-dose guard is entirely unfunded, which is exactly why it is underprovided relative to its importance. And on the disputed regulatory question, the most reliable signal is that both sides have material stakes, which means the exposure-specific evidence is worth more than any framework-level argument from either direction.

Sources (16)

moderateLast updated July 2026
Open in the Library: search, filter, every entry →

We set no cookies and run no ad trackers. We count visits with Cloudflare's cookieless, privacy-first analytics. The only thing stored on your device is which example you last viewed.