Alzheimer's is not an event. It is a process in time.
Welcome to Adult Cognitive Disease. This wiki is an attempt at a comprehensive model of the disease, built from the published scientific literature — a century of it — on a platform meant to serve scientific inquiry and the public interest equally.
That model has two axes, time and space: a fifty-year sequence that begins decades before any symptom, and a fixed route through the brain in which one region is reached before the next. It is built from first principles on a single measurement — where neurons and their synapses are lost, and when — which is what allows findings from fields that never cite one another to be placed on the same two axes. The four charts below give the whole argument in a sitting, and the papers that make it sit underneath. If you came looking for one thing in particular — APOE4, the locus coeruleus, ferroptosis, the perineuronal net — it has a page of its own here, and the papers that argue about it.
The evidence underneath is published, peer-reviewed research, cited paper by paper and checked against the record rather than recalled. Where it is thin, the papers grade it and say so. Where the frameworks contradict one another, they say that too. And where a question is genuinely open, they name the experiment that would settle it — which is the part most likely to be worth something to anyone actually working on this.
The reading was done by frontier language models, under a method called Organic Network Synthesis. What that produces is a synthesis of the literature, not a machine’s opinion of it. Where it interprets a living research programme it goes back to the researchers behind it — dozens whose work is cited here have read it, and the corrections have come at the margins rather than the core. The list of researchers consulted is available on request. The citations themselves are checked by machine and the check is repeatable: every identifier this site publishes — 3,480 DOIs and PubMed IDs — has been resolved against PubMed or Crossref and compared with the reference printed beside it. None points to a different paper. It is still a map rather than a verdict: the phases, and the bridges between them, are hypotheses, argued in the papers and graded there.
The Sin of Silos
This is not a paper about the disease. It is a paper about the field that studies it, and about why this project had to exist at all. Every component of the thesis above had been described in the published literature by the mid-1980s; assembling it took seventy years.
The Sin of Silos asks why, and answers from the field's own record: three research communities — noradrenergic, glial, matrix — that barely intersected with mainstream Alzheimer's research between 1950 and 2018, while the amyloid cascade absorbed the funding, the journal pages and the prestige. It reads 151 independent prize submissions, the NIH award record for 2000–2020 and the citation structure around them, argues the thesis was assembleable by 2005, and puts a scenario-based figure on what the delay cost, in person-years of dementia. The argument is diagnostic, not accusatory.
Everything else on this site is the attempted remedy.
Read The Sin of Silos →The shape of the disease
Four charts · start hereAlzheimer’s is usually introduced as a list of things found in a brain after it has stopped working: plaques, tangles, shrinkage. A list has no shape. This corpus argues the disease does have one — that it happens in a particular order, over a particular stretch of a life, moving through particular places.
How to read these four charts
They are diagrams, not measurements. No one has watched a single brain lose its connections for sixty years, and these curves are not a record of anyone doing so. What they draw is the order and the shape this corpus argues for — which place is reached before which, what falls early, what falls late, and how far apart the two ends of that sequence sit. The heights are illustrative. The sequence is the claim, and it is a claim, not a settled fact. Two further charts, folded into the section on symptoms below, draw an operation rather than an order: what a working repair looks like, and what its absence looks like. In every case: the evidence behind each step, and how firmly it is held, is in the papers these charts summarise.
Where each claim comes from. The sequence and its two transitions: The Temporal Architecture of Collapse. Phase I, the brainstem and its bioenergetics: Bioenergetic Collapse and the locus coeruleus. Phase II, the microglial turn: Homeostatic Microglial Collapse. Phase III, the matrix and the synapse: perineuronal nets. The same curves with the underlying data, cell type by cell type: Collapse Cascades. Every claim above is graded where it is argued.
1It begins long before anyone notices
The first thing lost is not a memory. It is a connection — one of the roughly hundred trillion junctions, called synapses, where a brain cell passes a signal to the next one. Losing them is not an event. It is a slow subtraction, and on this account it starts in early adulthood, in one small nucleus in the brainstem, and runs for decades in complete silence¹,²,⁵.
That silence is the single most important feature of the disease, and the reason it has been so hard to treat. By the time a person is sitting in a clinic being asked to remember three words, the process being named that day has been running for something like forty years².
The long silence
Connections from the locus coeruleus — the earliest place affected — plotted against age.
The shaded stretch is the part of the disease no one sees. On this account, the earliest-affected connections are approaching half gone by the time a person turns 50, and the diagnosis arrives near the far end of the curve — not at its beginning.
2It travels a fixed route
The disease does not appear everywhere at once, and it does not appear at random. It starts in one place and moves along the wiring to the next, in an order that is much the same from brain to brain³,⁴. That regularity is the reason it is possible to talk about a shape at all.
The route begins in the locus coeruleus — a speck of tissue in the brainstem, no bigger than a grain of rice, that supplies the whole cortex with noradrenaline and does the work of waking you and holding your attention. The second place reached is its counterpart, the nucleus basalis, which supplies acetylcholine. Both are small, ancient nuclei that send a single thin axon an enormous distance to keep the rest of the brain in a working state; both are, on this account, in trouble before the structures they serve.
Only then does the disease appear in the places that produce the symptoms — the doorway into memory, memory itself, the amygdala — and last of all in the neocortex, the thinking surface, where the losses everyone recognises finally show up.
A fixed route through the brain
The age at which each place has lost about a quarter of its connections.
Read top to bottom, this is the order — locus coeruleus, nucleus basalis, entorhinal cortex, hippocampus, amygdala, neocortex. Read left to right, it is the timetable: roughly 22 years separate the first station from the last, which is why changes in sleep, mood and arousal can precede a memory complaint by decades and be dismissed as ordinary ageing. The rows are sorted by the curves themselves, not by assertion.
3Every place keeps its own clock
Put those six places side by side and the disease stops looking like a switch and starts looking like a wave. Each panel below is one place, drawn against the other five in grey. The curve moves rightward down the sequence: the brainstem is already well down its slope while the neocortex has barely started.
This is what is meant by the disease having a shape in time and space at once. There is no single moment at which someone “gets” Alzheimer’s, and no single place where it lives. There is a front, and it moves.
The wave, place by place
The same measure, the same axes, six times over.
Grey lines are the other five places, repeated in every panel so each can be read against the whole set. The coloured line is the panel’s own.
4And inside each place, it is selective
Zoom into a single region and a fourth dimension appears. The disease does not thin out a piece of cortex evenly, like paint fading. It takes specific connections from specific kinds of cell⁶, and leaves others comparatively alone — for a while.
The clearest case is the PV interneuron, a cell whose job is timing: it inhibits its neighbours in fast, precise rhythm and keeps cortical activity from running away with itself. It is wrapped in a protective mesh called a perineuronal net, and for six decades it is the best-protected cell on the chart. Then the mesh is digested, and it becomes the worst-affected one.
Not every cell falls together
The neocortex, broken out by the kind of cell losing its connections.
For most of a life the PV interneuron loses less than anything around it. The crossing point, late on, is the moment its protection fails — and it is the argument for why the last phase of the disease arrives so much faster than the first.
5The first symptoms are the ones nobody counts
The three places the disease reaches first are not memory structures. They are small nuclei that set arousal, motivation and orientation, and each of them fails long before anyone is tested. What that produces is not forgetfulness. It is broken sleep, apathy and getting lost — the three complaints most reliably written off as ordinary ageing.
They are not soft signs. In a cognitively normal person, apathy carries roughly three times the odds of progressing to mild cognitive impairment and twice the odds of progressing to dementia⁷; in someone who already has mild cognitive impairment it raises the hazard of conversion by half again⁸. Each of the three has a nucleus behind it, and a mechanism that can be drawn.
Broken sleepthe locus coeruleus
The nucleus in the first chart is the brain's noradrenergic arousal nucleus. It sets the depth of wakefulness and the architecture of sleep, and it is the first place the disease touches — losing cells from early adulthood⁵.
The direction of that claim matters, and this corpus is careful about it: it does not argue that poor sleep causes Alzheimer's disease. The tau pathology of the locus coeruleus precedes any plausible sleep lesion. The lesion comes first, and disturbed sleep is among the earliest things it produces.
What the disturbed sleep then does is measurable in the other direction. How badly sleep is disturbed forecasts the rate at which amyloid accumulates over the following years⁹ — which makes the interval both a symptom of the lesion and a lever on what comes after it.
The full argument: The Restorative Interval.
Apathy, and the day that does not stickthe ventral tegmental area
A second small midbrain nucleus, the ventral tegmental area, was tied to dementia in 1988¹⁰ and then left largely to Parkinson's researchers.
In amyloid-bearing mice its dopamine neurons begin dying at three months, before a single plaque forms¹¹ — and only the mesolimbic ones. The neighbouring substantia nigra, the nucleus that fails first in Parkinson's disease, is untouched. Every standard explanation of why dopamine neurons are fragile predicts the opposite ranking, so whatever is killing these cells is not the thing those explanations describe.
Dopamine sets the price an organism will pay for a goal, which is why its withdrawal reads as apathy rather than sadness: a reduced willingness to spend effort for a reward. The same loss has a second consequence in the hippocampus, where the dopamine supplies something closer to a permission. A memory that has been formed is not yet a memory that will be kept; whether it persists depends on a dopaminergic signal arriving in the hours around the event¹².
The permission that is never given
The strength of a new memory trace across the day that follows it.
Encoding is not the step that fails; both traces are laid down. Only one is granted the signal that converts it into something which survives the day — the shape of the complaint families actually report. The conversation happened. By evening it is gone.
It is not a one-way street: in mice, preventing the loss of those neurons preserves the memory¹³.
The full argument: The Price and the Permission.
Getting lostthe somatostatin interneuron
The oldest finding in this disease that is not about acetylcholine is that somatostatin — a peptide, and the name of the interneuron that carries it — is depleted from the Alzheimer's cortex. Davies, Katzman and Terry reported it in 1980¹⁴. It has never been overturned, and it has never been placed.
The intuitive account is that these cells maintain the brain's spatial metric — the hexagonal grid in the entorhinal cortex that tracks where you are — and that losing them turns the map to noise. That account is wrong. Silencing somatostatin interneurons leaves grid firing intact; it is the parvalbumin cells that gate the hexagonal code¹⁵. The correction is worth making, because what the cell actually does is more specific, and easier to test.
A position computed from your own motion accumulates error, and that error is cleared on contact with the world — at a wall, at a doorway, at a landmark¹⁶. In the hippocampus, a somatostatin interneuron sets the weighting between the two inputs that matter: the one arriving from entorhinal cortex, which carries the world, and the one from CA3, which carries memory¹⁷. It is the device that alternates between correcting by the world and computing from memory. Health is not an accurate map. It is a repairable one.
The map that cannot be reset
Error in the internal estimate of position, with and without the correction that clears it.
Both lines drift at the same rate. The whole difference is the reset. Lose it and the map is not erased — it becomes unresettable: the landmark is seen, and cannot be bound to the position it ought to correct.
That predicts a particular kind of behaviour, and the measurements agree. More than five thousand unassisted trips by nursing-home residents, recorded on video, were 87 per cent direct travel — efficiency, not quantity, tracked cognitive status¹⁸. A review of 325 community missing-person cases found them "temporally appropriate but spatially disordered", usually beginning as an ordinary permitted errand¹⁹. The walking has an object. It is a search for the landmark that would reset the map.
The full argument: The Homeward Vector.
Why the shape matters
If this account is right, then most of what makes a person recognisably ill happens at the far right of these charts — and almost every drug ever trialled has been given there, in the last few years of a forty-year process. The shape suggests the target is further left.
It also carries a genuinely hopeful implication. What these charts plot is connections, not cells. A silenced neuron that has lost its inputs is in a different condition from a dead one, and this corpus argues that a good deal of what is lost in the late phase is the former — cells that have stopped speaking rather than cells that have gone. Whether that is recoverable is an open question, and one of the most consequential in the field.
The full argument, with the evidence for each step and a grade on how firmly it is held, is in The Temporal Architecture of Collapse. To move through the same material yourself, the interactive Temporal Chart lets you pick any place and take the phases apart.
SourcesNineteen references, each resolved against PubMed or Crossref
Braak H, Del Tredici K. The pathological process underlying Alzheimer's disease in individuals under thirty. Acta Neuropathologica 2011;121(2):171–181. DOI 10.1007/s00401-010-0789-4.
Braak H, Del Tredici K. The preclinical phase of the pathological process underlying sporadic Alzheimer's disease. Brain : A Journal of Neurology 2015;138(Pt 10):2814–2833. DOI 10.1093/brain/awv236.
Braak H, Braak E. Neuropathological stageing of Alzheimer-related changes. Acta Neuropathologica 1991;82(4):239–259. DOI 10.1007/BF00308809.
Braak H, Thal DR, Ghebremedhin E, et al. Stages of the pathologic process in Alzheimer disease: age categories from 1 to 100 years. Journal of Neuropathology and Experimental Neurology 2011;70(11):960–969. DOI 10.1097/NEN.0b013e318232a379.
Theofilas P, Ehrenberg AJ, Dunlop S, et al. Locus coeruleus volume and cell population changes during Alzheimer's disease progression: A stereological study in human postmortem brains with potential implication for early-stage biomarker discovery. Alzheimer's & Dementia : the Journal of the Alzheimer's Association 2017;13(3):236–246. DOI 10.1016/j.jalz.2016.06.2362.
Leng K, Li E, Eser R, et al. Molecular characterization of selectively vulnerable neurons in Alzheimer's disease. Nature Neuroscience 2021;24(2):276–287. DOI 10.1038/s41593-020-00764-7.
Fan Z, Wang L, Zhang H, Lv X, et al. Apathy as a risky neuropsychiatric syndrome of progression from normal aging to mild cognitive impairment and dementia: a systematic review and meta-analysis. Frontiers in Psychiatry 2021;12:792168. DOI 10.3389/fpsyt.2021.792168.
Fresnais D, Humble MB, Bejerot S, Meehan AD. Apathy as a predictor for conversion from mild cognitive impairment to dementia: a systematic review and meta-analysis of longitudinal studies. Journal of Geriatric Psychiatry and Neurology 2023;36(1):3–17. DOI 10.1177/08919887221093361.
Winer JR, Mander BA, Kumar S, et al. Sleep disturbance forecasts β-amyloid accumulation across subsequent years. Current Biology 2020;30(21):4291–4298.e3. DOI 10.1016/j.cub.2020.08.017.
Torack RM, Morris JC. The association of ventral tegmental area histopathology with adult dementia. Archives of Neurology 1988;45(5):497–501. DOI 10.1001/archneur.1988.00520290025008.
Nobili A, Latagliata EC, Viscomi MT, et al. Dopamine neuronal loss contributes to memory and reward dysfunction in a model of Alzheimer's disease. Nature Communications 2017;8:14727. DOI 10.1038/ncomms14727.
Takeuchi T, Duszkiewicz AJ, Sonneborn A, et al. Locus coeruleus and dopaminergic consolidation of everyday memory. Nature 2016;537(7620):357–362. PMID 27602521.
La Barbera L, Vedele F, Nobili A, et al. Nilotinib restores memory function by preventing dopaminergic neuron degeneration in a mouse model of Alzheimer's disease. Progress in Neurobiology 2021;202:102031. PMID 33684513.
Davies P, Katzman R, Terry RD. Reduced somatostatin-like immunoreactivity in cerebral cortex from cases of Alzheimer disease and Alzheimer senile dementia. Nature 1980;288(5788):279–280. DOI 10.1038/288279a0.
Miao C, Cao Q, Moser MB, Moser EI. Parvalbumin and somatostatin interneurons control different space-coding networks in the medial entorhinal cortex. Cell 2017;171(3):507–521.e17. DOI 10.1016/j.cell.2017.08.050.
Hardcastle K, Ganguli S, Giocomo LM. Environmental boundaries as an error correction mechanism for grid cells. Neuron 2015;86(3):827–839. PMID 25892299.
Leão RN, Mikulovic S, Leão KE, et al. OLM interneurons differentially modulate CA3 and entorhinal inputs to hippocampal CA1 neurons. Nature Neuroscience 2012;15(11):1524–1530. DOI 10.1038/nn.3235.
Martino-Saltzman D, Blasch BB, Morris RD, McNeal LW. Travel behavior of nursing home residents perceived as wanderers and nonwanderers. The Gerontologist 1991;31(5):666–672. PMID 1778493.
Rowe MA, Vandeveer SS, Greenblum CA, et al. Persons with dementia missing in the community: is it wandering or something unique? BMC Geriatrics 2011;11:28. DOI 10.1186/1471-2318-11-28.
The Fifty-Year Prescription
The temporal architecture says in what order the disease happens. It does not say why it begins. This paper is the most developed attempt here at the second question, and it is offered as a plausible account of priority rather than a settled one — a single mechanism followed without a break from a lifelong drift in the ecology of the gut, through the vagal relay and the locus coeruleus, to the withdrawal of a serotonergic tone that had been holding amyloid production down for decades before the first plaque, and on to the antidepressant trials that engage the right target half a century late.
It is a chain, and a chain of thirteen plausible links is not a plausible chain — so every claim is graded where it is used, and the paper closes with forty-one graded claims, eleven predictions, and the specific result that would break each link. The most upstream link, that gut ecology is a cause and not a companion, is the least established, and says so.
Read the paper →One lesion, a century apart
1907 — todayThe oldest idea in this corpus and one of its newest are the same idea. In 1907 Oskar Fischer read the plaque as a process — a lesion growing outward from a dying neuron. A century later two prize entrants re-derived his morphogenesis in molecular terms: amyloid accumulates inside the neuron, and the plaque is the gravestone of a cell killed from within.
Sources. Fischer's own papers and his eclipse: Fischer's Sidelining. What he actually staged — and one reading of him this corpus examines and declines: The Staged Deposit. The modern inside-out account: the inside-out plaque.
Fischer's Sidelining
In 1910 the disease Oskar Fischer had described across 275 brains was given another man's name. In 1941 the sanatorium he had run at Veleslavín since 1908 was signed over to a German intelligence agent whom Czechoslovakia had convicted and imprisoned, and whom the occupation set free. Fischer refused to sign. He was arrested for the refusal, held in the Small Fortress at Terezín — the Gestapo police prison, not the ghetto — and beaten to death there on 28 February 1942.
Rewritten from previously unpublished family papers: the notary contract he signed two days after German troops entered Prague, his own prison postcards giving the cell numbers, and a fellow prisoner's 1946 letter to his son. Earlier editions of this paper called the cause of death “illness, exhaustion” and the distinction bureaucratic; that correction is published in full. The paper also declines the obvious conclusion — the eponym did not lead to the Small Fortress, and the two erasures share a shape, not a cause.
Read Fischer's Sidelining →Every stage, in order
All documents →The full reading sequence — the molecular groundwork, the three clinical phases, then the spread, convergence, synthesis and frontier.
The same argument, told at length
Everything above is the evidence — papers, concepts, the apparatus for reading them. The seven monographs are the argument those papers add up to, written as continuous narrative and meant to be read in order.
Or find your own way in
A note on this edition
This wiki publishes the science: the papers, the concepts, the convergence nodes and the temporal stages. It withholds the working dossiers on individual researchers, which are graded internal notes — where the text names one, it appears as plain text rather than a link.