Phase Measurability

An epistemic guardrail for the corpus. The three-phase architecture is stated as a sequence of events in time. This note records, phase by phase, whether an instrument exists that could detect the claimed event in a living person — because a claim no instrument can currently reach has not survived scrutiny; it has escaped it.

The finding is uncomfortable and should be stated plainly: two of the three phases are, at present, unfalsifiable in living humans. The one phase that is well instrumented is the one the external literature attacks.

Phase I is below the instruments

Four independent barriers, and they compound:

Resolution. The locus coeruleus is 2–2.5 mm across (see LC Morphology What is Established). ADNI's harmonised tau-PET effective resolution was 8 mm FWHM, reduced to 6 mm in 2023 (Landau et al. 2025; Jagust et al. 2024); best-case intrinsic scanner resolution across the cohort scanners is 3.4–8.1 mm FWHM (Carbonell et al. 2025). The structure sits well below one resolution element in its limiting dimension.

The tracer binds the pigment. Worse than a resolution problem. Marquié et al. (2015) demonstrated flortaucipir "off-target binding to neuromelanin- and melanin-containing cells." The locus coeruleus is defined by neuromelanin. The field states the consequence outright — Koops et al. (2025): "The field still lacks spatially specific measures to investigate in-vivo tau accumulation in the LC, as most tau-PET ligands bind to neuromelanin in the LC and nearby choroid plexus, confounding any information on tau."

Zero published studies quantify tau-PET signal within an LC region of interest.

Structural MRI cannot see it either — and this is a contrast failure, not a resolution failure. Aganj et al. (2024), working at 0.7 mm isotropic MPRAGE — better than clinical standard — report "the lack of LC contrast on T₁W images" and fall back on purely geometric landmarks. Increasing T1 resolution will not help.

The dedicated method is fragile. Neuromelanin-sensitive MRI exists and is a mature subfield, but its histological validation is a single n = 7 post-mortem study establishing spatial correspondence without quantitative correlation (Keren et al. 2015); the contrast mechanism remains disputed (Priovoulos et al. 2018 argue magnetisation transfer rather than neuromelanin); test–retest ICC for the right locus coeruleus is 0.36 (Tona et al. 2017); and a 7T post-mortem validation reports MRI-to-histology overlap of just 2.5–23% (LaMore et al. 2025).

What this does and does not mean

It does not weaken the Phase I claim. Braak's post-mortem evidence is untouched by any of it: of 42 brains aged 4–29, 38 showed pretangle tau, and in 19 of 22 the material was confined to the coeruleus/subcoeruleus complex (Braak & Del Tredici 2011); in 2,332 unselected brains, 58 carried subcortical tau predominantly in the coeruleus with no cortical tau at all (Braak et al. 2011).

It means the absence of the locus coeruleus from data-driven staging models is not evidence against Phase I — no such model has ever included a brainstem region, so the silence is structural. And symmetrically, it means the corpus cannot cite in vivo imaging as confirmation.

The in vivo timing evidence is thinner than the corpus has implied

Exactly one study carries longitudinal LC imaging and longitudinal tau-PET. Decade-scale lead times are modelled from cross-sectional age curves.

Phase I should be stated as autopsy-anchored, with in vivo imaging cited as consistent and suggestive, thinly longitudinal.

Phase III cannot be measured in a living person at all

Larsen et al. (2023), Box 3: "there are no techniques available to investigate PNNs in the living human brain." In vivo perineuronal-net imaging is confined to animal models using genetically encoded labels and two-photon microscopy.

Phase III is doubly weakened: the corpus's own blind audit already demoted the terminal-cell identity (somatostatin interneuron involvement precedes parvalbumin), and there is no human in vivo instrument for the net. It should be stated as a post-mortem-and-model-derived hypothesis with a named measurement gap.

The structural consequence

The architecture is arranged so that its two most original claims sit where evidence cannot currently reach, and its one well-instrumented claim is the one under attack. That arrangement is not evidence of error — but it is the shape a framework would take if it had been built to be unfalsifiable, and the corpus has already established that the phase count was fixed for pharmacological rather than biological reasons.

The programme cannot claim the phases' resistance to refutation as a virtue when two of them are, at present, beyond refutation by construction. Naming its own unmeasurable regions is more credible than letting a reviewer find them.

What would change this

An LC-inclusive staging model is the tractable one, and it is unclaimed: the technique exists, the modality exists, and the field's own leaders have combined them — in Parkinson's disease (Zhou et al. 2023; Tan et al. 2023) and in TDP-43 proteinopathies (Young et al. 2023), never in Alzheimer's. A SuStaIn or event-based model over neuromelanin-MRI LC integrity plus cortical tau-PET regions would be the first analysis capable of placing Phase I in a data-driven sequence rather than asserting it.

For Phase III, no near-term instrument is visible. A human PNN ligand would be the unlock.

Primary sources

• Braak H, Del Tredici K (2011). The pathological process underlying Alzheimer's disease in individuals under thirty. Acta Neuropathol 121(2):171–181. PMID 21170538
• Braak H, Thal DR, Ghebremedhin E, Del Tredici K (2011). Stages of the pathologic process in Alzheimer disease: age categories 1–100 years. J Neuropathol Exp Neurol 70(11):960–969. PMID 22002422
• Marquié M, Normandin MD, Vanderburg CR, et al. (2015). Validating novel tau PET tracer [F-18]-AV-1451 on postmortem brain tissue. Ann Neurol 78(5):787–800. PMID 26344059
• Koops EA, Dutta J, Hanseeuw BJ, et al. (2025). Elevated locus coeruleus metabolism provides resilience against cognitive decline. Alzheimers Dement 21:e14385. PMID 39588792
• Aganj I, Mora J, Fischl B, Augustinack JC (2024). Automatic geometry-based estimation of the locus coeruleus region on T1-weighted MRI. Front Neurosci 18:1375530. PMID 38774790
• Keren NI, Taheri S, Vazey EM, et al. (2015). Histologic validation of locus coeruleus MRI contrast in post-mortem tissue. NeuroImage 113:235–245. PMID 25791783
• Priovoulos N, Jacobs HIL, Ivanov D, et al. (2018). High-resolution in vivo imaging of human locus coeruleus by magnetization transfer MRI at 3T and 7T. NeuroImage 168:427–436. PMID 28743460
• Tona KD, Keuken MC, de Rover M, et al. (2017). In vivo visualization of the locus coeruleus in humans: quantifying the test–retest reliability. Brain Struct Funct 222:4203–4217. PMID 28647901
• Bueichekú E, Diez I, Kim C-M, et al. (2024). Spatiotemporal patterns of locus coeruleus integrity predict cortical tau and cognition. Nat Aging 4(5):625–637. PMID 38664576
• Betts MJ, Cardenas-Blanco A, Kanowski M, et al. (2019). Locus coeruleus MRI contrast is reduced in Alzheimer's disease dementia. Alzheimers Dement (Amst) 11:281–285. PMID 30976648
• Larsen B, Sydnor VJ, Keller AS, et al. (2023). A critical period plasticity framework for the sensorimotor–association axis. Trends Neurosci 46(10):847–862. PMID 37643932
• Young AL, Vogel JW, Robinson JL, et al. (2023). Data-driven neuropathological staging and subtyping of TDP-43 proteinopathies. Brain 146(7):2975–2988. PMID 37150879
• Zhou C, Wang L, Cheng W, et al. (2023). Two distinct trajectories of clinical and neurodegeneration events in Parkinson's disease. NPJ Parkinsons Dis 9(1):111. PMID 37443179

Converges on

Locus coeruleus as ground zero · LC Morphology What is Established · Perineuronal Nets · Staged anatomical progression · Tau Propagation