EEG control experiments

Posterior alpha, spectral background, individualized frequency bands, and healthy-state controls.

Executed results. The source spatial difference persists under individualized bands. The score is also sensitive to ordinary eye opening. Biological causation and AD-specific diagnosis remain unestablished.

Paired healthy-state tests

One observation per participant and condition. Differences are condition minus reference. Holm correction across8planned state endpoints; pointwise bootstrap confidence intervals.

Individualized-frequency sensitivity

Exploratory; highest fitted posterior peak4–14Hz,±2Hz band. Missing peaks remain missing. Separate six-test correction.

Source spectral and regional contrasts

AD versus controls. These p-values account for all81planned group tests; the complete table is included in results/spectral_group_tests.csv.

Spectral-model quality

Fit gate:R²≥0.8 and error≤0.15log10units. Eight-second failures are not used to infer background or peaks.

Interpretation and methods

# Executed EEG control results

The source-cohort posterior/frontal alpha difference survives an individualized frequency-band sensitivity, and the frozen19channel score responds to ordinary eye-condition changes in older healthy adults. These results narrow the measurement interpretation; they do not identify a molecular cause or establish Alzheimer-specific diagnosis.

## Scope

Source:88participants, with36AD/29controls for primary contrasts and23FTD for specificity contrasts; first8seconds and first/middle/last60seconds. Florida:80AD/12controls, eight seconds only. LEMON:74older candidates, 34included paired participants after availability/channel requirements. CAP:16healthy candidates, 11recordings with eligible signals/epochs; each paired comparison reports its own eligible n. Exclusion receipts, parser corrections and protocol amendments are included.

## Main results

1. Source posterior absolute alpha is lower in AD in all three60second periods after Holm correction across81planned group comparisons. Frontal alpha has no corrected group difference in these periods. A nonsignificant frontal result does not prove equal frontal power.
2. The posterior/frontal ratio remains lower in AD using an individualized band around each participant's highest fitted4–14Hz posterior peak: all three exploratory contrasts survive the separate six-test correction. Candidate peak frequency itself does not survive correction. Sub-8Hz peaks are not automatically identified as slowed alpha.
3. Background-adjusted posterior alpha is lower in AD after correction in the first and middle periods, but not the last; the background-adjusted spatial contrast survives correction only in the middle period. Therefore the effect is not demonstrated to be wholly independent of spectral background in every period.
4. The original8second spectra fail the fixed spectral-model quality gate in both cohorts. Their raw band-power tests remain available, but their spectral-decomposition parameters are not interpreted. Longer source spectra mostly pass; no Florida long recordings are invented.
5. In LEMON, eyes opening increases the frozen deviation score in30/34participants and lowers posterior/frontal alpha in29/34. Mean score changeEO−EC=0.322,95%CI[0.212,0.421], Holm p≈0.0008. The common19channel average-reference score also changes. This is evidence of state sensitivity, not a clinical false-positive rate or proof that eye opening explains the AD/control difference.
6. CAP stage2 increased theta/alpha in all11eligible participants (mean log-ratio change1.224;95%CI0.764–1.701;Holm p0.0040). The posterior/frontal alpha contrasts and stage1 theta/alpha contrast did not survive correction. These results are not a worst-case microsleep bound. CAP wake labels do not establish eye closure; states can come from different times of night and recording contexts differ from the clinical cohort. ECG regression/sham results are descriptive contamination sensitivities, not evidence for autonomic causation.
7. None of the source AD-versus-FTD factor contrasts survives the81-test correction. This does not prove equivalence, but AD specificity remains unestablished.

## CAP state results

cohort,condition,reference,factor,n,reference_mean,condition_mean,mean_paired_difference,ci_low,ci_high,p,n_increase,n_decrease,holm_state_tests
CAP,S1,W,log_posterior_frontal_alpha,10,0.2990082772448659,0.0965727418195456,-0.2024355354253202,-0.4433896933647338,0.0060932048934045,0.1491850814918508,3,7,0.1491850814918508
CAP,S1,W,log_theta_alpha,10,-0.6520069040358066,0.1014538779854774,0.7534607820212841,0.2537154988385737,1.279861539024859,0.0334966503349665,8,2,0.0923907609239076
CAP,S2,W,log_posterior_frontal_alpha,11,0.2949641368944536,-0.0271049006630412,-0.3220690375574949,-0.5596802236729259,-0.0838881721169772,0.0307969203079692,2,9,0.0923907609239076
CAP,S2,W,log_theta_alpha,11,-0.6408358101140369,0.5833270287310536,1.2241628388450905,0.7639664592941038,1.7005302902014758,0.000999900009999,11,0,0.003999600039996

## Supported interpretation

An AD-associated posterior alpha-power/topography difference is reproducible in the analyzed recordings and is robust to the tested individualized-band definition in the source cohort. The score also responds to normal eye-condition changes. The literature provides biological hypotheses, but this package does not measure synaptic loss, cholinergic integrity, amyloid, tau, or AD alpha reactivity.

The proposed follow-up is paired eyes-open/closed EEG in AD and comparison groups, ideally linked to MRI or other independent biomarkers. Schumacher et al.2020 is a concrete precedent with EEG and NBM-volume measurements; its data are available on reasonable request, not as a verified direct download here. https://doi.org/10.1186/s13195-020-00613-6

## Reproduction

See README.md for the offline command sequence. Exact analysis waveform extracts, full Florida archive, scripts, requirements, formulas, input hashes, person-level outputs, and PNG/SVG graphs are included. Full-length source/LEMON/CAP recordings are not duplicated in this control package. Statistical intervals are pointwise and these are retrospective analyses with documented amendments, not a preregistered causal experiment.

Limitations for the paper

# Paper-ready methodological wording

We evaluated the state sensitivity of the frozen EEG score using within-participant eyes-closed/eyes-open recordings from older healthy adults and a separate, reduced-montage sleep-stage comparison. Sleep epochs were selected from externally supplied annotations rather than the spectral outcomes under investigation. A metadata-driven amendment replaced an initial first-30-minute restriction with the first eligible scored epochs per state; optional annotation-column handling was also corrected before the final sleep analysis. CAP sleep-stage comparisons are interpreted as vigilance sensitivity analyses, because 30-second labels cannot exclude brief within-epoch state changes and sleep-study posture, instrumentation, and context differ from the seated clinical recordings. These comparisons neither certify wakefulness in the AD cohort nor provide a validated worst-case bound on microsleep contamination. The AD data contain only eyes-closed recordings; therefore, impaired alpha reactivity or an AD-by-eye-condition interaction cannot be estimated from these data.

To examine sensitivity to fixed frequency boundaries, an exploratory analysis used the highest fitted posterior spectral peak with a center between4 and14Hz and integrated regional power within±2Hz of that peak. This range and procedure were applied identically to AD and control participants. Spectra without an eligible peak were reported as missing; no peak was forced or imputed. Because a lower-frequency spectral peak cannot independently identify its neuronal generator, these peaks are described as candidate posterior rhythms rather than unequivocally slowed alpha. Group contrasts in individualized posterior/frontal power persisted across all three sampled source-recording periods, whereas differences in candidate peak frequency did not survive correction. These results support robustness to this particular frequency-band choice, without establishing a unique biological mechanism.

The first eligible epochs of a state can be separated in time and do not necessarily form an event-aligned wake-to-sleep transition. CAP wake labels also do not independently establish eye closure. Therefore, these comparisons are within-person state contrasts, not perfectly matched eyes-closed transition experiments. Differences in time of night and unrecorded condition details remain possible contributors.

Literature context

# Interpretation of the literature supplied during this run

The literature supports posterior alpha abnormalities as an established research area. The contribution of this package is a reproducible audit of a particular spatial score, its spectral contributors and its state sensitivity; it does not claim the first discovery of lower posterior alpha in AD.

## Directly checked primary mechanistic study

Schumacher et al. (2020), EEG alpha reactivity and cholinergic system integrity in Lewy body dementia and Alzheimer's disease, https://doi.org/10.1186/s13195-020-00613-6 . The study measured eyes-open/closed EEG in21AD,41LBD and40controls, plus structural MRI in a subset. Reactivity was reduced in dementia groups and related to NBM volume across groups; the within-group association survived correction specifically in PDD, not AD. This supports a cholinergic hypothesis but does not establish it for our AD participants. Its data-availability statement offers data from the corresponding author upon reasonable request; no direct downloadable EEG/MRI dataset was provided there. No request was sent.

## Supplied review

Baek (2026), https://www.frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2026.1665301/full , is a mini-review, not a new experiment. It distinguishes resting alpha power from EC-to-EO reactivity and describes the marker as promising, with additional harmonized multimodal/longitudinal validation needed. The original primary study above was checked for the mechanistic claim.

## What our data can and cannot establish

Our AD inputs contain eyes-closed recordings. Healthy LEMON EC/EO data establish score sensitivity to condition, but cannot establish reduced AD reactivity or estimate an AD-by-eye-condition interaction. There is no paired MRI/PET, synaptic marker, cholinergic measure or drug intervention for our AD participants in this package. Structural and molecular mechanisms remain literature-supported hypotheses, not measured causes in this experiment.

A normal EC-to-EO change affecting the score does not imply that eye opening caused the AD/control difference: both source groups were recorded eyes closed. It demonstrates the need for state-standardized interpretation and does not erase the observed disease-group association.

The supplied Scientific Reports2016 paper (https://pubmed.ncbi.nlm.nih.gov/27546195/ ; doi10.1038/srep31859) concerns memory-task EEG and medial temporal atrophy predicting conversion from amnestic MCI. It should not be cited as a direct measurement of synaptic loss causing our resting scalp-power difference. The supplied INSIGHT-preAD2020 study (https://pubmed.ncbi.nlm.nih.gov/32111391/) is a relevant example of linking resting alpha to amyloid-PET-defined status and education; its design is different from our clinically labeled AD/control data.