# Lane 4 — Adverse Events & the Safety-Duty Design

**Waypoint research phase · 2026-07-16 · status: complete**
Feeds: spec §3 (confusion pairs: dark night vs depression; devotional intensity vs destabilization), §6 (phase markers), **§8 (teacher-flag design — the deliverable, at the end of this file)**.
Verification note: every citation below was checked against a live primary source (publisher page, PubMed, or the journal PDF) during this pass unless marked otherwise. Nothing is cited from memory alone.

---

## 1. The Varieties of Contemplative Experience (VCE) study

**Lindahl JR, Fisher NE, Cooper DJ, Rosen RK, Britton WB (2017). The varieties of contemplative experience: A mixed-methods study of meditation-related challenges in Western Buddhists. *PLoS ONE* 12(5): e0176239. doi:10.1371/journal.pone.0176239.** [verified — publisher page]

The taxonomy Waypoint's difficulty-marker family should be built on. 100+ interviews with Western Buddhist practitioners and teachers (Theravāda, Zen, Tibetan lineages), ~10 years, 3,000+ pages of transcript.

- **59 categories of meditation-related experience across 7 domains:** cognitive, perceptual, affective, somatic, conative, sense of self, social. Practitioners averaged 19.6 categories each; ~¾ reported experiences across 6+ domains. Difficulty is *multi-domain by default* — a marker bank that only listens on the affective channel will miss most of the signature.
- **26 influencing factors in 4 domains:** practitioner-level (trauma/psychiatric history, intentions, worldview), practice-level (amount/intensity, type, *response to* the experience, stage of practice), relationships (teacher, community, support), health behaviors (sleep, diet, exercise, substances). These are Waypoint's natural *flag-context features* — the same phenomenology means different things at different practice intensities and support levels.
- **Severity in this sample:** median challenge duration 1–3 years (days to 10+ years); 73% reported moderate-to-severe impairment in at least one domain; 17% reported suicidality; 17% required hospitalization; 88% said challenges spilled beyond formal practice into daily life.
- **Critical sampling caveat:** VCE deliberately criterion-sampled for challenging experiences (help-seeking-adjacent). These are *severity conditional on being a difficult case*, **not prevalence**. Use VCE for the taxonomy and severity gradients; use §3 below for base rates.
- **The appraisal finding (load-bearing for Waypoint):** identical phenomenology was appraised as progress by some and pathology by others; "interpretations of and responses to the experiences differed considerably," and Buddhist textual traditions themselves classify the same experience as marker-of-progress or obstacle depending on framework. Consequence: **the flag must be built on phenomenology + duration + impairment, never on the practitioner's own valence label or on our interpretive gloss.**

Related (cited in corpus digest 1, not re-verified this pass — flag as VERIFY-ON-USE): Lindahl & Britton (2019), "'I Have This Feeling of Not Really Being Here': Buddhist Meditation and Changes in Sense of Self," *Journal of Consciousness Studies* 26(7–8) — ~55% of reported self-boundary changes were distressing, ~45% came with functional impairment. Directly relevant to scoring awareness-pillar advances: **self-boundary change markers are simultaneously stage evidence and safety evidence.**

## 2. Britton lab: defining, measuring, and the clinical presentation

**Britton WB, Lindahl JR, Cooper DJ, Canby NK, Palitsky R (2021). Defining and Measuring Meditation-Related Adverse Effects in Mindfulness-Based Programs. *Clinical Psychological Science* 9(6): 1185–1204. doi:10.1177/2167702621996340.** [verified — publisher + PMC8845498]

44-item Meditation Experiences Interview (MedEx-I), independent assessor, n=96 across three MBCT variants:

- 83% reported ≥1 meditation-related side effect; **58% negative valence; 37% negative impact on functioning; 6–14% "lasting bad effects"** (definition-dependent: >1 day / >1 week / >1 month) — a rate comparable to psychotherapy generally (3–14%).
- **Lasting bad effects were associated with signs of dysregulated arousal: hyperarousal and dissociation.** This is the single most flag-relevant empirical regularity in the literature: the *persistence risk signature* is not sadness, it is arousal dysregulation.
- Methodological lesson Waypoint inherits: valence, causal attribution, duration, and functional impact are **separate dimensions queried separately**. A marker event schema that collapses them ("negative experience: yes/no") reproduces the measurement failure this paper was written to fix. Waypoint difficulty markers should carry `{phenomenology_category, valence, duration, impairment, practice_link}` as distinct fields.

**Cheetah House clinical taxonomy** (cheetahhouse.org/symptoms; founded by Britton; the VCE 59×7 taxonomy operationalized for meditators-in-distress) [verified — site + Tricycle interview]: the presenting-problem clusters, in order of clinical frequency, are (a) **hyperarousal cluster** — anxiety/panic, emotional flooding ("someone turned up the volume knob"), insomnia, involuntary movements; (b) less commonly, **dissociative/perceptual cluster** — DPDR-type detachment, perceptual distortions. People typically present when they *can no longer meditate or function*, not at first onset — meaning passive conversational evidence will usually show the pattern **before** a practitioner would self-identify as in difficulty. That is the whole case for the consolidation-pass flag.

## 3. Base rates (for threshold calibration and expected flag volume)

**Farias M, Maraldi E, Wallenkampf KC, Lucchetti G (2020). Adverse events in meditation practices and meditation-based therapies: a systematic review. *Acta Psychiatrica Scandinavica* 142(5): 374–393. doi:10.1111/acps.13225.** [verified — PubMed 32820538]
83 studies, 6,703 participants. Pooled AE prevalence **8.3%** (95% CI 0.05–0.12), splitting sharply by method: **3.7% in experimental studies vs 33.2% in observational studies** (ask-and-you-shall-find; passive trial monitoring undercounts). Most common AEs: anxiety (33%), depression (27%), cognitive anomalies (25%); gastrointestinal problems and suicidal behaviors ~11% each. **AEs occur in people with no prior mental-health history** — a clean psychiatric history must not lower the flag threshold.

**Goldberg SB, Lam SU, Britton WB, Davidson RJ (2022). Prevalence of meditation-related adverse effects in a population-based sample in the United States. *Psychotherapy Research* 32(3): 291–305. doi:10.1080/10503307.2021.1933646.** [verified — PubMed]
n=434 meditators, population-based (i.e., app-user-like, modest experience): **50%** endorsed ≥1 specific adverse-effect type; **10.4%** had effects lasting ≥1 month; **10.6%** some functional impairment; **1.2%** impairment lasting ≥1 month. Most common: **anxiety, traumatic re-experiencing, emotional sensitivity.**

**Calibration take (marked speculation, but anchored):** in an app population resembling Goldberg's sample, expect on the order of **~10% of active practitioners per assessment window to genuinely satisfy a "sustained difficulty" criterion**, and ~1% to be in lasting-impairment territory. A Tier-A flag rate persistently ≫15% or ≪2% of the active cohort is evidence the thresholds are miscalibrated, not that the cohort changed. Farias's experimental-vs-observational 9× gap also warns that *how* evidence is elicited moves measured rates enormously — probe items about difficulty will surface far more than passive listening; the aggregator must not treat probe-elicited difficulty reports as equivalent evidence to spontaneous ones (different likelihoods).

## 4. Dark night vs clinical depression — the differential, per whom

### 4a. The empirical anchor: teachers' and practitioners' actual criteria

**Lindahl JR, Cooper DJ, Fisher NE, Kirmayer LJ, Britton WB (2020). Progress or Pathology? Differential Diagnosis and Intervention Criteria for Meditation-Related Challenges: Perspectives From Buddhist Meditation Teachers and Practitioners. *Frontiers in Psychology* 11:1905. doi:10.3389/fpsyg.2020.01905.** [verified — publisher]

Eleven criteria teachers actually use: phenomenological match to tradition maps (ñāṇas, *makyō*, *nyams*); **distress level; controllability (can the practitioner modulate it by adjusting practice); duration; functional impairment; preserved critical attitude (reality-testing)**; cultural compatibility; health history; onset context (retreat vs daily life); longitudinal impact (growth vs deterioration); and the teacher's own skills/resources.

- **Most decision-reliable across communities:** suicidality and disruptive behavior warranted intervention *regardless of duration or framework*; loss of critical attitude, uncontrollability, sustained impairment were the consistent intervention triggers.
- **Least reliable / explicitly contested:** psychiatric history as an explanation (circular: it delegitimizes any experience a person with a history reports), and retrospective benefit ("it was purification") — useless at decision time.
- **The key reframe (also Cheetah House's differential-diagnosis page):** replace the binary "spiritual vs pathological?" with **"what kind of support does this person need?"** An experience can be valid path territory *and* need intervention. This is exactly the framing Waypoint's flag should encode — the flag never adjudicates ontology, it triages support need.

### 4b. What actually discriminates (both lenses, synthesized)

Contemplative-side sources (Mahasi *Progress of Insight* dukkha-ñāṇa map; Ingram *MCTB*; Shinzen Young's dark-night-vs-dukkha-ñāṇa distinction — dukkha ñāṇas cycle within days, a "dark night" proper is a months-long dissociative episode) and clinical-side sources (VCE/Britton corpus; Cheetah House diffdiag) converge on a usable feature set. Per corpus digest 4, the Mahasi 16-ñāṇa map (dukkha ñāṇas = stages 5–10) is already in the internal canon's staging vocabulary, so these features double as phase markers:

| Feature | Dark-night / dukkha-ñāṇa phase leans | Clinical-depression leans |
|---|---|---|
| Temporal coupling | Onset tracks practice intensification, retreat, or a named insight event; often follows a high-clarity period (post-Arising-and-Passing in Mahasi terms) | No systematic coupling to practice; may predate practice |
| Dynamics | **Cycles/stages** — moves through recognizable textures (dissolution→fear→misery→disgust→desire-for-deliverance) within days-to-weeks | **Static or slowly worsening**; same flat quality week after week |
| Perception | Vividness often preserved or heightened; anicca hyper-salience ("everything is dissolving/vibrating") | Blunting, dulling, anhedonic graying |
| Content | Existential/insight-flavored: meaninglessness *of constructs*, disenchantment, fear without object | Guilt, worthlessness, self-blame, hopelessness about *oneself* |
| Response to modulation | Improves when practice is reduced/grounded (backing off works) | Indifferent to practice changes |
| Motivation toward practice | Often aversion to practice *with* continued pull toward the path | Global anhedonia including the path |
| Sleep/appetite | Variable, often hyperarousal-flavored | Classic neurovegetative pattern |

Cautions that bind: (i) even experienced practitioners report being unable to tell the two apart from the inside (Dharma Overground practitioner accounts) — so **Waypoint must never assert the differential either**; it records *features* with both hypotheses and flags for a human. (ii) The two co-occur: a dukkha-ñāṇa phase can precipitate or coexist with a depressive episode. (iii) F-SNR corpus (digest 6) grounds the phase model: U-shaped destabilize-then-restabilize curves are *expected*, so "dark-night week ≠ regression" is scientifically defensible, but only *phase-typical, time-limited, non-impairing* dips get the benign reading.

## 5. Trauma-sensitive delivery constraints (Treleaven)

**Treleaven DA (2018). *Trauma-Sensitive Mindfulness: Practices for Safe and Transformative Healing*. W.W. Norton. (Foreword: W. Britton.)** [verified — publisher/retail + author site]

Five principles: **(1) stay within the window of tolerance** (the foundational construct — an arousal zone within which practice is safe; outside it, practice consolidates dysregulation), (2) shift attention to support stability, (3) work with dissociation, (4) practice in relationship, (5) understand social context. Plus 36 concrete practice modifications.

Waypoint/Wisdom translation — these become **flag-conditional delivery constraints**, not just teacher advice:
- While a difficulty flag is active: Wisdom biases toward **stabilizing moves** — external/grounding anchors (sound, feet, open eyes), shorter sessions, choice and opt-out language ("if it feels okay…"), titrated exposure; and **away from** deconstructive/intensive practices (2.4-style deconstruction, long silent sits, breath-retention work). This aligns with the existing rule that Stage 2.4 + dying contemplations need extra flagging.
- **Probes are suppressed while a flag is active** (spec §5 already mandates "never during flagged-vulnerable moments" — Treleaven is the citation that makes this defensible rather than merely cautious).
- Goldberg 2022's top-3 AEs (anxiety, traumatic re-experiencing, emotional sensitivity) are precisely the trauma-surface Treleaven addresses; **traumatic re-experiencing markers should route to the trauma-flavored flag template, not the dark-night one.**

## 6. Duty-of-care & escalation design for a non-clinical wellness product

### 6a. The regulatory boundary (what the flag may claim)

FDA, *General Wellness: Policy for Low Risk Devices* (guidance; **updated January 6, 2026** together with the CDS software guidance) [verified — fda.gov guidance PDF + legal analyses (Faegre Drinker, Hooper Lundy, Sidley, Jan–Nov 2025/2026)]:
- **Intended use is the trigger**, and it is read from labeling, marketing, UI copy, and functionality — not just formal claims. A mindfulness product stays general-wellness so long as it claims to maintain/encourage general health (stress management, wellbeing) and does **not** claim to diagnose, treat, cure, mitigate, or prevent a disease/condition.
- Consequence for flag copy: **the moment flag text says "depression," "PTSD," "psychosis," "dissociative disorder" as an assertion about the practitioner, we are generating diagnostic output.** Internal-only status helps but is not a shield if it drives product behavior marketed as therapeutic. Flag text therefore names *observed patterns in the practitioner's own words* + support-need level, never nosology.
- FDA's 2025–26 attention to generative-AI mental-health tools (Digital Health Advisory Committee, Nov 2025) signals the wellness/device line is being actively policed for exactly this product category.

### 6b. Human-in-the-loop norms

- **APA app-evaluation framework (Torous et al.)** — safety and privacy are evaluated *first*, before efficacy/engagement (Torous et al. 2018, *Psychiatric Services* 169(5); consensus standards: Torous et al. 2019, *World Psychiatry* 18(1):97–98) [verified — publisher pages]. A wellness product is increasingly measured against these criteria regardless of regulatory classification.
- Emerging crisis-support assessment instrument: **MHACSAF** (Mental Health App Crisis Support Assessment Framework), *Frontiers in Digital Health*, 2026 [verified — journal page; new, treat as directional]. Converging norm: tiered design — automated detection → human review → de-escalation or escalation; visible crisis-resource access (988 in the US) as a floor.
- Design consensus from the digital-mental-health literature: **automated systems detect and route; humans decide and act.** Waypoint's architecture already conforms: consolidation pass raises the flag → teacher (human) judges → outreach under LIFE Team identity (never Wisdom). Acute risk remains the constitutional classifier's authority (GREEN→RED severity ladder), which owns real-time crisis resources.

### 6c. The threshold-design lesson: triage present states, never predict events

**Belsher BE et al. (2019). Prediction Models for Suicide Attempts and Deaths: A Systematic Review and Simulation. *JAMA Psychiatry* 76(6): 642–651. doi:10.1001/jamapsychiatry.2019.0174.** [verified — PubMed 30865249]
17 cohort studies, 64 models, >14M participants: global classification accuracy good (≥0.80) yet **positive predictive value for the rare event ≤0.01 in most models** — "accuracy of predicting a future event is near 0." At low base rates, even excellent classifiers flag almost entirely false positives when framed as *event prediction*.

The transferable lesson for Waypoint: **do not build the teacher flag as a risk predictor** ("this practitioner will deteriorate"). Build it as a **nowcast of an observed, already-present support-need state** ("this practitioner has reported X-cluster phenomenology for Y weeks with Z impairment") — a triage of the present, where the "event" (current difficulty) has base rates of ~10%, not 0.01%, and where every flag is verifiable by a human reading the cited evidence. This single design choice is what makes the flag statistically defensible.

---

## 7. DELIVERABLE — Teacher-flag design for spec §8

### 7.0 Position in the safety stack

Two loops, cleanly separated: the **constitutional classifier** is the fast loop (session-time, acute risk, owns crisis response — unchanged, remains the safety authority); the **Waypoint teacher flag** is the slow loop (consolidation-time, sustained-difficulty triage). The flag never handles acute risk; acute markers found during extraction are routed to the classifier's severity machinery immediately, and *also* recorded as flag evidence.

### 7.1 Marker patterns that warrant a flag

Difficulty markers carry the Britton-2021 event schema: `{phenomenology_category (VCE 59×7), valence, duration, impairment, practice_link, elicitation (spontaneous|probe)}`.

**Tier A — standard flag, raised by monthly consolidation (persistent-difficulty patterns):**
1. **Dysregulated-arousal cluster** (the empirical persistence-risk signature, Britton 2021): ≥2 of {anxiety/panic, emotional flooding, sleep disruption, involuntary movements/energy phenomena} recurring across sessions.
2. **Dissociative cluster**: DPDR-type reports ("not really here," unreality, watching-from-outside) with negative valence or impairment. (Positive-valence self-boundary change without distress/impairment = stage evidence, not a flag — Lindahl & Britton 2019.)
3. **Trauma re-experiencing** during or triggered by practice (Goldberg 2022 top-3) — routes to the trauma-flavored template.
4. **Sustained dark-night phase**: dukkha-ñāṇa-typical phenomenology persisting ≥3 weeks OR any duration with functional-impairment reports OR features leaning depression-typical per the §4b table (static quality, self-directed worthlessness, global anhedonia).
5. **Unbinding-adjacent territory**: posterior mass ≥0.3 on bands 51+ at consolidation, or a corroborated `small_death` milestone — flags on *stage*, not distress (highest-support-need territory per the canon).
6. **Post-intensification destabilization**: difficulty-cluster onset within ~2 weeks after a retreat, practice-load jump, or gate progression (VCE practice-level factors).

**Tier B — expedited flag (teacher notified without waiting for monthly consolidation; the Lindahl-2020 consensus intervention triggers):**
1. **Loss of critical attitude**: reality-testing slippage sustained across exchanges (fixed non-consensual convictions acted on — distinct from reporting unusual *experiences*, which is expected path material).
2. **Uncontrollability**: practitioner reports being unable to stop, modulate, or exit the experience, including outside formal practice.
3. **Escalating functional impairment**: can't work / relationships breaking / stopped eating or sleeping properly, attributed to practice-related states.
4. **Suicidality or self-harm content**: classifier authority fires as usual (acute path); the flag additionally records it as Tier B so the teacher sees the longitudinal context. Per Lindahl 2020: concerning *regardless of duration or interpretive framework* — no phase annotation ever downgrades it.

### 7.2 Threshold logic

- **Conjunctive core:** Tier A requires (phenomenology cluster) AND (persistence: evidence in ≥2 separate sessions ≥1 week apart, or 1 session + a corroborating telemetry break — practice-pattern collapse, session-abandonment spike, time-of-day disruption) AND (≥1 of: negative valence, impairment report, uncontrollability report). Single-session single-marker events never flag; they accumulate.
- **Tier B is disjunctive** — any one trigger suffices; sensitivity dominates.
- **Phase-aware, not phase-excused:** a dukkha-ñāṇa phase annotation contextualizes the flag text ("phase-typical pattern") but only *suppresses* a Tier A flag when the pattern is phase-typical AND time-limited (<3 weeks) AND impairment-free. Phase never suppresses Tier B.
- **Independent of stage certainty:** flags fire even when the band estimate abstains (`data_sufficiency: insufficient`). Safety evidence does not wait for stage evidence.
- **Elicitation-weighted:** probe-elicited difficulty reports carry lower likelihood weight than spontaneous ones (Farias 3.7% vs 33.2% elicitation gap).
- **Hysteresis:** a raised flag clears only after two consecutive clean consolidations (no qualifying markers), preventing flap; clears immediately on teacher disposition "resolved."
- **History-neutral:** psychiatric history neither raises nor lowers thresholds (Lindahl 2020 circularity; Farias: AEs occur without history). It may appear in flag *context* only if the practitioner volunteered it.
- **Rate monitor:** expected Tier-A rate ~5–15% of active practitioners per window (Goldberg-anchored; SPECULATION to be calibrated at M4). Sustained rates outside ~2–15% trigger threshold review, not silent acceptance.

### 7.3 False-positive vs false-negative stance

**Asymmetric toward sensitivity, made cheap by design.** A false positive costs one teacher a few minutes reading a cited dossier snippet; a false negative is a practitioner destabilizing unwatched for a month (Cheetah House: people present only when they can no longer function). So: recall-first at Tier definitions, precision recovered by (a) the conjunctive persistence requirement, (b) every flag carrying verifiable quoted evidence, (c) the teacher-disposition feedback loop (each flag dispositioned useful / not-useful / resolved; dispositions refit marker likelihoods at M5–M6). The Belsher constraint bounds the whole design: the flag asserts **present observed state, never future risk** — no risk scores, no "probability of deterioration," ever. That framing keeps the effective base rate high (~10%, not 10⁻³) and the PPV defensible.

### 7.4 Flag text — what it must and must not say

Template (teacher dashboard only; never practitioner-visible):

> **Support check-in suggested — [pattern name, plain language]**
> [Practitioner] · flag tier A/B · raised [date] by monthly consolidation
> **Observed:** [marker pattern] across [N] sessions over [window], e.g. "[direct quote]" ([session ref]), "[quote]" ([ref]). Practice telemetry: [corroborating change, or "no change"].
> **Context:** estimated territory [band, CI]; phase annotation [phase, confidence]; recent [retreat/gate/load change if any].
> **Suggested action:** teacher check-in within [7 days / 48h for Tier B], via LIFE Team identity.
> *This is a supportive-triage signal generated from conversation and practice patterns. It is not a diagnosis, not a clinical assessment, and not a prediction. The practitioner has not seen this flag.*

**Must contain:** direct quotes with references (verifiability); duration and cross-session counts; phase + stage context; tier; a concrete suggested action with a timeframe; the not-a-diagnosis disclaimer; LIFE-Team-identity routing.
**Must never contain:** diagnostic labels asserted of the person ("is depressed," "PTSD," "psychotic"); treatment or medication suggestions; numeric risk of future harm; causal claims ("meditation caused…" / "this practice harmed…"); ontological adjudication ("this is (not) a genuine dark night"); anything practitioner-facing; comparative rankings against other practitioners.
Depression-overlap wording, when the §4b table leans clinical: *"pattern includes features that in other contexts overlap with low-mood presentations; teacher judgment on whether to suggest the practitioner consider outside support"* — the teacher, a human, makes any referral suggestion.

### 7.5 Wisdom behavior while a flag is active (Treleaven layer)

Stabilize-first delivery constraints switch on automatically: grounding/external-anchor practice bias; shorter sessions; opt-out language; **no deconstructive or intensive practices** (2.4 deconstruction, dying contemplations, breath retention); **all probes suppressed**; register shifts to warm-and-concrete. These constraints are advisory adaptation (spec §8: no hard gating pre-validation) except probe suppression, which is hard (already spec law, §5).

### 7.6 Open items for M2 (Surya calibration)

1. Surya to review the §4b differential table against the canon's own dark-night material (canon places hysteria-adjacent deconstruction at 58–59; the dukkha-ñāṇa map anchors much earlier territory — the same *words* at 25 vs 55 need different flag priors, which is exactly the stage-aware-classifier promise).
2. Set the Tier-A persistence window (3 weeks is a defensible default from Shinzen's days-vs-months distinction; canon may prefer another).
3. Confirm the Unbinding-adjacent posterior threshold (0.3 is a placeholder).
4. Decide whether Tier B expedited flags need an out-of-band notification channel (email/Slack to teacher) vs next-dashboard-visit.

---

## Source list (all verified live this pass unless noted)

- Lindahl, Fisher, Cooper, Rosen & Britton (2017), *PLoS ONE* 12(5):e0176239, doi:10.1371/journal.pone.0176239
- Britton, Lindahl, Cooper, Canby & Palitsky (2021), *Clinical Psychological Science* 9(6):1185–1204, doi:10.1177/2167702621996340
- Goldberg, Lam, Britton & Davidson (2022), *Psychotherapy Research* 32(3):291–305, doi:10.1080/10503307.2021.1933646
- Farias, Maraldi, Wallenkampf & Lucchetti (2020), *Acta Psychiatrica Scandinavica* 142(5):374–393, doi:10.1111/acps.13225
- Lindahl, Cooper, Fisher, Kirmayer & Britton (2020), *Frontiers in Psychology* 11:1905, doi:10.3389/fpsyg.2020.01905
- Baer, Crane, Miller & Kuyken (2019), *Clinical Psychology Review* 71:101–114, doi:10.1016/j.cpr.2019.01.001; follow-up: Baer et al. (2021), *Mindfulness* 12(3):763–774
- Treleaven (2018), *Trauma-Sensitive Mindfulness*, W.W. Norton
- Belsher et al. (2019), *JAMA Psychiatry* 76(6):642–651, doi:10.1001/jamapsychiatry.2019.0174
- Torous et al. (2018), *Psychiatric Services* 169(5), doi:10.1176/appi.ps.201700423; Torous et al. (2019), *World Psychiatry* 18(1):97–98, doi:10.1002/wps.20592
- FDA, *General Wellness: Policy for Low Risk Devices* (updated guidance, Jan 6 2026), fda.gov/media/90652/download
- MHACSAF: *Frontiers in Digital Health* (2026), doi:10.3389/fdgth.2026.1814547 [new instrument; directional]
- Cheetah House: cheetahhouse.org/symptoms, /diffdiag, /vce [org resource, not peer-reviewed]
- Lindahl & Britton (2019), *Journal of Consciousness Studies* 26(7–8) [cited via corpus digest 1 — VERIFY-ON-USE]
- Mahasi Sayadaw, *The Progress of Insight*; Ingram, *MCTB*; Shinzen Young dark-night commentary [traditional/practitioner sources; used for the differential table's contemplative lens, not as clinical evidence]