*(Historical record, filed 14 July 2026, before the v2.0 rename: "resolution" herein is the framework term since renamed "cohering", and claim labels are as of v1.6. Preserved unedited.)*

# Red-Team: Study B at the Item Level
**Constructive attack on the residue instrument · v0.1 · 14 July 2026**
**Referent: `The_Container_Problem_v0.1.md` (working note v0.1, published in this corpus)**
**Stance: hostile experimental-philosophy reviewer, then instrument designer. Every item drafted here was drafted to be killed.**

---

## 0. Scope and rules of engagement

Study B claims kill authority over the existence question: does a "non-spatial container" (NSC) factor — *experiences happen in something they are presented to* — separate from the perceptual spatial frame (PSF) and body-boundedness (BB) in non-meditators, and strengthen with introspective experience? The note's own falsification conditions require items good enough that their failure is informative. This document (1) drafts a candidate item pool, (2) attacks each item as a hostile reviewer would, (3) issues survival verdicts, (4) states design requirements, including the pre-registered discriminant between "no container factor exists" and "our items failed."

Response format assumed throughout unless stated: 7-point agree–disagree Likert with two off-scale options, **[?] "I can't tell"** and **[X] "This question doesn't make sense to me"**, both always available and analysed as data (see §4.2). All items instructed with: *"Answer about how things actually seem to you in your own experience, not about how people usually talk."* (This instruction is itself attacked in §3.5.)

**Verification status of sources.** Verified this session against the live literature: Sytsma & Machery (2010, *Phil Studies* 151:299–327 — folk ascribe "seeing red" to a simple robot but not felt states; folk conception tracks valence, not phenomenality); Sytsma & Ozdemir (2019, *JCS* 26(9–10):241–256 — problem intuitions not widespread); Forstmann & Burgmer, "The Cartesian Folk Theater" (N=2,057; consciousness conceptualised as spatio-temporally localised in a confined prefrontal region); Gamma & Metzinger (2021, *PLOS ONE* 16(7):e0253694; 12 factors incl. Witness Consciousness; Emptiness and Non-egoic Self-awareness; Sensory Perception in Body and Space; Luminosity); Starmans & Bloom (2012, *Cognition* 123:313–318 — self located at/near the eyes, incl. alien with eyes on chest); Bertossa et al. (2008, *Perceptual and Motor Skills* 107:323–335 — 83% locate the "I-that-perceives" behind the eyes; same result in congenitally blind and non-Western subjects); Choi & Bowerman (Korean tight-fit *kkita* vs loose containment crosscuts English in/on; language-specific by 18–29 months); Pennycook et al. (2015, *JDM* 10(6):549–563, pseudo-profound bullshit receptivity); Longo, Schüür, Kammers, Tsakiris & Haggard (2008, *Cognition* 107:978–998 — ownership, location, agency as separable embodiment components); Graham, Taylor, Olchowski & Cumsille (2006, *Psych Methods* 11:323–343, 3-form planned missing design); Yoon and related work on Korean *maum* vs English *heart/mind* (maum has no English counterpart; cognition sits in *meori* "head"). **Not re-verified this session** (marked where relied on): Kövecses on metaphor variation; Willis on cognitive interviewing; Billiet & McClendon on acquiescence factors; East Asian response-style literature (midpoint responding). Treat those four as standard-but-unchecked.

---

## 1. The item pool (24 candidate items + 1 forced-choice probe)

### 1.1 PSF — Perceptual spatial frame (5 items; 2 reverse-coded)

Construct: *things appear arranged around me*. This should be near-ceiling in everyone; its job is to be the thing NSC must separate from, and (with BB) to serve as an internal positive control (§4.3).

- **PSF1.** When I look and listen, things seem laid out around me — some to my left, some to my right, some near, some far.
- **PSF2.** Sounds seem to reach me from particular directions.
- **PSF3.** What I perceive seems arranged around a "here" that I am at.
- **PSF4 (R).** The things I see and hear don't seem to be in any particular place relative to me.
- **PSF5 (R).** What I perceive seems like a jumble with no arrangement — nothing nearer or farther, no directions.

### 1.2 BB — Body-boundedness (5 items; 2 reverse-coded)

Construct: *experience feels located in or near my body*. Anchored in Starmans & Bloom (egocenter at the eyes) and Bertossa et al. (behind the eyes, incl. blind subjects).

- **BB1.** My experiencing seems to happen where my body is.
- **BB2.** If I had to point to the place where my experiencing happens, I would point to a part of my body.
- **BB3.** I seem to be located somewhere behind my eyes.
- **BB4 (R).** My experiencing doesn't seem to be in my body or anywhere near it.
- **BB5 (R).** My experiencing seems no more located in my body than anywhere else.

### 1.3 NSC — Non-spatial container (8 Likert items + 1 forced choice; 2 reverse-coded)

Construct: *experiences happen in/to something they are presented to*, where that something is not (merely) the perceptual frame or the body. The pool is deliberately split into two **metaphor-family facets**, pre-registered as such, because — as §3.3 argues — the construct cannot be worded metaphor-free in English. The facets use *incompatible* metaphor families; convergence across them is the signal (§3.3).

**Dative facet (theater/audience family: presented-to, shown-to, registered-by):**

- **NSC1.** My thoughts, feelings, and perceptions seem to be presented to something.
- **NSC2.** Individual experiences come and go, but whatever they come and go *for* seems to stay.
- **NSC3.** Whatever it is that registers my thoughts seems to be the same as whatever registers sounds, sights, and feelings.
- **NSC7 (R).** My experiences don't seem to be presented to anything — they simply occur, one after another.

**Venue facet (container/medium family: happens-in, room-for):**

- **NSC4.** My experiences seem to happen *in* something, the way sounds happen in a room.
- **NSC5.** Even when almost nothing is going on — no thoughts, nothing much to notice — it doesn't seem like sheer nothing; something seems ready for the next experience.
- **NSC6.** When one thought has ended and the next hasn't started, something seems to remain.
- **NSC8 (R).** There is nothing more to my experiencing than the particular sights, sounds, thoughts, and feelings themselves.

**Forced-choice probe (administered before any Likert NSC item, no scale, no midpoint):**

- **NSC-FC.** Which is closer to how things seem to you? **(a)** Experiences simply happen, one after another. **(b)** Experiences happen *to* something that is there throughout. **(c)** I can't tell.

### 1.4 MV — Metaphor-inventory / vocabulary controls (4 items + 2 tasks)

These measure the respondent's *linguistic* container inventory, so that "container reports track metaphor inventory" (the note's Lakoff kill condition) is a testable regression, not a shrug. Framed explicitly as questions about *ways of talking*.

- **MV1.** It sounds natural to say a thought was "in the back of your mind."
- **MV2.** It sounds natural to say someone "holds an idea in their head."
- **MV3 (contrast, non-container).** It sounds natural to say thoughts "race" or "flow."
- **MV4 (contrast, non-container).** It sounds natural to say an idea "struck" you.
- **MV-P (production task).** "Complete in your own words: *The name was ___ my mind, but I couldn't get it ___.*" Coded for spontaneous container prepositions.
- **MV-V (vocabulary/exposure).** (i) 10-item general vocabulary test (Wordsum-type) as verbal-ability covariate; (ii) contemplative-vocabulary exposure checklist: "Have you ever encountered these phrases? *spacious awareness / field of awareness / witness consciousness / open awareness / the space in which thoughts arise*" — yes/no per phrase, plus source (app, book, class, never). This checklist is the single most important covariate in the study (§3.5, §4.3) and must be administered **last**, since it is maximally teaching.

### 1.5 JB — Jangle bait (3 items)

Syntactically well-formed, semantically vacuous, *profundity-flavoured* items in the style of Pennycook et al. (2015). Their job: give a "sounds-deep endorsement + acquiescence" method factor something to grab, so that NSC's discriminability from that factor is testable rather than assumed.

- **JB1.** My experiences seem to arrive already tuned to a hidden frequency of their own.
- **JB2.** Each moment of experience seems to cast a shadow that arrives before it does.
- **JB3.** My awareness seems to renew itself by passing back through its own outer edge.

JB3 is deliberately *container-flavoured* nonsense: if NSC items co-load with JB3 specifically, the method-variance story wins on the container's home turf.

---

## 2. Item-level attack: hostile review with verdicts

Verdict scale: **KILL** (unsalvageable) · **REVISE-OR-KILL** (fieldable only if the stated revision passes cognitive interviewing) · **SURVIVES** (provisionally, subject to the design controls in §4).

### 2.1 PSF items

**PSF1 — SURVIVES.** Near-tautological for sighted respondents, which is the point: it defines the ceiling against which NSC must show variance somewhere else. Hostile note: "laid out around me" is itself container-adjacent ("around"); acceptable because PSF *is* the spatial construct — contamination here is measurement, not error.

**PSF2 — SURVIVES.** Cleanest PSF item. One worry: literal-minded respondents answer about acoustics ("sounds come from their sources"), a Sytsma-style deflation — but for PSF the deflationary and phenomenological readings coincide, so the deflation is harmless. This is exactly why PSF can't just be cloned to produce NSC items: PSF survives deflation *because* it is about the world-as-arranged; NSC does not.

**PSF3 — REVISE-OR-KILL.** "A 'here' that I am at" flirts with the egocenter (BB territory: Starmans & Bloom's eye-point) and will cross-load on BB. Revision: *"Everything I perceive seems to be some direction and distance away from me."* Keeps frame, drops the point-self. Cognitive interviews must check that respondents aren't answering about their body position (which would still be fine — PSF/BB cross-loading is expected and modelled — but we should know).

**PSF4 (R) — SURVIVES**, barely. Reverse items describing absent structure invite "huh?" responses, and near-universal disagreement gives it almost no variance. Retained anyway: its function is acquiescence control within PSF, and its *endorsement* in the high-practice stratum is a bonus signal (frame destabilisation is reported in MPE data — Sensory Perception in Body and Space factor).

**PSF5 (R) — KILL.** "Jumble" describes disordered *content*, not absent *frame*; a respondent with a messy desk and ADHD can endorse this truthfully with the frame fully intact. Confound between disorder-within-the-frame and absence-of-frame is fatal and no rewording fixes it without duplicating PSF4. Replace with a second reverse item only if PSF4's cognitive-interview performance forces it.

### 2.2 BB items

**BB1 — SURVIVES.** Direct statement of the construct. Deflationary reading ("obviously — I'm where my body is") and phenomenological reading agree for most respondents; divergence is exactly what the high-practice stratum should show. Acceptable.

**BB2 — REVISE-OR-KILL.** The pointing frame is borrowed from Bertossa et al., but their result required a structured interview converging over many minutes; as a cold survey item, the *instruction to point* may **create** the point it claims to measure — a one-item demand characteristic. Revision: precede with a filter: *"Does the question 'where does your experiencing happen?' seem answerable to you?"* [yes/no/not sure], then BB2 only for yes-respondents. If the filter can't be added (form-length), kill; the created-point risk contaminates the trajectory claim (point → container) at its origin.

**BB3 — SURVIVES**, with a flag. Strongly supported anchor (Starmans & Bloom; Bertossa; Forstmann & Burgmer's confined prefrontal locus). Flag: it measures the *egocenter point*, a sub-facet of BB, and may split from BB1/BB2 (body-volume) in factoring. Pre-register: BB may be two-dimensional (point vs volume); either structure is fine for Study B's purposes, but the pre-registration must say so now, or a reviewer will call the post-hoc accommodation a degree of freedom.

**BB4 (R) — SURVIVES.** Floor effects expected in non-meditators (good — that's the prediction), and it does honest acquiescence work.

**BB5 (R) — KILL.** "No more located in my body than anywhere else" requires the respondent to run a comparative quantification over locations of experience — a philosopher's operation. Cognitive interviews will return paraphrases like "my mind wanders a lot." Nothing here that BB4 doesn't do more simply.

### 2.3 NSC items — the hard part

Cross-cutting attacks first; per-item verdicts after.

**Attack A (Sytsma deflation).** For every dative item there exists a reading on which "something" = *me, the person* — and on that reading every item is trivially true and phenomenologically empty. Sytsma & Machery's result is precisely that the folk operate with an agent-level concept and no distinct phenomenal one; Sytsma & Ozdemir extend it cross-culturally. A hostile reviewer's default model: **NSC endorsement = reading-comprehension noise around the trivially-true person-level reading.** No wording eliminates this; adding "something other than yourself, and other than the experiences themselves" teaches the frame outright (it's a meditation instruction with the serial numbers filed off). The only honest countermeasures are (i) cognitive-interview coding of *which* reading respondents used, with a pre-registered ceiling on person-level readings (§4.1), and (ii) the prediction that deflationary reading rates should *fall* with practice strata — which is itself trajectory evidence if it holds.

**Attack B (metaphor contamination — and the honest concession).** Can NSC be stated without spatial vocabulary? **No.** Attempted wordings exhaust two metaphor families: the *container/medium* family (in, within, room, space, field — Lakoff & Johnson's MIND IS A CONTAINER directly) and the *theater/audience* family (presented, shown, appears-to, witnessed, registered — the Cartesian theater, Dennett's own term of abuse, and note that Forstmann & Burgmer's title concedes the folk already live there). Every "clean" paraphrase we generated in drafting reduced to one family or the other; the dative escape from spatial prepositions lands squarely in theater vocabulary. **Design consequence rather than defeat:** pre-register the two facets as deliberately incompatible metaphor families and let structure adjudicate. If a real phenomenological structure exists upstream of metaphor, dative and venue items should form one factor *cutting across* families. If metaphor inventory drives responses, items should cluster *by family* and covary with MV scores. This turns the impossibility of neutral wording into the test itself. It also means NSC needs both facets fielded; neither alone is interpretable.

**Attack C (no basis → midpoint pileup).** Respondents with no relevant phenomenology will not leave blanks; they will select 4-of-7, and midpoint pileup factored across NSC items *looks like* a weak common factor. This is the single most likely way Study B produces a false-positive container. Countermeasure: the off-scale [?]/[X] options, forced-choice NSC-FC (no midpoint), and the pre-registered rule that an NSC "factor" built disproportionately on midpoint responses does not count (define: factor invalid if median absolute item response of its top-loading tercile is the scale midpoint and IRT information peaks at trait levels where [?]/[X] rates exceed 25%).

Per-item verdicts:

**NSC1 — REVISE-OR-KILL.** The flagship, and Attack A hits it hardest: "presented to something" → "yes, to me" (trivial) or "to my brain" (Cartesian Folk Theater reading — Forstmann & Burgmer show that reading is the folk default, which would make NSC1 load with BB3, not separately). Revision: *"My thoughts, feelings, and perceptions seem to be presented to something — not just stored or happening, but presented."* The contrast clause gives the deflationary reader an exit (mere happening) without teaching the container. Must pass cognitive interviews with <30% person/brain readings; else kill, and NSC-FC inherits the flagship role.

**NSC2 — SURVIVES**, the best item in the pool. Persistence-through-change is stateable with a dative ("come and go *for*") and has a natural deflationary reading ("I'm still here between thoughts") that is *not* trivially identical to the target reading — the difference is testable in interviews. It is also the item closest to the darkening mechanism's actual claim (a standing remainder). Weakness: "for" is doing subtle work and will strain translation (§3.4).

**NSC3 — KILL.** "Same thing registers thoughts and sounds" is trivially true under the person reading ("yes — me"), and near-universal endorsement is guaranteed; ceiling effects carry no discriminating information, and the item is a paraphrase of the MPE Witness Consciousness factor's contemplative framing — i.e., it is harvested vocabulary, exactly what the note warns Study A items would be. It cannot separate trained vocabulary from structure. Kill without replacement; NSC2 covers the defensible content.

**NSC4 — SURVIVES, but reclassified.** As a measure of NSC it is indefensible: "in something, the way sounds happen in a room" is an explicit container simile — it *teaches the frame in the item stem* and is maximally Lakoff-contaminated. Retain it **as the pre-registered contamination anchor**: NSC4 should be the *most* metaphor-driven item, so its loading pattern calibrates the contamination axis (predicted: highest correlation with MV of any NSC item; if the "real" NSC items behave like NSC4, the Lakoff story wins). It must appear *last* among NSC items in every form (§4.2) since it teaches. Never score it on the NSC factor for the existence verdict.

**NSC5 — REVISE-OR-KILL.** Target: capacity/readiness without "space" or "room" — the drafting successfully avoids both, which makes it precious. Two live objections: (i) "something seems ready" invites a dispositional reading ("I can always start thinking again" — a fact about capacities, not phenomenology; deflation again); (ii) the setup clause ("no thoughts, nothing much to notice") asks non-meditators to report on a state they may never have noticed occupying, so [X] rates will be high — which is acceptable *if* [X] is analysed as data (predicted to fall with practice). Revision: split the compound — *"Moments when almost nothing is going on still don't seem like sheer nothing."* Field both versions in the pilot; keep the better interviewer.

**NSC6 — KILL.** "When one thought has ended and the next hasn't started, something seems to remain" is a gap-observation instruction — it is recognisably *the* pointing-out instruction of several contemplative lineages (the note's own Okumura provenance should make the author wince). For anyone with app-level exposure it is a vocabulary-recognition item; for the untrained it demands a micro-phenomenological observation a survey cannot scaffold. Either way it measures training, not structure, and it teaches the very trajectory the study wants to test. This is the most dangerous item in the pool because it would probably "work" — producing exactly the practice-correlated loading the note predicts — for entirely artefactual reasons.

**NSC7 (R) — SURVIVES.** Necessary acquiescence control for the dative facet. Note: under Attack A, disagreement with NSC7 is the trivially-safe response ("of course they happen to *me*"), so its reverse-scored distribution will be skewed; retained anyway, with the skew pre-registered as expected.

**NSC8 (R) — KILL.** "Nothing more to my experiencing than the particular sights, sounds…" is a bundle-theory exam question. It requires the respondent to (i) form the totality of their experiences, (ii) consider whether something exceeds it — a reflective operation that cognitive interviewing will show ordinary respondents cannot parse stably. Worse, sophisticated respondents will answer from *belief* (religious or physicalist commitments), not seeming. Replace the reverse-coding burden on the venue facet with NSC7's dative twin: *"Nothing seems to stay the same from one experience to the next"* — call it **NSC8′**, verdict REVISE-OR-KILL pending interviews (it may collapse into a mood/instability reading).

**NSC-FC — SURVIVES.** The forced choice removes acquiescence and midpoint refuge entirely, and (b) is stated datively without container prepositions. Its position (before all Likert NSC items, for every respondent who receives it) is load-bearing. Hostile note: option (b) is longer and more interesting-sounding than (a) — pseudo-profundity pull inflates (b). Countermeasure: match lengths, and validate against JB endorsement (respondents high on JB who choose (b) discount its evidential weight in the pre-registered model).

### 2.4 MV items

**MV1, MV2 — SURVIVE.** Straightforward naturalness judgments; anchored in the conventional inventory Lakoff & Johnson catalogued.
**MV3, MV4 — SURVIVE.** Non-container contrasts are what make MV a *profile* rather than a general metaphor-tolerance score; without them, MV is confounded with acquiescence and verbal ability.
**MV-P — SURVIVES**, with coding burden noted (two coders, kappa ≥ .8 pre-registered). Production beats recognition: it samples the respondent's *active* inventory, which is what the Lakoff kill condition is about.
**MV-V — SURVIVES**; the exposure checklist is non-optional (it is the direct measure for the note's second objection — vocabulary acquired mid-training) and must sit at the very end of every form.

### 2.5 JB items

**JB1 — REVISE-OR-KILL.** Flawed lure: "arrive already tuned… before I notice them" has a *true* scientific reading (preconscious processing) that a thoughtful respondent can legitimately endorse. A lure with a defensible reading isn't a lure. Revision toward pure Pennycook vacuity: *"My experiences seem to balance their own becoming against the texture of what they are not."* Ugly, which is correct.
**JB2 — SURVIVES.** Temporal nonsense with no rescue reading; profundity-flavoured; good.
**JB3 — SURVIVES.** The container-flavoured lure is the pool's best trap and must be retained exactly because it is unfair to our own hypothesis.

### 2.6 Verdict summary

| Item | Verdict | | Item | Verdict |
|---|---|---|---|---|
| PSF1 | SURVIVES | | NSC4 | SURVIVES (as contamination anchor only) |
| PSF2 | SURVIVES | | NSC5 | REVISE-OR-KILL |
| PSF3 | REVISE-OR-KILL | | NSC6 | **KILL** |
| PSF4 | SURVIVES | | NSC7 | SURVIVES |
| PSF5 | **KILL** | | NSC8 | **KILL** (→ NSC8′ REVISE-OR-KILL) |
| BB1 | SURVIVES | | NSC-FC | SURVIVES |
| BB2 | REVISE-OR-KILL | | MV1–4 | SURVIVE |
| BB3 | SURVIVES (flag: point vs volume) | | MV-P | SURVIVES |
| BB4 | SURVIVES | | MV-V | SURVIVES (must be last) |
| BB5 | **KILL** | | JB1 | REVISE-OR-KILL |
| NSC1 | REVISE-OR-KILL | | JB2 | SURVIVES |
| NSC2 | SURVIVES | | JB3 | SURVIVES |
| NSC3 | **KILL** | | | |

Net: 5 kills (PSF5, BB5, NSC3, NSC6, NSC8), 7 revise-or-kill, 13 provisional survivors. The NSC scale enters piloting with, at best, 2 clean survivors (NSC2, NSC7) + NSC-FC + 3 conditional items — thin, and that thinness is a finding: **the construct resists itemisation**, and Study B's pre-registration must budget for a second drafting round after cognitive interviews rather than pretending this pool is final.

---

## 3. Cross-cutting attacks

### 3.1 Sytsma-style deflation (summary of Attack A)
Every NSC item has a person-level or brain-level reading under which endorsement is trivial. Given Sytsma & Machery (2010) and Sytsma & Ozdemir (2019), the *default* hypothesis must be that non-meditator NSC responses are noise around such readings. Study B's verdicts are uninterpretable without interview-based reading-coding and the pre-registered deflation ceiling (§4.1). Additionally, Forstmann & Burgmer's result cuts *against* the note in one respect worth stating plainly: the folk theater they document is **in the brain** — a located venue. If NSC items separate but correlate strongly with BB3, the "non-spatial" qualifier dies even if "container" survives, and the note's trajectory becomes point → *located theater* → collapse, a materially different claim. Pre-register this as a named alternative outcome, not a failure.

### 3.2 Midpoint/no-basis artefact (Attack C)
See §2.3. The false-positive risk is as serious as the false-negative risk the note worries about, and the invalidation rule must be pre-registered symmetrically.

### 3.3 Metaphor contamination and the impossibility of neutral wording (Attack B)
Conceded and weaponised: two incompatible metaphor families, factor structure adjudicates, MV profiles the inventory, NSC4 anchors the contamination axis. One further requirement: the *Lakoff regression* must be specified now — NSC factor scores regressed on MV recognition, MV-P production, MV-V exposure, and verbal ability; the note's kill condition ("container reports track metaphor inventory") fires if MV-family variables absorb the NSC–stratum relationship (pre-register the mediation model and the proportion-mediated threshold, suggested ≥60%).

### 3.4 Translatability: the Korean test
Korean is the chosen concrete test language: non-Indo-European, huge survey-panel infrastructure, and two directly relevant, well-documented properties.

**(i) The mind-lexicon problem.** Korean offers no neutral "mind": *maum* (마음) is the heart-mind — chest-associated, moral-emotional, explicitly documented as having no English counterpart — while cognition sits in *meori* (머리, head); *uisik* (의식, "consciousness") is clinical register (losing/regaining consciousness); *allachalim* (알아차림, "awareness") is meditation-app register and would import the contemplative vocabulary the study must control. Consequence: any NSC item mentioning mind/awareness forces a lexical choice that pre-loads either BB (chest vs head anchoring — note *maum-sok*, "inside the maum," is a high-frequency conventional collocation, so the venue facet is *more* idiom-saturated in Korean than English) or training vocabulary. Mitigation: the surviving NSC items (NSC1-rev, NSC2, NSC5, NSC7, NSC-FC) already avoid mind-words entirely — this must be preserved as a hard drafting constraint precisely because of the Korean test. That the best English items happen to be the translatable ones is weak evidence the facet-split is carving something real.

**(ii) The grammar problem.** Choi & Bowerman: Korean containment semantics carve by tight fit (*kkita*) vs loose fit, crosscutting English in/on, and Korean children are locked into these categories before age three. "Experiences happen *in* something" translated with loose-containment forms (*ane*, *soge*) evokes a *bounded physical* container more concretely than English's bleached "in" — the venue facet gets *more* spatial in translation, a directional prediction: if NSC is metaphor-driven, Korean venue items should diverge further from dative items than English ones do; if structure-driven, the facet correlation should survive translation (pre-register this interaction). Meanwhile the dative facet hits animacy: the recipient markers *-ege/-hante* prefer animate recipients, so "presented to something" forces either an odd animate-marked *something* — which cognitive interviews must check for spirit/deity readings (religiosity covariate mandatory in both languages, amplified here) — or the locative-ish *-e*, which partially collapses the dative facet into the venue facet. **The theater/container distinction that carries the whole English design is grammatically harder to draw in Korean.** This is not a reason to drop Korean; it is the reason to include it — a facet structure that only exists where the grammar makes it cheap to express is Lakoff's result, not ours. Requirements: forward-and-back translation with committee adjudication; full cognitive-interview cycle in Korean (not translated probes); measurement invariance testing (configural → metric → scalar) pre-registered as a gate before any cross-language comparison; midpoint-response-style differences (East Asian response-style literature — **unverified this session**) handled via the forced-choice item and response-style modelling, not ignored.

### 3.5 Demand and teaching effects from item order
The PSF and BB blocks ask *where* questions; twelve consecutive where-questions train a spatial frame, and NSC's "something" then inherits a location by momentum. Within NSC, NSC4's room simile teaches the container reading of every later dative item; the MV block teaches container idioms; MV-V's phrase checklist is effectively a syllabus. Even the general instruction ("how things actually seem to you, not how people usually talk") is a demand cue — it announces that a non-obvious answer is wanted, inviting Gricean depth-seeking. None of this is fully removable; all of it is measurable (§4.2).

---

## 4. Design requirements

### 4.1 Cognitive interviewing (gate 1 — nothing fields without it)
- Two rounds minimum, n = 30 per language per round, stratified (≥60% never-meditators), think-aloud plus scripted verbal probing (Willis-style protocol — **unverified this session**): "What is this question asking, in your own words?", "What did you think of when it said *something*?", "Can you tell me about a time it seemed that way?"
- Coding scheme fixed in advance: for each NSC item, code the reading as {person/self, brain, container-venue, dative-witness, religious, dispositional, no-stable-reading}. **Gates:** ≥80% stable paraphrase adequacy; ≤30% person/brain readings for any NSC item retained on the factor (NSC4 exempt — it is the contamination anchor); items failing get one revision cycle, then die. Two coders, kappa ≥ .8.
- Interview data are also the first trajectory datum: pre-register the prediction that deflationary readings decrease across practice strata.

### 4.2 Fielding design: exposure control and order arms
- **3-form planned missingness** (Graham et al. 2006): NSC Likert items split across forms so no respondent sees more than 3 of them (X-block: NSC-FC + demographics + strata, seen by all). This caps within-scale teaching (earlier NSC items scaffolding later ones), shortens forms, and the missingness is MCAR by design for FIML/MI estimation.
- **Between-subjects order arms** (randomised): Arm 1 NSC-first (cleanest existence test — no spatial warm-up); Arm 2 standard order (PSF→BB→NSC); Arm 3 NSC-last. Pre-registered moderation test: NSC loadings and means across arms; an arm effect exceeding the pre-set threshold (suggested: ΔM > 0.3 SD or loading Δ > .15) establishes the teaching confound, and only Arm 1 data then count toward the existence verdict.
- NSC-FC always precedes all Likert NSC content for those who receive it; NSC4 always last within NSC; MV-V always last in the whole instrument. [?]/[X] options on every Likert item; their rates analysed as substantive data (predicted: NSC [X]-rates fall with practice stratum; flat high [X] across strata = items failed signal).
- No definitions, no examples, no glosses anywhere. If an item can't stand without a gloss, it's dead (this killed NSC6).

### 4.3 The pre-registered discriminant: "no container factor exists" vs "our items failed"
The hardest problem. Proposal — a run only *counts* (either direction) if the instrument demonstrates, in the same sample and method, that it can resolve distinctions of comparable subtlety:

1. **Positive-control construct pair, matched for abstraction:** include 6 trait-adapted ownership/agency items (from Longo et al. 2008's separable components, restated outside the illusion context, e.g. "My body feels like it belongs to me" vs "I feel in control of what my body does"), drafted at the same reading level and abstraction as NSC items (verified by matched Flesch scores and interview [X]-rates within 10 points). **Gate:** ownership and agency must separate (distinct factors, latent r < .8, ΔBIC ≥ 10 for the 2-factor model). If they don't, the sample/method can't resolve subtle phenomenal distinctions and the run is void — no verdict either way.
2. **Internal positive control:** PSF and BB must separate from each other. If they don't, void.
3. **Negative control:** JB items must be absorbed by a method factor (modelled via random-intercept/acquiescence factor — Billiet & McClendon-style, **unverified this session** — plus JB group factor); every retained NSC item's loading on the JB/method factor must be < .3. NSC merging with JB = "items failed," at best.
4. **Known-groups coherence gate:** a supplementary experienced-meditator arm (n ≈ 300, >500h). NSC items must cohere there (ω ≥ .75) and separate from PSF/BB. If NSC doesn't even work where the construct is *known to be reportable* (MPE-92M documents the vocabulary and factors in this population), the items failed — no existence verdict. If it works there and gates 1–3 pass, then absence in non-meditators is *interpretable*: either genuine absence (supporting point-first trajectory) or the strongest possible evidence the note can get short of longitudinal data.
5. **Equivalence criterion for declared absence:** "no container factor" may be declared only when, in the non-meditator arm: the model adding an NSC group factor fails to improve on the PSF+BB(+method) model (ΔBIC ≥ 10 favouring the simpler model), NSC-specific ω-hierarchical < .20, *and* the midpoint-invalidation rule (§2.3, Attack C) does not itself explain the null. Power for this must be simulated pre-registration-side: ≥90% power to detect an NSC factor with mean standardised loading .50 at the planned N under the planned missingness pattern (simulation code in the pre-registration).
6. **Decision table (pre-registered verbatim):**

| Controls 1–3 | Meditator coherence (4) | Non-meditator NSC factor | Verdict |
|---|---|---|---|
| fail | — | — | Run void; redesign |
| pass | fail | — | Items failed; redraft, no existence verdict |
| pass | pass | present, separable, strengthens with strata | Claim survives; proceed to C/D |
| pass | pass | present but collapses into BB3/located-theater | "Non-spatial" dies; trajectory revised (§3.1) |
| pass | pass | absent under criterion 5 | Genuine absence in untrained cognition: point-default confirmed; container-as-trained-structure becomes the claim, and the vocabulary-vs-stage question (note's objection 2) becomes the *whole* question |
| pass | pass | present but Lakoff regression (§3.3) absorbs strata effect | Phenomenon is Lakoff's; drawer |

### 4.4 Sample and stratification
- **Cognitive interviews:** 2 × 30 per language (§4.1).
- **Pilot:** n = 400 per language, full item pool incl. revise-or-kill variants; item-level screening (distributions, [?]/[X] rates, preliminary EFA).
- **Main study:** n ≈ 1,800 per language (English, Korean), powered by simulation for the §4.3(5) equivalence test under 3-form missingness. Strata by introspective experience, quota-sampled: never practiced and no app exposure (50%); casual/app < 50h (20%); 50–500h (20%); > 500h incl. retreat experience (10%). Plus the dedicated meditator arm (n ≈ 300, > 500h) for gate 4. Practice hours self-reported with instrument-anchored questions (apps used, retreat-days), because "introspective experience" must itself not be measured with contemplative vocabulary.
- **Covariates, all arms:** vocabulary (Wordsum-type), contemplative-vocabulary exposure checklist (MV-V), religiosity, pseudo-profound receptivity (JB scores double as short-form), age, education; Korean arm: Korean-dominance screening (exclude balanced English-dominant bilinguals from primary analysis — bilinguals show spatial-category shift; flag, don't discard).
- **Cross-language:** measurement invariance sequence as a gate (§3.4); comparisons of latent means only if scalar invariance (or justified partial invariance) holds.

### 4.5 What this red-team changes in the note, if accepted
1. NSC must be fielded as two metaphor-family facets with cross-family convergence as the signal; the note's single-sentence operationalisation ("experiences happen in something they are presented to") conflates the two families in one breath — *happen in* (container) + *presented to* (theater) — and should be rewritten as the disjunction it is.
2. The "non-spatial" qualifier needs a pre-registered fate for the located-theater outcome (§3.1); Forstmann & Burgmer make that outcome likely enough to name.
3. The falsification condition "no separable container factor in B" is under-specified as written; replace with the §4.3 decision table, or the drawer-verdict will be contestable in exactly the way the note is trying to avoid.

---

## 5. Key sources for this document
Sytsma & Machery 2010 (*Phil Studies*); Sytsma & Ozdemir 2019 (*JCS*); Forstmann & Burgmer, Cartesian Folk Theater (*Consciousness & Cognition*, N=2,057); Gamma & Metzinger 2021 (*PLOS ONE*, MPE-92M); Starmans & Bloom 2012 (*Cognition*); Bertossa et al. 2008 (*Perceptual & Motor Skills*); Choi & Bowerman (Korean spatial semantics, tight/loose fit); Yoon, *maum* vs *heart/mind* (NSM contrastive semantics); Pennycook et al. 2015 (*JDM*); Longo et al. 2008 (*Cognition*); Graham et al. 2006 (*Psych Methods*); Lakoff & Johnson 1980. Unverified this session and flagged inline: Kövecses; Willis (cognitive interviewing); Billiet & McClendon (acquiescence factor); East Asian response styles.