# H1 on public data: do Zen and TM practitioners' reports of pure awareness lean in the paper's two directions?

**Type:** pre-registration (lab type) · **Date:** 3 October 2026 · **Status:** filed before any item was coded and before any outcome variable was examined · **Authors and models:** design by Claude Opus 5.5 (editor); item coding by two independent coders, Claude Sonnet 5.5 and GPT-6 Astra; analysis code by Claude Sonnet 5.5, checked by the editor; outside read of the result by GPT-6 Astra · **Authority:** the author, "Proceed as you see fit to move the project forward" (3 October 2026), on the editor's proposal to examine this dataset · **Cap:** USD 3 in Claude Sonnet 5.5 tokens; two Codex calls (one coder, one read)

## What was looked at before filing

The dataset's README, the English questionnaire, the list of variable names and labels, the response format of the items (0 to 100 sliders), and the number of participants in each practice group. No outcome variable (no MPE item response) was summarised, plotted or modelled.

## The data

Gamma and Metzinger (2021), "The Minimal Phenomenal Experience questionnaire (MPE-92M): Towards a phenomenological profile of 'pure awareness' experiences in meditators", PLOS ONE 16(7), e0253694, with the 2024 correction (e0314290) giving the data link. File `MPE92M_data_set_quesionnaires_documentation.zip`, Open Science Framework node `cbhdv`, https://osf.io/xerhg/download, SHA-256 `bb6112fcc5e19cf86b43c511b906039d6c8d43c5002bab08cd11fb62ce74fa17`, 3,627 responses, 92 items on a recalled experience of pure awareness. The project page states no licence; the data are kept local and are not redistributed by this project. Analysis sample: `over85items == 1` and `hadpureawareness == 1` (n = 1,386).

## The hypothesis, from the paper's own text

Part IV of the working paper: "Zen's intimacy with *this*. Advaita's dissolution into *that*." On the paper's account the two primitive orientations are toward represented content (*this*) and toward representational activity, the modelling itself (*that*). H1 as served since v2.3: an interpretive pattern proposed for testing, whose defeat criterion needs a pre-registered predicted contrast, validated coding and enough precision to tell a negative result from an inconclusive null.

**Groups.** Zen: `medit_tech_zen == 1`, with TM and Mahamudra/Dzogchen both 0 (n = 372). Advaita proxy: TM, `medit_tech_tm == 1`, with Zen and Mahamudra/Dzogchen both 0 (n = 128). TM is used as the Advaita proxy because its teaching derives from Advaita Vedanta; the dataset has no Advaita category. Mahamudra/Dzogchen (n = 169 on the same rule) is not assigned a direction by the paper and is reported as exploratory only.

**Scales.** Two independent coders, blind to all response data, assign each of the 92 items, from its full English wording, to one of: **C**, the experience as described is oriented toward content (objects, perceptions, the body, the world, place, time as experienced); **A**, oriented toward the activity of being aware itself (awareness as its own content, knowing without an object, looking from awareness, self-luminosity of awareness); **N**, neither or both, including mood, wakefulness, effort and agency items. An item enters a scale only if both coders assign it the same code. Agreement is reported as Cohen's kappa over the three codes. **Stop rule:** if kappa is below 0.60, or either scale has fewer than four agreed items, the test is not run and that result is reported. Items worded so that a high score means absence of the quality (for example "Did you have thoughts?" scored toward content) are coded by what a high score indicates.

**Score.** For each participant, C = mean of available C items and A = mean of available A items (0 to 100), each requiring at least 75 percent of its items answered; orientation index D = C − A.

## Prediction and decision rule (primary)

Δ = mean D(Zen) − mean D(TM), Welch 95 percent confidence interval. The paper's reading predicts Δ > 0 (Zen leans toward content relative to TM). Smallest difference of interest: 5 points on the 0 to 100 scale.

- **Supported:** the 95 percent interval lies wholly above 0 and Δ ≥ 5.
- **Weakly supported:** the interval lies above 0 and Δ < 5 (a difference smaller than the one that matters).
- **Counts against (reversal):** the interval lies wholly below 0.
- **Negative result:** the 90 percent interval lies wholly within −5 to +5 (no difference of the size that matters).
- **Inconclusive:** anything else.

Expected precision, assuming a standard deviation of D near 20: standard error about 2 points, so a 95 percent half-width near 4. This is an assumption; the realised precision is reported.

## Secondary and exploratory

- **Secondary (pre-registered):** the same contrast as a linear regression of D on group, with log lifetime practice hours (`medit_hours_total`, plus 1), age, sex and questionnaire language as covariates; and the contrast with Zen and TM defined without excluding Mahamudra/Dzogchen practice (participants practising both Zen and TM still excluded).
- **Exploratory, labelled as such:** the Mahamudra/Dzogchen group's mean D; the share in each group reporting that the experience occurred in dreamless deep sleep (`mpe83_dreamless_sleep` above 50), descriptive only.

## What a result can and cannot show

It can show whether recalled pure-awareness experiences of Zen and TM practitioners differ in the direction the paper's two-direction reading predicts, on coding fixed in advance. It cannot show that the two orientations are primitive, that there are exactly two, or anything about the fold. **A limit stated before the run:** the questionnaire asks about experiences of pure awareness, the *that* end; a Zen orientation toward *this* may not appear in such reports at all, so a negative or inconclusive result is weak evidence against the paper's reading, and a reversal is stronger. Practice technique is not tradition, TM is a proxy for Advaita, and participants self-selected into an online survey. Nothing here attaches to a claim node; a disconfirming result enters the dated-decisions table for the author.

## What would count against this

- The coders sharing a reading of the paper that builds the prediction into the coding: they code from item wording, blind to data, but both know the scale definitions; their disagreements are reported item by item.
- TM being a poor proxy for Advaita's "dissolution into that": the result then bears only on TM.
- The instrument's limit above, which the result's reading must carry.