# Mehrabian 7-38-55 research packet

- Object ID: `em:research-note:sha256:29c1b707457182707c21766d15835e5d0a76fc840c67d1f9242b51001c099285`
- Kind: `research-note`
- Repository path: [`research/how-we-know/mehrabian-7-38-55/README.md`](https://github.com/yoheinakajima/epistemedia/blob/f92846570180dfa4511263f8ba98ecd18f7772c9/research/how-we-know/mehrabian-7-38-55/README.md)
- Content digest: `c9f576ddad5127111a56d41a8615ca4e2e417e38357814b6c643475017ac9d5d`

**Also filed under:** [Research Program](https://epistemedia.org/topics/research-program/)

## Source content

# Mehrabian 7-38-55 research packet

Status: author packet complete; fresh-clone independent review pending. This is research input,
not an admitted How We Know dossier, public verdict, or deployment.

Task: `EM-0033`

Target question:

> What did the 7-38-55 weighting represent in the examined primary editions, how was it
> derived, and what supports or undercuts its circulation as a general rule of communication?

Evidence cutoff: `2026-08-27`

## Current author recommendation

**GO, pending independent review**, for a later Case 004 dossier about a narrow historical
weighting that was generalized beyond its source conditions. The packet's useful result is not
another context-free percentage:

- P1 crossed isolated words with vocal tone in an inconsistent attitude task;
- P2 crossed facial and vocal attitude using the neutral word `maybe`;
- P2's discussion proposed `.07/.38/.55` by integrating the linked studies rather than testing
  three channels in one experiment;
- the 1971 book calls the numbers approximate and limits the equation to one feeling dimension;
- the author's public qualification says the equation does not apply outside feelings or
  attitudes;
- a direct all-channel rebuttal reported message-specific equations rather than one fixed rule;
  and
- the exact source calculation for the `.07` verbal coefficient remains unresolved and receives
  no credit.

The broad proposition that 93 percent of all communication is nonverbal is explicitly prohibited.
Institutional and web-sample recirculation objects receive zero scientific evidence credit.

## Packet identity

- candidate packet:
  `em:research-packet:sha256:a73dc29f0a0c3f05a112b7c392d115cfcc38a1136a683325bf74bebf0c6b2e40`;
- 12 source records and 40 quote-minimal or explicitly structured spans;
- 11 scoped claims and 6 deterministic derivations;
- 6 lineage groups, 5 participant-data roots, and 11 typed evidence-linked dependence edges;
- 3 propagation objects with zero scientific-rule credit; and
- 5 bounded follow-up dispositions.

The two 1967 papers remain distinct participant-data roots within one linked UCLA research
program. Shared author, grant, stimulus, material, method, scale, and citation relations prevent
them from being presented as independent replications of the circulated rule. Unknown participant
or speaker overlap remains unknown.

## Files

- `source-records.json` — source, edition, capture, license, span, claim, lineage, propagation,
  negative-search, and limitation records;
- `candidate-packet.json` — deterministic content-addressed packet;
- `build_packet.py` — offline deterministic builder and structural source validator;
- `verify_packet.py` — exact set/count/math/lineage verifier, adversarial receipt tests, and
  fail-closed independent-review gate; and
- `independent-review-receipt.json` — absent until a separate reviewer completes exact-head
  review.

No source PDF, HTML, XML, recording, photograph, participant data, or restricted full text is
committed. Readable but unlicensed carriers are represented only by identity, digest, access
history, locator, and quote-minimal spans.

## Reproduce

```bash
python research/how-we-know/mehrabian-7-38-55/build_packet.py --check
python research/how-we-know/mehrabian-7-38-55/verify_packet.py --self-test
python research/how-we-know/mehrabian-7-38-55/verify_packet.py --require-review
make check
```

Before independent review, the third command must fail with
`independent review receipt missing`. After review, it must bind the exact base, author head,
tree, packet bytes, complete source/capture/span/claim/calculation/lineage coverage, command-output
digests, limitations, clean state, and recommendation. Forged, incomplete, self-authored, stale,
or open-finding receipts fail closed.

## Hard boundary

This packet does not authorize a Case 004 dossier, catalog admission, feature, Pages deployment,
provider call, credential use, spend, public launch, or a general claim about communication.
Those remain separate governed tasks.