# NUMA Research Protocol (Draft)
Version: 0.1 · October 2026

## Research question
Can a small fixed numerical concept registry and functional grammar improve cross-language comprehension after equivalent training, compared with alternative symbol systems?

## Hypotheses
1. Trained participants recover the intended agent, action, and object more accurately than untrained participants.
2. Parsing and digits-only transmission round-trip without information loss on valid test messages.
3. Polysemy, cultural assumptions, and grammar omissions lead to measurable misunderstanding that must be documented.

## Experimental design (planned, not completed)
1. Recruit adult participants who speak English, French, Spanish, and at least two unrelated language families.
2. Obtain informed consent; avoid collecting unnecessary personal information.
3. Randomize training material and test message order.
4. Present 40 unseen semantic messages with role permutations and negation/tense operators.
5. Ask participants to select the intended meaning from balanced choices; measure accuracy, time, and confidence.
6. Report per-language and overall results, confidence intervals, exclusions, and failures.
7. Publish message corpus, grammar revision, limitations, and machine tests.

## First technical milestones
- Freeze v0.2 proof-of-concept grammar and test corpus.
- Write concept definitions including argument roles and examples.
- Expand pilot registry only after feedback on ambiguous meanings.
- Add explicit versioning and a non-ambiguous large-number format.
- Conduct independent scientific review before claiming universality.

**No field research, human-subject testing, external review, or scientific validation is claimed to have been completed.**