Reliable machinery. Unfinished science.

• A full blind re-code produced strong intra-rater agreement, and the within-subject result survived both independent codings.
• Where an external answer key existed, the label matched it on all 193 eligible records.
• All 193 admitted packets passed six quality checks; 21 of 215 candidates were rejected upstream for artifact.
• An independent numerical re-derivation matched the pipeline on 194 of 194 source files.
• The admission gates reject candidates under pressure, including candidates favored by the analysis itself.

• No Archetype has yet been externally judged correct by independent expert annotators.
• No Archetype has yet demonstrated cross-corpus generalization.
• The neural features tested so far have not improved phenomenological classification over baseline.
• The experiments, analysis, and reports are not yet independent of the system being evaluated.

The evidence supports the reliability, reproducibility, and selectivity of parts of the pipeline. It does not yet establish that an Archetype is scientifically correct or generalizable.

The original subject-confound rule used one fixed threshold. A corpus-specific null test showed that threshold was too permissive. Under the recalibrated standard, two of four Archetypes failed and were demoted rather than deleted. Their prior status and the reason for reversal remain part of the record.

That correction does not validate the two Archetypes that remain. It demonstrates something different and important: the validation system can reject its own outputs when the evidence or the standard changes.

1. Replace the fixed subject-confound threshold with a corpus-calibrated null.
2. Use independent SAN annotators and an adjudicator to test content distinctions.
3. Pool independent corpora so candidate forms are spread across more people and protocols.
4. Repeat cross-corpus testing with human coding rather than lexical matching.
5. Preserve the registered held-out test until the sample is adequately powered.

Treat current Archetypes as testable, versioned research objects—not settled discoveries. Trace each one to its Exemplars and source Projects, check its evidence state, and cite the exact Edition used. Read the Noetic Taxonomy explanation or the working paper for the broader methodology.