There is no universal dictionary of phase directions.
Across unrelated families, the full twelve-field rank ordering is essentially uncorrelated: -0.007 for expansion and -0.008 for compactification.
That means “field 03 is good, field 08 is bad” is not generally true. Most of a field’s meaning comes from how it meets one organism’s own hidden composition, body geometry, and developmental flow.
But the dictionary is not completely private. Field 06 keeps surfacing as a signed expansion axis.
The same hidden field points toward expansion across families.
HELD-OUT EXPANSION
[+0.027, +0.580]; positive in 11 of 16 families.
HELD-OUT COMPACTIFICATION
[-0.626, +0.085]; the same field tends to suppress the reciprocal direction.
SIGNED POLARITY
[-0.025, +1.181]; positive in 12 of 16 families.
Each held-out family was omitted before field selection. The other fifteen families selected field 06 in every fold. The field was then scored only in the omitted family.
Compactification does not share a winner—it anti-generalizes.
BEST-ELSEWHERE FIELD IN A NEW FAMILY
[-0.553, -0.227]; only 1 of 16 families improve.
WHY THIS MATTERS
For compactification, the field favored by the other families is systematically a poor choice in the omitted family. Its useful directions are body-specific, not merely noisy versions of one universal ranking.
The hidden code is neither a universal program nor arbitrary noise. It is an interaction language: mostly private syntax around a few possible shared axes.
Destroy only direction identity and ask how surprising the transfer is.
The gray ranges come from 20,000 relabelings. Every family keeps its exact twelve outcomes; only the claim that “field 06 here is field 06 there” is destroyed. The expansion effect is at the 92.0th percentile, compactification anti-transfer at the 4.4th, and signed expansion polarity at the 94.6th.
The expansion candidate is compelling but not yet isolated from all chance vocabularies. An independent local-label atlas even produced one accidental transferable-looking compactification label (+0.325). That is why field 06 now goes to new organisms.
Take the word outside the language class where we found it.
Apply field 06 to a genuinely fresh organism cohort, alongside new shared fields and independent local scrambles. Ask only whether it preserves its signed expansion-over-compactification polarity.
If it does, we have a transferable developmental coordinate. Then rotate, reflect, translate, and partially dose that field to learn its geometric meaning. If it does not, the apparent word was a cohort-specific coincidence—and the stronger discovery is individualized developmental memory.
Either way, the next experiment tells us what kind of code this is.