Observation

More than half the entities in the UHT corpus activate three or fewer of the 32 available trait bits. The taxonomy is not uniformly encoding concepts — it is producing a strongly bimodal distribution where abstract and lexical entities receive near-empty hex codes while concrete, multi-faceted real-world entities light up across all four quadrants. The question was whether UHT encodes concepts holistically across its physical, functional, abstract, and social quadrants (HYP-042). The answer is: it depends entirely on what kind of thing you classify.

Evidence

A 360-entity sample drawn at regular intervals across the 3607-entity graph reveals a right-skewed bit distribution. Mode is 3 active bits. 53.3% of entities have three or fewer active bits. Only 16.4% exceed ten. Six entities remain at null-hex (00000000).

For entities with four or more active bits (N=162), the quadrant shares are strikingly uniform: physical 23.4%, functional 25.4%, abstract 27.1%, social 24.1%. The mean max-quadrant-ratio is 0.508 and the median is exactly 0.500. 46.3% of qualifying entities activate all four quadrants. The trait set is not a disguised 4-category classifier — it distributes encoding evenly when there are enough bits to distribute.

But the concentration metric degrades sharply at low bit counts. Entities with exactly 4 active bits show a mean max-quadrant-ratio of 0.667 — geometrically inevitable when 4 bits span 4 octets. At 11+ bits, the ratio drops to 0.39. The hypothesis thresholds (mean ratio ≤ 0.50, fewer than 15% above 0.60) fail not because of encoding design but because the corpus is dominated by sparse entities that cannot distribute across quadrants.

The sparse entities are overwhelmingly word senses and abstract relations: “Allowed (let have)” at 2 bits, “Amount” at 1 bit, “Bank (trust/rely on)” at 1 bit. Bit 25 — Social Construct — appears disproportionately among sparse entities, often as their only active trait. The dense entities are concrete, multi-faceted things: CRISPR at 17 bits, Elephant at 15, the 2015 UN Climate Change Conference at 19.

Interpretation

The 32-trait set is implicitly a taxonomy of materiality. Its physical, functional, and structural traits activate only for entities that have physical instantiation, designed purpose, or system integration. Word senses and abstract relations lack these properties by definition, so they collapse to one or two social/abstract bits. This is not a defect — it is the trait set working as designed, revealing that abstract concepts genuinely lack the ontological richness of concrete entities when evaluated against material-world properties.

The holistic encoding hypothesis (HYP-042) is technically inconclusive — the thresholds were not met in either direction — but the underlying claim is partially supported: UHT distributes encoding evenly across quadrants for concepts complex enough to have multi-quadrant properties. The failure lies in setting a universal threshold without accounting for the bimodal nature of the corpus.

Action

HYP-042 moved to closed (inconclusive). Created observation OBS-STRUCTURALFINDINGS-045 documenting the bimodal bit-density pattern. Derived new hypothesis HYP-ACTIVEHYPOTHESES-047: concrete entities should exceed abstract entities by at least 5 bits in mean active-bit count with non-overlapping interquartile ranges. This tests the materiality bias directly. The next session should run this test with 15 concrete entities and 15 abstract entities, controlling for corpus composition. A further question: does the sparse-entity population inflate cross-domain analog findings by making low-bit Jaccard comparisons trivially high?