What a score
keeps.

This is a small contract for the current studies. Another work may define different conventions. English supplies names and explanations; those names are not independently recoverable from the shapes or pitches alone.

Entities

Identified things or states in this work. English labels explain their intended referents; identifiers alone do not communicate cultural meaning.

Relations

Named, directed links between entity IDs. A relationship name is a local authored convention, not a built-in ontology.

Temporal behaviours

Explicit clocks, intervals and dependencies available to any rendering. Order in a JSON list is not temporal order unless specified.

Expressive weights

Typed local salience or a declared attention budget. These are not measurements of probability, emotion or understanding.

The machine identifier is holosem-core. The core sits at structure.semanticCore and makes selected operational relationships reusable. visualConfig is also authored score data; it supplies a particular performance’s geometry. Meaning was present in earlier JSON, but some timing was tied to that renderer.

• The declared profile validates shape, ranges, reference integrity and its conservation rules.
• For 001: retain the common origin, simultaneous availability and shared attention budget. For 002: retain each trace’s distinct decay, residual and recurrence timing. For 003: retain a shared reference and independent field cycles.
• The score states intended relationships. Validity is not evidence that another mind decodes them.

• A rendering may choose colour, position, pitch, timbre and interpolation where the mapping is disclosed.
• A reinterpretation that changes a semantic invariant should identify itself as a new score or explicitly changed version.
• Other contributors may use a different documented profile. This small core is a local contract, not a mandatory Holosem grammar.

All three futures share an origin and a total expressive attention of 1. Each retains at least 0.1. These are authored shares, not estimated probabilities or emotional measurements.

Set one target to a clamped value between floor and total minus the other floors. Redistribute the remaining above-floor budget in proportion to the other targets’ previous above-floor amounts (equally if that surplus is zero). Restore each floor.

With Motion On, the same surplus is modulated by a smooth periodic function and normalised back to the budget. Motion Off freezes the state. Adjusting a slider holds time and begins a new cycle from that allocation.

Visual opacity and line weight increase with a branch’s attention. Three fixed sine tones use linear gains from the same values, with a low shared origin tone beneath them. Pitch, colour and stereo placement are artistic choices. Sound adds duration and potential masking; the drawing adds geometry. The projections have not been shown to be perceptually equivalent.

• Catalogue 1.0–1.4 fields and permanent study URLs remain valid. The core is optional for older or differently structured works.
• Studies 001–004 use structure.semanticCore. visualConfig remains public authored data; its legacy timing fields are derived compatibility projections. Study 005 preserves its distinct canonical structure in structure.canonicalScore and at /scores/study-005-canonical.json, with separate rendering choices. It is not converted to holosem-core.
• Human Approximation remains an array of paragraphs; translationNote is the legacy catalogue name for Translation Loss. translationGain is an optional array. Withheld blind-study interpretations use an empty array and explicit interpretationStatus.
• Study 001 revision 1.2 and catalogue schema 1.4 remain readable at versioned URLs. Revision 2.0 adds the shared budget; it does not silently relabel the earlier independent animation.

Airlock packets are transport envelopes. Strings map deliberately to catalogue paragraphs; Translation Loss maps to the retained translationNote field. A review draft has no publication authority. Contact, courier and claimed authorship remain separate review data.

Exact review mapping and worked example / JSON ↗
A valid preparation packet / JSON ↗