music/MUSIC_COMPOSITION_DOCTRINE.md
CANON SUBORDINATION — this document is PROPOSAL-TIER: it serves canon and never outranks it.
The canon: the CVD · the T1 foundation docs · the T0 registries (registries/) · the spine
(docs/spine/CH_*.md) · the region pages (_source/02_Tier_2_Region_Pages/). Authority order:docs/DOC_MAP.md§ 0.
Canon served (scanned from this document's own citations — widen it by hand where it is thin):
T0_Chapter_Index·T0_SFX_Registry·T0_Theme_Registry
READ THAT CANON FIRST — open it and derive from it before you build anything from this document.
If this document disagrees with canon, CANON WINS and this document is the defect — fix the
document, never the canon. Nothing here is applied until it is ratified into canon.
The composition lane's runbook. Queued by PIPE_AUDIO_MUSIC section 0 as the artifact that must exist
BEFORE any hero theme is authored. It folds six researched sections (leitmotif architecture,
hummability and the Kondo school, wordless vocals and atmosphere, cultural registers, orchestration
and the concert deliverable, adaptive scoring architecture) and applies a full adversarial critic
pass over them. Where the critic refuted a claim, the corrected claim stands here and the refuted one
does not survive anywhere in the document. Where the critic itself was wrong or two findings
conflicted, the reconciliation is recorded in Appendix F, one line each.
How to read the confidence tags used throughout:
decision rule.
authored from a PENDING-SOURCE row. This tag is a hard gate, not a note.
Two documents this one does not replace: the CVD (which wins every conflict) and the per-region
Section 10 composition briefs (which consume this doctrine, node by node).
PIPE_AUDIO_MUSIC section 0 (2026-07-29, ledger 28al) binds this lane. Its laws, quoted faithfully:
broad — SNES-era, Zelda, Undertale, Sonic, Super Mario 64, Skyrim, Halo, Pokemon, Animal Crossing,
Donkey Kong — scores that live outside their games. They endure on MELODY and leitmotif, not
production sheen.
cannot be hummed after one hearing, it is not the theme yet. Composer-in-loop is the MECHANISM for
this bar (the AIVA closed-as-skip ruling stands); themes iterate until they pass, judged with
evidence like everything else.
period, using instruments, chants, vocals where appropriate but mostly no vocals other than hums
chants and harmonies, ambient noises." Period-appropriate instrumentation per region; vocals
default WORDLESS (hums, chants, harmonies); ambient texture is part of the score.
consecrated performance never is (DESIGN_GAP_REGISTER 1317, 1789; adjudication 18).
VOLUME inside the per-culture register; the identity themes people hum in 2056 are the
composer-in-loop lane's deliverable. The bar never routes down-tier.
operas, and Broadway — the Harry Potter and Star Wars, Distant Worlds, Symphony of the Goddesses,
Symphonic Evolutions lineage. The hero-theme lane's deliverable is a COMPOSED WORK, not just
rendered audio: real notation (MusicXML and MIDI plus orchestration) that a human orchestra can sit
down and play. Rendered stems serve the game build; the score serves the concert hall. Both come
from the same composition.
create this, understand the patterns, and how everything comes together."
Three standing project laws compose with these and are load-bearing here:
is a defect. Rename and reframe, never remove. The 17 hard-line floor is the only absolute.
obligation is therefore fourteen nodes, not eight.
registered. Music is a player-facing surface.
good. Everything below is that sentence made testable.
practice, never by an exotic-signifier scale bolted onto a Western tune.
has already declared what is lawful there.
register cards with named sources and named reviewers. A described claim is not a result.
dependency in this doctrine carries an owner, a trigger milestone, a cost estimate, and a
solo-executable interim proxy that is labelled proxy-tier and never claimed as the real thing.
Wagner never authorised "Leitmotiv" — he wrote Hauptmotiv and Grundthema. The term first appears in
print in F. W. Jähns' 1871 Weber study; Hans von Wolzogen's 1876 Ring guide named the individual
motifs and popularised the practice. Wagner's own framing was dramatic import, not structural
bookkeeping, which is the correct frame here too.
The single operational law the practice hands us: a motif must be identified clearly enough on first
statement that it retains identity when modified in rhythm, harmony, orchestration or accompaniment,
and it must be combinable with other motifs to signal a new dramatic condition. Everything in this
section is that law made buildable.
Scale reference for budgeting: Frank Lehman's Star Wars catalogue documents dozens of entries across
nine films (a derived research corpus counts 64 recurring leitmotifs, Episodes I through IX), yet
audience recognition load is carried by roughly five — Main Title, Force, Imperial March, Yoda, Leia.
Plan for a large catalogue and a small recognisable core. Those are different budgets.
The rules below are induced from verified exemplar anatomy, not asserted.
in trumpets and answers with a leap of a seventh; the harmony alternates tonic with quartal
sonorities; every phrase ends on the dominant, so the tune never closes. Triplets are the rhythmic
backbone at two metric levels, making it a march in all but name.
four two-bar ideas, the second closing on a MAJOR IV (raised 6, melodic-minor borrowing) that reads
as hope and refuses finality; the third idea condenses the first two into an ascent climaxing on
the upper tonic over a major VI. Its non-closure is exploited: Williams frequently deploys only the
first two phrases as a motif.
up a perfect fifth to B-flat — so the tune survives a loud orchestra; the accompaniment pairs i
with a modally darkened flat-VI (E-flat minor), chromatic-mediant territory, Lehman's pantriadic
wonder vocabulary inverted to menace.
comes from two imported chromatic degrees, flat-2 (F natural) and sharp-4 (A sharp), producing a
diminished third and a tritone, and a bar-6 sonority B-D sharp-F-A sharp that sounds like an
altered dominant seventh and resolves nowhere ordinary. Two 16-bar sections, A A B A.
Cross-cut with the empirical work: Jakubowski, Finkel, Stewart and Mullensiefen (2017), Dissecting an
Earworm, matched 100 frequently named involuntary-musical-imagery tunes against 100 controls across
83 melodic features and found faster tempo, MORE common global contour (the rising-falling arch), and
LESS common local gradients between turning points. That is the exemplars' recipe stated as data:
ordinary shape, one strange thing.
THE HEAD-CELL LAW (binding, and scoped). Every canonical theme whose motif_class is "tune" is built
on a 3-5 note head cell that is singable inside a ninth, rhythmically distinct enough to be tapped
without pitch, and either triad-outlining (survives dense mixes) or purely stepwise (survives fragile
solo statements) — chosen deliberately per motif, never both. Exactly one deviant element carries
identity: one unusual leap, one borrowed degree, one metric oddity. Two deviants and it stops being
hummable.
The law is scoped to motif_class = tune because the doctrine's own Tier-A roster contains identities
that are deliberately not tunes. See 1.3.
Every registry row carries motif_class. There are three, and each routes to its own acceptance test,
because a single checklist cannot judge all three.
Aquatic Ambience are the exemplars; the second is a 30-year survivor that is not a one-hearing
hummable tune at all, and its endurance is timbral and harmonic-ambiguous. Takes the ATMOSPHERE BAR
(section 2.4), never the hummability checklist.
something, with no fixed tune. The vril and wonder identity is one of these. Takes the PROGRESSION
RECOGNITION TEST (Appendix T.4), which must be authored and dry-run before the vril theme starts.
A row with motif_class unset cannot be scored and cannot be rendered.
Twenty-two threads do not get twenty-two independent themes. That exceeds any listener's recognition
budget and produces mush.
registry in 1.5. These are the only motifs allowed a full 16-32 bar statement with an A A B A
architecture.
a shared harmonic signature. A thread earns its own Tier-B motif only if it satisfies all three
tests: it recurs across five or more nodes; it has an on-screen agent (a person, place or object
the player can point at); and it has an arc that changes state, because if nothing transforms there
is nothing to score. Threads failing any test INHERIT — they take the family cell with one interval
substitution (change exactly one interval, for example the family's ascending fourth becomes a
fifth) and a distinct carrier voice. Inheritance is the default; independence is earned.
region, boss and faction cells built from a Tier-A or Tier-B cell in the local register. They never
need to be hummable, only coherent with their parent.
Rule: any Tier-C cue whose region also carries a Tier-A payoff must quote the parent's head cell at
least once, or the payoff has no soil to grow in. Per section 6.7, that quotation is a
composer-authored stem laid over generated material or fed to ACE-Step as lego conditioning — never a
prompt, because the generator has no melodic conditioning that could honour it.
The live T0_Theme_Registry carries exactly twelve CARDINAL rows — since the 28au apply the roster
query is row_class = cardinal_theme (composer_in_loop_required is True on all 14 rows, the two
Cassius cue_variants included, because Cassius material is never generated; the count-by-flag
identification below is the PRE-APPLY state, preserved as the reconciliation's input) — which
matches the dossier's "Canonical themes (12)". The research section proposed a roster of seven or
eight described themes against an open range of "8-12", and the two lists did not match: four
described themes had no row, eight live rows were unaddressed, and three of the twelve live rows were
Cassius phase variants that the transformation catalogue itself treats as transformations of one
motif rather than as separate identities.
RECOMMENDATION (proposed to the director; a registry change, so recorded here as a work order rather
than applied): pin the cardinal roster at exactly twelve, as follows.
✅ APPROVED — Josh, 2026-07-29 twenty-second sitting: "yes approve." The reconciliation
below is RULED: the Cassius collapse (CASSIUS_PHASE_1 survives as theme_id; _2/_3 re-class as
cue_variant transformations) + HOME_AND_LOSS and JOURNEY_WORLD minted into the freed Tier-A
slots; the roster pins at exactly twelve. Verbatim:
docs/spine/DECISIONS_PENDING_JOSH.md twenty-second sitting.
✅ APPLIED (ledger 2026-07-29au, same day): author → fresh critic GO-WITH-FIXES (1 MAJOR +
3 MINOR, all applied) → landed. The registry is 14 rows × 28 cols; the roster query is
row_class = cardinal_theme (12); evidence at docs/pipeline_review/motif_roster_apply_2026-07-29/.
pressure: its own notes field records that thematic kinship to GRAND_SAGE_REVEAL makes a pre-reveal
leak high-risk, so the shared material is governed by 1.8.
one cardinal motif with three logged transformations (social, combat, resolution). Retain
CASSIUS_PHASE_1 as the surviving theme_id to avoid foreign-key churn; re-class _2 and _3 as
row_class = cue_variant with parent_theme_id pointing at it. This frees two cardinal slots.
state; it meets all three independence tests outright.
the score.
identity is a harmonic-transformation identity whose cardinal payoff surface is the Vimana
activation. Retitle the row's theme_name accordingly; the anchor stays as canon has it.
closure-change realisation of PROTAGONIST_THEME under its integrity band. They are cardinal because
the closure-change payoff is the game's one authentic cadence and each band needs its own
resolution; they are not three new tunes.
Two described themes are deliberately NOT cardinal. The endgame realm becomes a Tier-B family head,
because the realm design directive builds realms last and a cardinal slot cannot be held open that
long. The timbral voice motif is registered on its own row with row_class = timbral_signature and
motif_class = timbre, outside the twelve tune slots, judged on the atmosphere bar; it is a colour that
threads the score, and forcing it into a cardinal tune slot would put it in front of a checklist it
fails by construction.
STRONGEST OBJECTION to this reconciliation: collapsing the Cassius phases loses the ability to give
each phase a distinct identity in a three-phase final fight, which is exactly where distinct identity
pays. The answer is section 6.5's instrumental-migration model — the phases escalate by carrier and
transformation, which is how God of War's Baldur encounters work and how Ori's boss cues are staged,
and a listener perceives three identities anyway. If playtest evidence contradicts that, the ruling
reverses cheaply because the two freed rows are new material and can move.
Thematic transformation as Liszt and Berlioz formalised it — permutation (transposition, inversion,
retrograde), augmentation, diminution, fragmentation — plus the film and game extensions. Every
operation has a reserved dramatic meaning. An operation used without a dramatic reason is drift.
most legible; ration it. In non-tonal registers the equivalent is the register's own declared
substitution: pathet shift, qenet substitution, or a third-sixth-seventh swap over a fixed drone.
combat-layer form of a theme.
sonorities, modal drones) — wonder, uncanny, revelation. The Hedwig and Lehman lever.
governed by 1.9.
expensive and most earned event: a new dramatic condition.
Budget the operations. A Tier-A motif should appear in six to ten distinct transformations across the
game (INFERRED): fewer and it is wallpaper, more and it becomes noise. Log every one; the
transformation log is the audit surface.
Four mechanisms, all evidenced.
engines: they recur indefinitely without exhausting because they never finish. Reserve the
authentic cadence on the tonic for one node in the whole game. Write the resolution first, then
delete it from every earlier statement. In cyclic non-cadential registers this is satisfied by
construction and the payoff is carried instead by the register's own arrival device (a seleh
landing, a completed cycle stated once in full, the entry of the interlock partner).
heard. The Zelda evidence, scoped correctly: the six player-input CALL songs (Zelda's Lullaby,
Epona's Song, Saria's Song, Sun's Song, Song of Time, Song of Storms) are each a three-note cell
stated twice, while the six WARP songs are through-composed phrases of roughly five to eight notes
(Minuet of Forest six, Serenade of Water five, Nocturne of Shadow seven, Bolero of Fire eight).
The law we take is therefore not a required 3+3 shape: the diegetic surface carries the head cell,
at whatever length the surface allows. Plant every Tier-A cell in at least one diegetic surface —
an instrument minigame, a chant, a puzzle lock, an NPC's whistle — inside the first three nodes it
appears AFTER its reveal_gate. Before the gate, see 1.8.
harmony and intensity across the whole game and returning at Apotheosis in the home key with a new
tempo and pulse. That is the model for PROTAGONIST_THEME across 79 nodes: the Prologue statement
and the Ch-77 statement must be the same tune, and the player must feel the distance.
operations are harmonic revision, double-time tempo, instrumentation matching and rhythmic
adaptation. Reserve motif COMBINATION for relationship reveals — it states the relation with no
dialogue.
Cross-title precedent worth adopting: Soule's Nerevar Rising returns as Dragonborn not as a variation
but as a full orchestral-choral re-arrangement — proof that a tune can be re-scored into an entirely
different civilisational voice and still be the same tune. That is the mechanism for a 79-node world
tour, and it is bounded by 1.9.
Nothing in the research sections encoded reveal discipline, and at least two cardinal themes are
reveal-gated: HOUSE_OF_VELHEIM_THEME (no antagonist-pattern leak before roughly Ch 38) and
GRAND_SAGE_REVEAL (CVD 17.5, no pre-Ch-76 hints), with PROTAGONIST_THEME carrying the kinship risk in
its own notes field. A composer following the plant-diegetically rule literally would plant the
antagonist head cell in its first three nodes and build a 17.5 violation that no other gate here
would catch.
kinship must be inaudible to a fresh listener.
statement and confirms no kinship is perceptible, logged with the capture. This matches the
standing reveal-discipline re-verify.
Re-scoring a motif into a real culture's register is reference-composition — instrumentation, mode or
scale, ensemble shape and era descriptors, authored in that tradition's own terms. It is never
sampling of consecrated performance, and never a fine-tune on one.
inside his own culture's register, with the same craft and beauty as that region's protagonist
cues. What is prohibited is deforming the register itself — systematic darkening, inversion or
menace-coding applied to the regional identity — or making that register the recurring cue for
threat. The House motif carries the darkening operations and travels with it into any region. This
is CVD 17.1 collective protection expressed in music: it protects the culture, and it explicitly
does not make an individual antagonist from that culture unscoreable in his own idiom. A blanket
register ban would produce the stranger result that every villain in the world sounds
Western-orchestral while the local register is reserved for scenery, which reads as timidity to a
player from that region.
region-as-threat check. A motif with an empty deny list has not been reviewed.
Care means weight and authenticity. A generic world-music wash under a cardinal theme is a care
failure, not a taste note.
1. Head cell notated (MusicXML) and contour string recorded; one named deviant, not two. Applies to
motif_class = tune only; timbre and harmonic rows record their own identity spec instead.
2. The recall gate of Appendix T.1 passed at the tier declared for the deliverable (interim proxy or
human panel), with the capture archived.
3. Recognisable at its DECLARED MINIMUM CARRIER UNIT with everything else muted, at 50 percent tempo,
and under the register's declared mode substitution. For interlock-emergent registers the minimum
carrier unit is the interlock group, not one stem — see 1.11.
4. Family membership and inheritance justification recorded; independence claims meet all three
Tier-B criteria.
5. Plant node, diegetic surface, payoff node and closure policy all assigned before first render.
6. Allow and deny cultural-register lists non-empty, with the region-as-threat check recorded.
7. Concert bar: the statement is orchestrable for a live ensemble without the game mix.
8. REVEAL CHECK per 1.8, with the fresh-context critic's capture-backed verdict logged.
Adaptive delivery is vertical remixing (layers whose volumes are driven by game parameters) and
horizontal re-sequencing (loopable sections that transition on cue). Two consequences bind the
composition, and both are corrections to the naive form of the rule.
carries the head cell; that carrier is named in the state-machine row via adaptive_role =
motif_carrier and the field motif_carrier_stem. The mute test is run once per STATE, not once per
cue. This is what makes section 6.5's instrumental-migration escalation legal: migration changes
which stem holds the role and never leaves the role vacant.
the melody exists only in the composite: the Aka method deals a theme's notes across three or four
voices so no performer carries the tune; Shona inherent patterns are melodies nobody plays; kotekan
is polos plus sangsih. A one-stem recognisability test is unpassable there, and enforcing it would
force the composer to rewrite the theme as a single-voice line — the Westernising retrofit arriving
through the stem contract instead of the checklist. So: the registry carries stem_group_id and
atomic_group; for these registers the minimum carrier unit is the interlock pair or trio
(kushaura plus kutsinhira, polos plus sangsih, the hocket group), that group is declared
indivisible, and no runtime parameter may split it. An intensity fader that could mute one half of
a hocket pair destroys the tune rather than thinning it, and the atomic group is what prevents it.
clean seams and no baked fades.
cue, or the engine will fade the one cadence the game has been saving.
It is not a vibe. It is a memory-encoding property with measurable correlates, and the checklist below
is built on three of them, from Jakubowski, Finkel, Stewart and Mullensiefen (2017), Dissecting an
Earworm, which matched 100 frequently named involuntary-musical-imagery tunes against 100 controls
across 83 melodic features.
controls — the ordinary rise-then-fall arch — while having LESS common average gradients between
melodic turning points. The shape of the phrase should be the shape every listener already owns;
the steps between the peaks should not be.
paper's own illustrative cases are the Smoke on the Water riff and the My Sharona and In the Mood
instrumental figures.
Huron's work on Western folksong establishes the melodic arch (convex, inverted-U) as the modal phrase
contour, with descending second, ascending third and concave rarest — a finding about a Western
folksong corpus, and treated as such. It is the default expectation for tonal registers, and each
non-Western register card supplies its own documented contour norm instead.
THE POST-SKIP REVERSAL CORRECTION. Narmour's implication-realization model predicts that listeners
expect a large leap to be followed by a reversal of direction, and as a LISTENER-EXPECTATION finding
that half survives. As a COMPOSITIONAL rule it overgenerates. Von Hippel and Huron (2000), "Why do
skips precede reversals? The effect of tessitura on melodic structure", showed the compositional
pattern is largely an artifact of regression toward the mean of a melody's tessitura: skips tend
toward range extremes, from which a melody has little choice but to retreat. So the doctrine's
scoreable rule is the tessitura rule, not the reversal rule — a leap that lands near a tessitura
extreme reverses; a leap toward the centre need not. Both sources are named on the checklist item so a
composer who disagrees can see why the rule is shaped this way.
Koji Kondo states the design goal in his own words at GDC 2007: he strives to make each piece
recognisable within the first few notes, so the music immediately signals location or status. That is
the hummability test restated as a production requirement — the identity lives in the head of the
phrase, not in its payoff.
textbook Caplin sentence: simple triple metre, binary A B designed to loop, built as basic idea,
basic idea, continuation, cadence. The A section is 16 bars (two sentences), B is 8; the first
sentence's basic idea spans B-D-A in bars 1-4 over an alternating half-quarter-half rhythm, bars 5-8
continue by fragmentation of that same rhythmic template, closing on a half cadence with V7 under
the melodic A. The B section performs liquidation, reducing the motif toward its fundamental
components, and inverts the basic idea's fifth in the continuation's final notes. Kondo emphasises
fourths and fifths and withholds tonic resolution so the loop never lands finally. The prototypical
sentence proportion is 2 + 2 + 4 bars inside an 8-bar theme.
harmonic rhythm of exactly one chord per bar, perfectly regular, while the chords themselves are
chromatic: parallel-minor borrowings (minor v, flat-VI), a non-cadential tonicization to D-flat
rather than a real modulation, and a return through V-of-V to V. This is the school's central trick.
Rhythmic and metric regularity buys harmonic adventurousness: the listener can hum the tune because
the grid never moves, and the tune stays interesting because the harmony under it does.
with swing feel and prominent syncopation over a Latin rhythm; its first three bars are the
recognisable unit. Kondo cites T-Square's fusion as the influence because the rhythms were easy for
listeners to follow. Ease-of-following is the criterion, not simplicity.
average on chord complexity (sevenths, add9s), borrowed non-diatonic tones and progression novelty
while the top line stays singable — the same regular-grid, rich-harmony split. Sonic's Green Hill
Zone is C major with an A section, a key-shifting B section, and a deliberately contrasting
counter-melody under the main line.
theme accrues meaning by recurrence rather than by length. Kondo's Ocarina of Time Hyrule Field
music is the systems analogue: 12 phrases randomly stitched into 8-phrase sequences, so the loop
varies without the identity moving. That is the model for our own field music, and it is cheaper
than generating variation volume.
inside one hymn-length phrase. Totaka's stated Animal Crossing principle: no sound that was
unnecessary.
one-hearing hummable tune. Its endurance is timbral and harmonic-ambiguous, built on wave-sequenced
single-cycle samples and weeks of sample engineering. The checklist below is never applied to
ambience beds or to motif_class = timbre. Two lanes, two bars.
Scope: a candidate hero theme with motif_class = tune, scored on its PRIMARY STATEMENT only (the first
8 to 16 bars), already written inside its correct register. The checklist is never a Westernising
retrofit of a correctly registered theme, and the register-relative construction below is what makes
that promise enforceable rather than aspirational.
Several classic checklist items presuppose Western functional tonality and Western phrase norms. A
theme written correctly inside this doctrine's own registers fails them by construction: slendro and
pelog have no scale degrees 1-3-5 and no functional harmony, and seleh pitches are chosen by pathet;
Hijaz has a structural flat-2; the corrected ambassel has structural flat-2 and flat-6; a 48-pulse
mbira cycle has no cadence to withhold; a West African timeline has no shared downbeat, so regular
hypermetre mis-scores it; several registers have no chords at all; and the yodel register-breaks
required in the San, Aka and Shona huro registers deliberately violate a centred-tessitura rule. With
a no-zeros gate, an unscoped checklist would systematically reject correctly-registered themes and
reward Westernised ones — enforcing exactly the retrofit the care line bans.
So EVERY REGISTER CARD SUPPLIES ITS OWN SUBSTITUTION SET, and a diagnostic is scored only against that
set. The substitutions each card must declare:
swap over the fixed Sa and Pa (Sinhalese and South Indian), or an explicit NONE.
registers; seleh cadence pitches; the reciting tone; the mbira harmonic-segment landing.
construction, scored 2).
An item with no equivalent in a register is scored N/A and DROPPED FROM THE DENOMINATOR. The gate is a
percentage, not a fixed 28 out of 34.
can fail a theme on listener evidence alone.
credit. The four: mode flip or its declared register equivalent; augmentation and diminution;
re-instrumentation into another cultural register drawn from the motif's allow list; and
fragment-as-stinger at two seconds or under.
CLAUSE: the cardinal set is drafted as a COHORT, B3 is evaluated at cohort close against the
declared Tier-A and Tier-B family cells, and the first theme authored satisfies it against its
family head cell rather than against a finished sibling. Without this clause the first cardinal
theme is unpassable by construction, because no other canon leitmotif exists yet.
WRITTEN range (section 5.2), and singable by a wordless choir at that section's own tessitura after
transposition.
Applies to motif_class = tune only; timbre rows are judged on the atmosphere bar instead, where
timbral dependence is the point.
first three to five notes on a second hearing. Kondo's stated bar.
or matches the register's declared contour norm. Concave shapes are the rarest and least sticky:
justify or reject.
larger, or an insistent repeated-note figure. Two or more anomalies is not a hook, only difficulty.
Threshold INFERRED.
centre need not. Sourced to Von Hippel and Huron 2000 beside Narmour.
within the middle third of that span. Absolute pitch is NOT part of this gate — it belongs to the
per-deliverable transposition note: sing-back tests run at a transposition comfortable for the test
panel, concert statements are transposed to the performing section's own tessitura, and the
per-region tonal_centre rule of section 3.6 governs the bed. The congregational-singing percentage
quoted in the research section is dropped for want of a source.
degrees; chromatics or non-anchor pitches are passing rather than structural, unless the register
declares them structural (Hijaz flat-2, ambassel flat-2 and flat-6, anchihoye flat-2).
rhythm to the right theme out of three.
timeline or colotomic registers, correct alignment to the declared cycle instead. Odd groupings must
be a deliberate character marker, not an accident.
not in WHEN. N/A where the register has neither chords nor colotomy.
compensating hook elsewhere. Threshold INFERRED.
smaller and accelerated — or by the register's own documented variation procedure (irama density
change, kutsinhira entry, hocket voice count).
finished.
GATE. All five binary gates PASS, and the diagnostics reach 77 percent or more of the APPLICABLE
maximum with no applicable diagnostic scoring 0. On a full thirteen-item register that is 20 out of
26. Below the gate the theme returns to the composer-in-loop lane with the failed items named. Every
iteration's scores are logged as the evidence trail.
The percentage threshold and the 0-1-2 anchors are INFERRED calibration, tunable once the first ten
hero themes have produced real score distributions. They are not findings and are never cited as such.
Ambience and timbral identities are scored on a different bar entirely, because the Aquatic Ambience
counter-case proves the hummability checklist would reject a 30-year survivor.
same pitch material. If the colour is not the identity, there is no timbral motif here.
melodic statement, judged on captured playback.
Hummability techniques are structural: sentence form, contour, the tessitura rule, regular event rate.
Cultural identity comes from instrumentation, tuning, timeline and performance practice as specified
in the register card — never from bolting a pentatonic scale plus parallel fourths onto a Western
tune. That specific combination is the documented Orientalist cliche.
Two facts, kept apart, because collapsing them is itself the error the rule warns against: the
so-called "oriental riff" is a Western stage invention traceable to the Aladdin Quick Step from the
1847 Grand Chinese Spectacle of Aladdin, spread through Tin Pan Alley and 1930s cartoons; separately,
the 1889 Paris Exposition Universelle is where Western composers, Debussy among them, first heard
JAVANESE gamelan. Citing a Southeast Asian encounter as the origin of an East-Asian-signifying cliche
reproduces the interchangeable-foreignness move the rule exists to ban, and it undercuts the rule's
authority with anyone who knows the history.
Two rules follow. Never use pentatonicism as a shorthand for foreignness; a culture's mode is used
because that culture's own repertoire uses it, cited on the register card. And the checklist is
applied to a theme already written inside the correct register, which section 2.3.1 makes structurally
true rather than merely promised.
Wordless voice is the only orchestral colour that is unmistakably human and carries zero semantic
freight. It localises for free, it never dates, and it is fully notatable — which matters because the
hero-theme deliverable is a concert-performable score, not a render. The concert tradition is deep and
citable: Debussy's Sirenes closes on a wordless female chorus seated within the orchestra; Holst's
Neptune scores a six-part double chorus of wordless women's voices placed in an adjoining room whose
door is closed slowly during the last bar, the earliest great fade-out in orchestral literature,
achieved by architecture rather than a mixer; Rachmaninoff's Vocalise lets the singer choose the
vowel. A wordless chorus is a first-class notated object. Write it as one.
Exemplar readings, exact:
4/4 at 95 bpm; the middle section shifts to 12/8 at 120 bpm before the chant returns at 95. The
monks were not a choir — O'Donnell, Salvatori and three jingle singers, overdubbed and reverbed. Two
lessons: the identity lives in a short modal melody with a tight ambitus, and five good voices
multitracked beat a thin real choir. The chant's internal phrase structure is CONTESTED in our own
research (one reading gives 28 notes in four phrases with the second and fourth of seven notes each;
another gives 28 as seven times four with inner phrases of seven), so no composer copies the shape
until a transcription check pins one description with a citation. What survives either reading and
is safe to build on: four phrases, a tight ambitus, a modal final, roughly 95 bpm. Care note: the
wailing solo over the strings is qawwali-flavoured, and qawwali is living Sufi devotional practice —
we take the register, free melisma over a modal drone, never a qawwali text or a sampled
performance.
over taiko-derived percussion and strings, singing in Dragon-tongue, a constructed language. The
structural insight: to a player an invented tongue is functionally wordless, while giving singers
hard consonants for rhythm and attack that pure vowels cannot supply.
from their sound. Same lever, same hazard.
first, treated the Japanese pentatonic as a liberating constraint, and recorded taiko in Japan. He
also hit our exact failure mode — his invented phonetic syllables read as vulgar to the Japanese
taiko performers in the room. That is a binding QA rule for us, not an anecdote.
Homeric Greek, French and Japanese. We do not do this. In play, Journey's identity is carried by
cello and a wordless call — that is the part we take.
Vowel choice is an acoustic decision. Every vowel is a formant pattern, and matching vowels across a
section tunes the singers' vocal-tract resonances so they reinforce. Front vowels raise the second
formant (ee is bright, cutting, forward). Back closed vowels lower both (oo is dark, distant,
covered). Open ah raises the first formant and is loud, present and maximally projecting.
floor of any vocal writing; it sits under dialogue without ducking.
tutti.
as pitch rises, and the defect is the effect.
on a specific beat.
Vowels cannot be held fixed across the range: the louder or higher a vowel is sung, the more it
migrates. Never write a sustained high soprano ee and expect blend; write ah or aw up top and let the
darker vowels live below the passaggio.
Working choral ranges that hold blend and endurance, at SOUNDING pitch: soprano C4 to G5 (write G4 to
E5 for tessitura), alto G3 to D5, tenor C3 to G4 (sits E3 to E4), bass E2 to C4 (sits G2 to C4). Tenor
parts are NOTATED AN OCTAVE ABOVE SOUNDING in the standard treble-clef-with-8 convention, so a tenor
line written C4 to G5 sounds C3 to G4; a doctrine that states choral ranges without saying which
convention it is using produces parts an octave wrong. Tessitura governs, not the extremes.
Doubling is the identity switch. A choir doubled by the full string body becomes texture; a choir
alone, or doubled at unison or octave by a single sustaining colour (cor anglais, solo viola, one
horn), keeps the tune and reads as identity. Exposure is identity; doubling is atmosphere. Divisi
thins the sound and makes dissonance legible rather than muddy — use divisi for ambient and harmonic
layers, unison and octave for statements of a theme. Follow the Halo lesson on forces: a small
ensemble multitracked in several passes is the affordable path to a hundred-voice impression, and it
is what generated beds are layered AROUND, never a substitute for.
Steal the grammar, invent the melody. The grammar is generic and unownable; the melody is ours.
accent. Write the tune before any harmony exists.
modes ONLY IN UNTRANSPOSED NOTATION; a mode is defined by its interval pattern above its final and
may be transposed to any pitch. Halo is Dorian transposed to a final of E, with the characteristic
sixth as C sharp — it is not Phrygian, which is what a literal reading of "E identifies Phrygian"
would produce, inverting the affect the section is trying to specify. Registry consequence: a row
records mode_final AND mode_interval_pattern as a pair, never a bare final.
tone (sustained, most syllables), then mediant (mid-phrase inflection), then termination (returns to
the final). Invent inside that four-part shape and it reads liturgical instantly.
about a seventh; a full octave span is rare.
are ordinary; the minor sixth is rare; the octave is rare but permitted; the melodic seventh and the
tritone are not used. Reading the ambitus rule as "prefer a seventh leap to an octave leap" inverts
chant practice exactly, and since deliberate violation of these rules is our corruption and Hollow
signal, a misreading here would plant a Hollow-flagged interval in ordinary sacred material.
four), then melismatic (five, six, sixty notes on one sustained vowel). Wordless writing gets this
for free: melisma is vocalise.
set out in Musica enchiriadis, around 895), vox principalis carrying the tune against a vox
organalis, with oblique motion — one voice holding while the other moves — used to dodge the
tritone. For the vast cathedral setting, Notre-Dame melismatic organum: a held tenor under a florid
upper line.
the echos, ranging from fixed-for-the-whole-piece to phrase-mobile, but never as mobile as the
melody, and dropping to stay below the melody before returning. Take the behaviour. The ison is
living Orthodox liturgical practice; we never transcribe one.
The research sections applied this test asymmetrically: Gregorian chant and the Byzantine ison were
treated as free public grammar to invent within, while Ethiopian Orthodox zema — an equally
documented, equally public liturgical tradition — was barred even from original composition in the
idiom, leaving Europe's cathedral with a composed vocal identity and Lalibela with reverb. That is the
colonial register arriving through timidity, and it is the care doctrine's own named failure mode:
care-tier inflation is a defect, and the tool is rename and reframe, never removal. It also contradicts
the sourcing law's own test, which permits original material referencing the surrounding idiom.
The doctrine adopts ONE rule across all traditions.
behaviour, drone behaviour, timbre class, tuning system, timeline.
pieces, Solesmes propers, named bira pieces, odu verses, named healing songs, royal drum
appellations); real liturgical or scriptural text in any language; and any sampled or
model-conditioned recording of consecrated performance.
hidden from the uninitiated — that is a NAMED PER-REGISTER EXCEPTION recorded on the card with its
reason, never a silent default and never generalised from one form to a whole tradition.
The consequence for section 4 is concrete: every tradition in the slice gets a lawful composed vocal
lane, or its card states in writing which named exception removed it and why.
The research section flattened four different risk classes into one prohibition and mislabelled one of
them. They are separated here, and each line carries the rule it comes from.
CARE-EXCLUDED (never, in any lane, on care grounds):
melody still sung in a living liturgy is a care decision, not a copyright one. The 1961 Liber
Usualis is public domain and Solesmes' editorial rhythmic signs still carry editorial rights, and
neither fact settles anything that matters here.
OUT OF SCOPE BY THE WORDLESS RULING (a style decision, revisitable by Josh for the concert lane only,
and not an ethics question):
Western concert-choral repertoire and of the video-game-concert lineage this doctrine aims at.
Excluding them follows from the wordless default, not from appropriation. Filing them under care
would both misteach the distinction and block a texted concert movement as though it were an ethics
problem.
PROCESS RULES for invented material:
BEFORE any recording session. This is the Eshkeri failure, and it fails in the room in front of the
performers we hired.
never as pseudo-anything-real. Faux-Latinate mumble is banned.
Untuned ambience is SFX; tuned ambience is score. The dividing line is a declared pitch. Every region's
ambience carries a tonal_centre equal to its theme's modal final (or its fourth or fifth, the standard
transposition move for widening a pitched bed). Beds without a declared pitch belong in
T0_SFX_Registry; beds with one belong in T0_Theme_Registry as row_class = tuned_bed and are scored.
Bed stack, three layers, three homes:
register, unattended.
bocca chiusa. Score. This is the ison-behaviour layer, and per section 6.7 it is generation-eligible
only where the register card declares a 12-TET-compatible tuning.
(roughly every 20 to 40 seconds). This is what makes a bed feel composed instead of generated, and
it pre-loads the hummability test before the player hears the full theme. It is ALWAYS a
composer-authored stem: motif_source = authored, laid over the generated layers, never a prompt.
Three handoff mechanisms, all implementable as vertical layering plus Quartz quantization, with
horizontal re-sequencing as the fallback when bed and theme disagree on mode:
No crossfade, no seam, no duplicated bass. Requires tonal_centre equal to the theme's final.
different scale degree — final, then reciting tone, then third. The note does not move; the meaning
does. The cheapest emotional turn in the system.
low-pizzicato event quantized to the Quartz bar boundary one or two bars BEFORE the melody, so the
melody never sounds like it entered late.
Both the drone pivot and common-tone re-reading assume shared temperament, so neither may be used
across a tuning-system boundary. Section 4.12 makes that a rule with teeth.
The reverse handoff is the half everyone skips. Do not fade the theme out. Subtract layers in a fixed
order — melody, then counterline, then inner harmony, then percussion, with the drone remaining — and
let the surviving drone be the bed's pitch centre. Same object, thinned; the player never hears a cut.
LOOP-SEAM LAW FOR VOICES. Sustained wordless choir betrays a loop seam worse than any other layer,
because the ear tracks one continuous vowel. Never write a tutti breath: stagger the breathing across
divisi so the vowel never fully stops, and place the loop point at a hum onset or consonant, never
mid-vowel. No baked fades, already binding in the stem contract — a fade destroys the Quartz boundary.
Fields this section owns in the consolidated schema (Appendix S): vocal_register, vowel_ipa,
choir_forces, tonal_centre, mode_final, mode_interval_pattern, reciting_tone, text_status,
phonetic_screen_by, and the per_stem_role split of bed_untuned against bed_tuned.
Tests: Appendix T.1 (recall, sung on one vowel for wordless material), T.3 (drums-muted culture test,
loop-seam test, handoff test), and the phonetic screen recorded in the row before any session is
booked. Where a test in this section previously stated its own listener count or delay, that variant
is deleted in favour of the single protocol in T.1.
THE SOURCING LAW, restated as a per-cue test. Reference-COMPOSITION to a culture's general musical
language is permitted; sampling or reproduction of consecrated performance never is. Ask of the model
being used: is it a liturgy, an initiation, a healing rite, an ancestor-calling, or a divination? If
yes, compose original material referencing the surrounding idiom instead. If no, compose freely in the
idiom. The canonical failure is Deep Forest, which used field recordings of Central African polyphony
without consent and mis-credited a Solomon Islands lullaby as Pygmy music.
THE GRAMMAR-AND-REPERTOIRE SPLIT (section 3.4) is what makes the law executable rather than
paralysing. A card that names a palette and then excludes that same palette four lines later gives a
composer nothing to act on, and the likely reading — do not touch this culture's music at all — is
omission dressed as care. So every card below states the FREE grammar and the EXCLUDED repertoire as
two separate lines that do not cancel each other.
EVERY CUE CARRIES A REGISTER CARD with these fields, and a cue whose card is incomplete is not
authorable:
The central quality claim — the register must be recognisable to someone from that culture — cannot be
discharged by an AI critic listening to a render. The drums-muted test and the register card catch
structural defects; they do not catch a wrong embat, a tala felt as a grid, a gamaka-less straight
tone, or an idiom that reads as parody. Only a tradition-bearer hears those.
been read by a named tradition-bearer or specialist ethnomusicologist for that tradition, recorded
in register_reviewer and register_review_date, with the reviewer's objections recorded VERBATIM in
the card, including any that were overruled and why. Where section 5.8 already books a
tradition-bearer soloist, that person is the natural reviewer and the consult is paid work, not a
favour. Owner: the director. Trigger: at the same time the soloist session is booked. Cost: one
consultation fee per register, budgeted with the session.
specialist-literature verification pass by a fresh-context research critic that pins every card row
to a named source and kills every PENDING-SOURCE tag. This unblocks authoring; it does not unblock
release.
The ruled slice is PROLOGUE through CH 13 — fourteen nodes. The research section supplied executable
registers for eight and named CH_02/03 only negatively, which left the first two chapters the player
ever hears, and the first two chapters built, with no music doctrine of any kind. This section covers
all fourteen. Cards vary in depth: those carried from the research pass are deeper than those
authored here, and the newly authored ones carry more PENDING-SOURCE rows. The interim-tier pass of
4.0.1 is what levels them, and no cue is authored from a PENDING-SOURCE row.
Three slice chapters pin a pre-cultural or deep-time stratum: CH_02 (Homo floresiensis, roughly 60 to
100 thousand years before present), CH_10 (Olduvai, Oldowan and Acheulean horizons), CH_13 (Blombos).
NO LIVING PEOPLE'S IDIOM MAY SCORE A DEEP-TIME STRATUM. Scoring the Paleolithic with a present-day
Indigenous register is the "primitive" coding this doctrine exists to prevent, and it is also simply
false. Deep-time strata are scored with harmonic-series material, breath and struck-stone timbres,
untuned environmental layers, and invented non-idiomatic material; the card records the stratum as
CULTURE-NONE with that instruction, and the adjacent living register is explicitly barred from it.
after all mythological realms, on the matured concept-art pipeline; the slice is built Ch 2 to 13
first and the Prologue and Ch 1 close it. Their register cards are authored at that point, from the
realm program's own art and physics decisions.
name register, never cozy-domestic English, no phrase-names. The eleventh-realm precedent (invent
our own otherworld rather than reproduce a real tradition's) governs the musical language too: this
register is INVENTED, and inventing it is the point.
in Ch 77, and section 1.7 requires them to be the same tune. The theme is therefore composed EARLY,
in the slice, in a register that the fairy-realm card will later be built to fit — not the other way
round. Record this dependency in the theme row.
pre-colonial Flores roughly 1000 to 1500 CE; CH_03 pre-colonial Manggarai roughly 1000 to 1500 CE
through Portuguese 1500s and Dutch colonial 1800s to 1900s. Sub-regions: Liang Bua, Kelimutu,
Inerie, Labuan Bajo (CH_02); Ruteng, Wae Rebo, Compang Ruteng, Todo (CH_03).
here is a category error of the same class as scoring Wales with a sitar. The registers are separate
gong-and-drum traditions with their own local tuning.
measured cents for whichever gong set the score models, or declare the set invented-but-consistent
and pin its own cents table. tuning_system: LOCAL-GONG-SET, cents_table_ref required.
(gendang) leads and cues, the gongs mark the cycle. State the cycle length and stroke positions on
the card with the indexing convention explicit, as in 4.10.
performed for onlookers); communal circle song and dance forms (danding and sanda in Manggarai,
mbata with struck bamboo or plank); bamboo flutes and jaw harp; Ngada and Nagekeo bamboo ensembles
and the paired flute class. All named forms are PENDING-SOURCE until the interim pass pins them.
circle-song texture (unison and heterophonic response with clap and stamp), bamboo timbres.
ancestors' destination) and its offering rites; the Manggarai penti thanksgiving ceremony and its
ritual repertoire; the Ngada reba ceremony; and the ceremonial speech-verse classes (torok and
go'et) which are owned spoken art, not public melody — no fabricated invocation, in any language.
cave resonance, breath tone and harmonic-series material. It is explicitly barred from using
Manggarai, Ngada or Lio idiom, and the living register is explicitly barred from being read as
archaic.
cycle-marking pitches. Periodicity: the gong cycle. Harmonic rate: colotomic period. Non-closure:
cyclic non-closure. Contour norm: PENDING-SOURCE.
every pitched layer is authored or sampled at real tuning.
authored and reviewed FIRST, ahead of every other register in the slice.
across the chapter's time-fluctuation windows. Sub-regions include Besakih, Gunung Agung, Gunung
Batur, Tirta Empul, Uluwatu, Goa Gajah, Tanah Lot, Gunung Kawi.
pelog is seven unequal pitches with five chosen per piece. No prescribed tuning exists — each
ensemble owns its laras and its root. Build two cents tables, Java and Bali, deliberately different.
A 12-TET gamelan is an instant tell. Bali adds paired detuning, ngumbang and ngisep a few hertz
apart, for the ombak shimmer.
halves those, ketuk and kempyang subdivide again. State colotomy per form as beats per gongan plus
kenong and kempul placement, in balungan beats, naming the keteg conversion: lancaran 16 keteg and
8 balungan beats with kenong every 2; ketawang 16 balungan beats with kenong at 8 and 16; ladrang 32
balungan beats with kenong every 8. GATRA is defined separately: four balungan beats, the basic
melodic unit. Gatra and colotomy are independent, and the kenong span equals one gatra only in
lancaran.
dadi 1 to 4, wiled 1 to 8, rangkep 1 to 16 (elaborating notes per balungan beat). Map explore,
tension and combat onto tanggung, dadi and wiled so all three rungs exist and the ladder keeps its
bottom rung. Intensity rises with no key change and no new tune, which is exactly what section 6.5
wants.
2-3-5-6-7) selects the five pitches and governs seleh cadence pitches. Wayang shifts pathet across
the night's three watches, so we shift by dramatic stage. Texture is stratified polyphony — one
melody at many simultaneous densities — with no functional harmony and no V-I. Kotekan interlock
(polos and sangsih) is an ATOMIC STEM GROUP per section 1.11.
north Bali — 700 years wrong. The same rule disqualifies two ensembles the research pass listed as
survivors: gong gede and semar pegulingan are later court ensembles of the post-Majapahit Gelgel and
Klungkung era, roughly 16th century and after, and are POST-STRATUM — not for the 1200 CE pin. They
become lawful only inside a declared time-fluctuation window that the chapter's own era row opens,
and the cue must be marked as such.
Aga and pre-court strata, used as INSTRUMENT CLASSES with original material; plus suling (flute) and
rebab colour, each flagged for its own period check.
ensemble.
corpus; staging a consecrated ceremony as spectacle; and looting or trivialising a consecrated
instrument (the instruments receive offerings and are consecrated objects — no lootable gamelan
prop). The class is free, the repertoire is not, and no ceremony is a set-piece.
gongan. Harmonic rate: colotomic period. Non-closure: cyclic non-closure, satisfied by construction.
Contour norm: balungan motion.
authored or sampled at real tuning, and only the untuned noise floor is generative.
The live registry runs CH_05 well beyond Java: Sumatra (Palembang, Toba, Pagaruyung, Bukittinggi),
Java (Borobudur, Prambanan, Merapi, Candi Sukuh), and a mainland excursion through the Malacca Strait
and Kra Isthmus to Angkor and the Tonle Sap, with a Champa touch at My Son. Those are three registers,
not one, and each gets its own card.
composite.
Balinese one, and using one table for both is the tell.
and more sustained; the rebab, suling, gender and gambang carry elaboration; the pesindhen and
gerong vocal lines are part of the texture (scored wordless per the ruling).
context); transmitted gendhing repertoire by name; wayang ritual functions as distinct from wayang
as public theatre — the theatre frame is free, the ruwatan exorcistic function is not.
and the Bayon, the 1177 Cham naval sack and the reconquest.
— arched harps, gong-chimes, conch, drums, cymbals. Score from the depicted instrumentarium rather
than from the modern ensemble, and the period claim becomes checkable against stone.
pinpeat cue rendered in equal temperament is the same class of tell as a 12-TET gamelan.
and thom (gong circles), sampho and skor thom (drums), chhing (small cymbals). THE CHHING MARKS THE
CYCLE and is the timeline instrument: composed parts are written in relation to its closed and open
strokes. Mahori is the softer string-led entertainment ensemble and is the freer lane for
non-ceremonial scenes.
mahori idiom.
ceremonial repertoire including the buong suong invocation and the sampeah kru teacher-homage pieces;
pinpeat in its temple-ceremonial function as distinct from its theatrical one.
devastated under the Khmer Rouge and their reconstruction is living cultural-survival work carried
by named surviving masters. Pastiche here is not a taste failure. Score it with the craft the
reconstruction deserves, and credit the tradition-bearer soloist per 5.8.
small, sharply characterised ensemble whose drum pair drives everything.
procession idiom.
(Bani) and Balamon liturgical repertoire; anything led by the initiated ong gru and ong ka-ing. The
Cham are a living minority under real pressure and their ceremonial music is restricted practice, so
this exclusion is a named grammar exception under 3.4, not a blanket ban: the secular procession and
ensemble idiom stays free.
Anuradhapura, Sigiriya, Sri Pada, Mahavihara and Abhayagiri, Vedda forest, Fa Hien Cave.
mid-17th century, is where the 72 melakarta parent scales appear. Do not score Anuradhapura as a
Carnatic concert (kriti, violin, concert tani). Use raga as modal COLOUR over a sruti drone plus
Sinhala ritual-adjacent drumming.
changes by swapping the third, sixth and seventh over the SAME pedal fifth. That is the orchestral
trick and it is also this register's mode-flip substitution.
Straight tone over a drone reads as generic New Age, which is the failure mode.
a grid.
speaking drum; thalampata cymbals holding tala; horanewa double-reed shawm; hakgediya conch for
procession.
tovil healing rites; Pali paritta and pirith chant — never fabricate a pirith line.
should be replaced with wordless environmental texture. That is deleted. Converting a living
Indigenous people into ambience is a recognised colonial move — scoring a people out of the frame
while keeping their forest — and endangerment is an argument for accuracy and consultation, not for
substituting birdsong. The lawful lane, in order of preference: compose original material in the
documented secular Vedda vocal idiom (short cyclic phrases, narrow ambitus, unison and heterophonic
response, hand-clap and body pulse), cited to published ethnomusicology on the card and screened by
a community reviewer under 4.0.1; or, if no lawful lane survives that check, represent Vedda presence
through named characters, instruments and diegetic performance rather than through pitchless
texture. The choice and its reason are recorded in the card either way. Excluded in both cases:
ceremonial and invocation repertoire.
Pa plus the raga's declared nyasa. Periodicity: the tala avartana. Harmonic rate: N/A (drone, no
chords). Non-closure: return to sam without final cadence. Contour norm: gamaka-shaped approach.
produce it.
Brihadeeswarar at Thanjavur, Palani, Golconda, the Western Ghats, and the Kerala Kalaripayattu home
territory.
kriti repertoire is 18th to 19th century. Neither belongs at a Sangam or medieval-Siddha pin. Work
instead from the PANN modes, the older Tamil melodic system documented through the Tevaram corpus,
and from Sangam-era instrument evidence.
through reconstruction — an exceptional colour and a real period tell; the kuzhal reed pipe; the
murasu and parai frame drums; the udukkai hourglass drum; later-stratum periya melam (nadaswaram
with tavil) and the Kerala chenda, edakka and panchavadyam ensembles for festival scenes inside
their own window.
The ottu (drone shawm) is the period-honest drone beside the nadaswaram.
public festival melam idiom, martial-training percussion idiom.
initiatory practice; Tevaram and Divya Prabandham as sung by odhuvar in temple service; sopana
sangeetham in its shrine function; the Sabarimala and Ayyappa devotional repertoire; and the kalari
ground's own invocatory practice, since the training space is consecrated and the gurukkal lineage
owns its rites. Festival melam performed publicly is free as original composition; the same
percussion inside a ritual function is not.
stigmatised through that association, and is now the centre of a reclamation movement. It is scored
with weight and virtuosity, never as a generic or archaic drum, and never as the sound of poverty.
Represent it as the technically demanding instrument it is.
form is PENDING-SOURCE until the interim-tier pass pins it.
Sub-regions: Kilwa Kisiwani, Mombasa, Zanzibar Stone Town, the Mijikenda kaya hinterland.
1870 to 1888), who imported Egyptian musicians and instruments (oud, qanun, violin, riqq), and
Swahili-language taarab only took hold in the 1920s. On a 1200 to 1400 chapter that is a 500-year
anachronism. Taarab belongs to a modern Stone Town stratum or the CH_59 revisit.
raft-rattle, whose hiss doubles as coastal ambience; the zumari or nzumari double-reed shawm; the
side-blown siwa ceremonial horn as an instrument CLASS (the 2.15 metre Pate ivory siwa is
17th-century, so it is a class reference, not a dated prop); the early Indian Ocean lute family;
clap and response.
and plucked strings, with no Western triads; heterophony, one tune with many simultaneous
ornamentations; a fixed drum timeline under call-and-response where the leader OVERLAPS the chorus
entry.
the taarab anachronism ruling with a blanket ban on the entire Islamic vocal complex, which leaves a
literate, mercantile, Arabic- and Persian-connected civilisation with a drums-and-rattle register.
East Africa equals drums is the canonical colonial flattening, and arriving at it through
over-caution is still arriving at it. The lump was also miscalibrated: adhan and Qur'anic recitation
are genuinely consecrated recitation, while qasida and maulidi are publicly performed, commercially
recorded devotional-festival genres (the Lamu Maulidi, Zanzibar maulidi ya homu) that the sourcing
law's own test routes to original material referencing the surrounding idiom. So the exclusion
splits.
vocal register — solo plus chorus response, melisma over a Hijaz-type tetrachord, a duff-class
frame-drum pulse — with era-correct instrumentation. A Kilwa-court and mercantile vocal cue is
added to the CH_08 palette explicitly, so the chapter's identity is not carried by percussion alone.
The drums-muted culture test (Appendix T.3) is what proves it, and the card says so.
score instrumentally and environmentally only, from outside.
madh and qasida vocal register as original composition.
tetrachord's tonic and fourth. Periodicity: the drum timeline. Harmonic rate: N/A. Non-closure:
overlapping call-and-response entry. Contour norm: descending melismatic phrase over a fixed pulse.
authored, because maqam-class intonation is not 12-TET.
12th to 13th century. Sub-regions: Aksum, Lalibela, Simien, Ras Dashen, Saint Mary of Zion, Yeha.
wide. The rows: tizita major C D E G A (0-2-4-7-9); tizita minor C D E-flat G A (0-2-3-7-9); bati
major C D F G A (0-2-5-7-9); bati minor C D-flat F G A (0-1-5-7-9); AMBASSEL C D-flat F G A-flat
(0-1-5-7-8) — a pentatonic subset of Phrygian, CORRECTED from the research pass's 0-2-3-7-8, which
was a different mode entirely and would have had a composer write the wrong mode for a whole
chapter; anchihoye C D-flat F G-flat A (0-1-5-6-9). Affect: tizita is nostalgia; bati and ambassel
carry lively dance tunes; anchihoye is dramatic and solemn. THE WHOLE TABLE IS PENDING-SOURCE: it
ships with no citation, sources genuinely disagree because the practice is unequal-tempered, and no
CH_09 cue is authored until every row is pinned to one named source class (Weisser and Falceto's
qenet study, or Shelemay's New Grove entry) and the citation is recorded per row.
to a plain major pentatonic and reads Scottish or Chinese if played wrong.
VAULT the gap instead of stepping around it.
strings ring continuously — melody floats over a static cloud, never over changing chords.
speech-shaped bowing. That portamento identifies the music more than the scale does.
minor, used as leaning tones. (Restated from the corrected ambassel; the original phrasing credited
ambassel with a flat-6 over a natural 5 while its listed pitches contained neither a flat-2 nor the
characteristic degree the mode is known for.)
the tizita idiom, the pulse classes, the begena's buzzing drone TIMBRE.
(roughly 505 to 571, Aksumite), notated in melekket — its transmitted repertoire, its melekket
pieces, and the aquaquam liturgical dance with kebero, tsenatsel and meqwamiya as a playable set
piece. No sampling.
Europe's cathedral with a composed vocal identity while giving Lalibela only reverb is the failure
this doctrine names. So CH_09 gets an original wordless vocal lane built on the GRAMMAR of the three
zema modes — Ge'ez (plain, low, ordinary days), Ezil (slow and tender, fasts and funerals), Araray
(ornate and joyful, great feasts) — as an AFFECT and contour distinction, not as transmitted melody;
with begena buzz-drone timbre underneath; plus the azmari secular lane. Lalibela is SCORED, with its
stone architecture as the reverb and space of the mix rather than as a substitute for composition.
minor). Anchor degrees: the qenet's own gap-framing pitches. Periodicity: the 6/8 or 12/8 pulse
cycle. Harmonic rate: N/A (static krar cloud). Non-closure: continuous strum field with no cadence.
Contour norm: gap-vaulting with speech-shaped portamento.
Acheulean roughly 1.76 million; African Humid Period 11000 to 5000 BCE; continuous San tradition.
Sub-regions: Olduvai, Turkana, Tanganyika, Victoria, Bogoria, Naivasha, the Bale Mountains, a
Blombos forward-link, Lake Eyasi (Hadza) and Kondoa (Sandawe).
struck stone, breath, harmonic-series material and untuned environment. No living people's idiom
scores them, and the card states that in words so no composer reaches for the nearest Indigenous
register.
unison and heterophonic texture with hand clap and body pulse, the one-string zeze class,
shoulder-shaking dance rhythm. NAMED SACRED EXCLUSION: the epeme night ceremony and its songs, which
are restricted adult practice performed in darkness. No playable epeme, no imitation of its songs,
no depiction of the ceremony as spectacle.
of ordinary speech and song (screened per 3.5), simbo whistling and falsetto hunting communication
as a signalling colour. NAMED SACRED EXCLUSION: the trance-healing dance repertoire.
texture, ox-praise and herding song idiom. Excluded: initiation repertoire.
at 4.11a. The two chapters SHARE ONE CARD with per-chapter era pins; nothing is duplicated and
nothing is silently reassigned.
Periodicity: the clap or stamp cycle. Harmonic rate: N/A. Non-closure: cyclic. Contour norm:
PENDING-SOURCE.
what the live registry pins, and the research pass scored two peoples with a third's repertoire and
a third's UNESCO case. Aka (Bantu-speaking, Central African Republic and Congo), Baka (Ubangian,
Cameroon and Gabon) and Mbuti (Ituri) are distinct peoples with distinct languages and distinct
repertoires. They-all-do-hocket-and-yodel is precisely the flattening this doctrine bans elsewhere.
So the card splits into three rows keyed to the registry pins.
the yelli women's polyphonic hunting-call idiom as a TECHNIQUE class, mouth bow, forest ambience as
part of the score rather than under it. Excluded: named life-cycle and spirit-ceremony repertoire.
integration. NAMED SACRED EXCLUSION — and it is the most restricted item in the pinned set, omitted
entirely by the research pass: the MOLIMO, a sacred trumpet and a month-long ceremony whose
instrument is kept hidden from women and children. No playable molimo, no imitation of the trumpet's
forest-voice, no depiction of the ceremony as a spectacle, no molimo-shaped prop.
pieces document — layering, cyclicity, voice-breaks. Excluded: hunting and life-cycle songs, which
are named and functional. The UNESCO inscription and the Deep Forest appropriation case both attach
here: zero sampling of any field recording, zero ethnic-vocal library patch built from one. Our
singers, our material, their technique.
transcriptions are the analytic reference for interlock structure; applying that analysis to Baka or
Mbuti material is a stated inference, recorded as one, not a silent generalisation.
leitmotif system): OSTINATO WITH VARIATION, up to four parts over a fixed cycle, passacaglia-like,
every voice a variant of a shared model rather than a fixed written part; HOCKET AS METHOD, not
effect — write the theme, then DEAL its notes across three or four voices so no performer carries
the tune; YODEL and diphony, rapid chest-to-head register alternation, yielding a second
instrument-like line from one throat. Orchestral translation: flutes and clarinets in hocket,
pizzicato strings for interlock, no chordal pads. Harmony is a BYPRODUCT of interlock — fourths,
fifths, octaves with passing seconds and thirds.
voice of it.
the fixed cycle. Harmonic rate: N/A. Non-closure: cyclic. Contour norm: the shared model's contour.
generated "ethnic vocal" here would reproduce the exact appropriation case the register is famous
for, whatever the training data.
The live registry pins FOUR cultures to CH_12: Yorubaland (Nigeria), the Dahomey kingdom (modern
Benin), the Dogon of the Bandiagara escarpment (modern Mali), and Akan territory (modern Ghana). The
research pass supplied one register under a Yoruba heading and folded Fon and Dahomey under it as a
later stratum of the same music, which they are not. Three of the four cultures had no card, no palette
and no named sacred exclusion — and under this doctrine's own law, three of CH_12's four cultures were
therefore unshippable, with the highest-restriction material in the chapter unguarded. Four cards
follow.
THE TIMELINE IS THE LAW, and the figure must be right, because every other part is composed in
relation to it and an error propagates into every cue.
ONE-INDEXED — equivalently 0, 2, 4, 5, 7, 9, 11 zero-indexed — with the inter-onset interval string
2-2-1-2-2-2-1. The indexing convention is stated explicitly on every card because the two
conventions differ by one and a silent switch corrupts the figure.
STRING is what may not change. A set whose intervals are not 2-2-1-2-2-2-1 in some rotation is not
the standard pattern and is not a West African timeline.
which closes the 16-pulse cycle and keeps the seven strokes.
card rather than asserted.
three and three groups of four, with three to five parts each entering at its own point. There is no
shared downbeat consensus, which is why section 2.3.1 replaces hypermetre with cycle alignment here.
The dundun hourglass tension drum glides pitch as the drummer squeezes the osan cords, tracking
Yoruba's level tones and glides — the drum speaks. Iya ilu leads; gudugudu, omele isaaju, omele
atele and kanango hold the pattern; sekere and agogo colour it. ORCHESTRAL TRANSLATION: a solo
portamento instrument (trombone, cello) carrying a speech-contour motif — a theme that says a NAME,
native to this register.
spreading with the rise of Oyo. For the earliest Ile-Ife stratum lean on agogo, sekere, agba-class
drums and voice; bring the full dundun ensemble in for the Oyo stratum.
technique.
practice; Ifa divination verse (odu) and oriki praise-texts are living Orisha religion. No bata
liturgy, no fabricated odu, no invented oriki, no sampling. Shrine music is heard from outside, at a
distance, in our own material.
agbadza-class social dance idiom, gankogui bell and axatse rattle timeline, call-and-response.
vodun orders; initiation music. Ouidah's Temple of the Pythons is a living religious site, scored
from outside as our own material. Vodun is a living religion, not a horror register, and coding it
as menace is a care failure of the same class as the region-as-threat rule in 1.9.
method itself, which is a grammar); bell timeline; dance-drumming texture.
appellations and drum-texts are OWNED SPEECH — chiefly property attached to named stools and
lineages, not public-domain melody. No reproduction of a real appellation, no invented text
attributed to a real stool, and no state-occasion repertoire. The technique is free; the texts are
not.
and celebration song forms.
mask repertoire, and their instruments, which are restricted ceremonial property. No playable mask
ceremony, no reproduction of sigi or dama music, no mask instrument as a lootable prop.
interim pass.
The live registry pins CH_13 to Drakensberg, the Cape, Blombos, the Kalahari G|wi continuation, the
Witwatersrand and Kimberley mining region, AND Great Zimbabwe with the Zambezi and Mutapa interior
traverse on the northward leg to Ch 14. Both of the research pass's CH_13 sections were written as if
each owned the chapter, and the two registers are mutually incompatible on exactly the axes the
doctrine calls load-bearing. They are split here, and the general rule they force is 4.13.
declined by the communities themselves, and a care document should not carry it as a title.
The card says so rather than presenting another people's material as the pinned community's.
through the Bleek and Lloyd archive, with the descendant Khomani San community (whose name carries a
click letter that the orthography gate below must verify before it reaches a composer) engaged in
revival.
Claiming unbroken continuity and declaring extinction are both care errors; the archive framing is
the accurate one and it is what the card records.
typically a whole tone or semitone apart); the melody is the OVERTONE series selected by the mouth
cavity, roughly harmonics 4 through 10 over a fixed drone. Score as harmonic-series writing:
sustained low fifth, flageolet melody above, overtone flute or jaw-harp colour. Add the four- and
five-string pluriarc looping short cycles, single-string zither, light idiophones.
hand-clap polyrhythm whose timeline deliberately disagrees with the melodic downbeat. Drive is clap,
stamp and leg-rattle; drums are not the engine. Just and harmonic intonation, no heavy string
vibrato, which smears the overtones.
technically dense, never as primitive.
playable trance ceremony, no imitation of a named healing song, no sampling of archival Ju|'hoansi
or !Kung healing recordings — availability is not permission.
before it enters the card or reaches a composer. Misrendered click letters are themselves a care
defect, and at least one term in the research pass did not match standard rendering.
may enter on any of the 48 pulses. Compose a theme that survives without a V-I, because there is
none.
displaced and weaves through rather than doubling; the composite is played by neither player. Adding
the kutsinhira layer IS the intensity mechanism, and the pair is an ATOMIC STEM GROUP per 1.11.
extracts melodies nobody plays. Write the hummable theme AS one of those emergent lines and hide it
inside the interlock — this register's own melody-first, and the reason its recognisability test runs
on the interlock group rather than one stem.
clean plucked patch is wrong. Hosho gourd shakers run a continuous triplet stream. Vocal styles huro
(high yodel) and mahon'era (low humming) are wordless by tradition — an exact fit for the ruled
default.
(Kumadzulo iron strips indirectly dated roughly 500 to 700 CE); iron-working refinement roughly 1000
to 1100 CE enabled complex multi-key instruments; Joao dos Santos described a nine-key iron ambira
in 1586. For a period tell, use fewer keys than the modern 22-key dzavadzimu.
triplets, huro and mahon'era wordless vocal styles.
spirits and named pieces (Nhemamusasa, Nyamaropa among them) are that repertoire. No playable bira,
no named ceremonial melody.
CH_13 forces a general rule, because one grid contract per chapter cannot express two incompatible
registers: San is harmonic-series and just-intoned with clap-driven polyrhythm and no drum engine;
Shona is a 48-pulse non-cadential cycle in its own mbira tuning with continuous hosho triplets. A
composer asked to pin "CH_13's tempo family and tonal centre" cannot answer.
grid contract row (section 6.4).
WRITTEN TAILS. Crossfade is banned across a tuning change, because the drone-pivot and common-tone
handoffs of section 3.6 assume shared temperament and a crossfade between two tuning systems is
audible as being out of tune, not as a transition.
era pins and reciprocal cross-references, never two drifting copies.
Cross-register composer rules, binding everywhere:
6.7 makes it a pass or fail axis rather than an aspiration.
identifiable by INTERVAL SHAPE and RHYTHM alone — the same constraint that makes it hummable.
dundun part count, chhing subdivision. That is what the adaptive stems switch between.
real sacred language.
the sourcing law checkable instead of merely stated.
Every hero or canonical theme is authored ONCE as a composed work and rendered TWICE: as game stems,
bar-aligned, loopable and layerable under Quartz; and as a concert score and parts a human orchestra
can sit down and read. The score is the master artifact; the game render is a build of it.
So the theme carries real instrument assignments, ranges and bar structure from the first sketch. A
mockup written only to sound good will not survive reduction to playable parts: MIDI mockups are
generally not idiomatic when displayed in notation form, and layered patches — arco plus flautando
plus harmonics plus con sordino on one line — must be deleted down to one real section without losing
the demo's sound.
RANGES AT SOUNDING PITCH (practical, not virtuoso extremes). Piccolo D5 to C8; flute C4 to D7; oboe
B-flat3 to G6; cor anglais E3 to C6; clarinet in B-flat D3 to B-flat6; clarinet in A C-sharp3 to A6;
bass clarinet B-flat1 to F5; bassoon B-flat1 to E-flat5; contrabassoon B-flat0 to E-flat4; horn in F
B1 to F5; trumpet in B-flat E3 to C6; tenor trombone E2 to F5; bass trombone B-flat1 to B-flat4; tuba
D1 to F4; violin G3 to E7; viola C3 to E6; cello C2 to C6; double bass E1 to D4; harp C-flat1 to
G-sharp7.
WRITTEN RANGES FOR THE TRANSPOSING INSTRUMENTS, which is what actually goes on the page and into the
Dorico template, since section 5.5 adopts MOLA and MOLA requires transposing parts in the transposed
key. Piccolo D4 to C7 (sounds an octave higher); cor anglais B3 to G6 (sounds a fifth lower); clarinet
in B-flat E3 to C7 (sounds a major second lower); clarinet in A E3 to C7 (sounds a minor third lower);
bass clarinet C3 to G6 (sounds a major ninth lower — on a low-E-flat instrument the written floor is
E-flat3, sounding D-flat2); horn in F F-sharp2 to C6 (sounds a fifth lower); trumpet in B-flat
F-sharp3 to D6 (sounds a major second lower); contrabassoon B-flat1 to E-flat5 (sounds an octave
lower); double bass E2 to D5 (sounds an octave lower). Harp, flute, oboe, bassoon, trombones, tuba and
the strings other than double bass are written at sounding pitch.
WHY THIS DISTINCTION IS A GATE, NOT A PEDANTRY. A composer told to keep a trumpet part inside "E3 to
C6" while believing that to be a written range produces a part a whole tone flat of intent whose
bottom note is below the instrument's written floor — unplayable, and discovered in the room. The same
class of error appears in choral writing: tenor parts are notated an octave above sounding (section
3.2). Write inside these ranges and reserve the extremes for one deliberate effect per piece.
BALANCE IS REGISTER-DEPENDENT, NOT HEADCOUNT-DEPENDENT. The rule of thumb — four woodwinds balance one
heavy brass, two woodwinds balance one horn — is a starting heuristic only. Twenty flutes in their
lowest register cannot compete with a fortissimo trumpet, while one piccolo up top balances a trumpet
alone. Balance depends on instrument, number, dynamic, REGISTER, style, tempo, players, seating and
room. A close-miked sample template hides all of this; the concert deliverable exposes it.
VOICING, DOUBLING, VOICE-LEADING. Four-part writing still governs: include all, or the most important,
harmonic pitches; avoid over-doubling the third of a triad; space wider in the bass and tighter in the
treble; reduce any dense page to a legible two- to four-part skeleton with consistent doublings and
clean voice-leading. Default to BLEND (interlocking, register-compatible doublings) and diverge to
CONTRAST deliberately. Use ANTIPHONY, material handed between sections, and DOVETAILING, a line
continued seamlessly across instruments — the cheapest ways to make a hummable 16-bar tune survive
four minutes of concert time.
IDIOM PER FAMILY. Strings: divisi is not extra power — dividing celli or violas against loud brass
loses, it does not gain. Winds and brass: write the breath in; a flute line of 12 bars at a quarter
note equals 120 with no rest is a defect, and the orchestrator's test is to breathe along with the
playback. Percussion: notate high to low by relative pitch, with pitched staves adjacent.
MUSICXML IS THE INTERCHANGE SPINE, maintained by the W3C Music Notation Community Group; 4.0 stable
(June 2021), 4.1 a draft CG spec (April 2025); .musicxml uncompressed, .mxl compressed; more than 250
supporting applications. MIDI carries timing and performance; MusicXML carries notation semantics —
enharmonics, transposition, articulations, layout. MusicXML between notation programs, MIDI into the
DAW, and never rely on MIDI alone to reconstruct a score.
PROGRAMS. Dorico 6 (30 April 2025; 6.2 March 2026, 6.2.30 June 2026) is the engraving seat: condensing
for doubling players, cutaway scores, a proofreading panel that flags notation inconsistencies, cycle
playback. MuseScore Studio 4.6.x is free and cross-platform with active MusicXML import and export —
the tool for reading copies and outside collaborators. Finale's development ended, so nothing targets
Finale files; Sibelius stays valid if a hired orchestrator lives there.
THE PATH, IN ORDER. Sketch the theme melody-first at piano or short score. Author the full score in
Dorico on the real orchestra template. Export MIDI and MusicXML to the DAW for the game mockup, with
CC and keyswitch data living in the DAW only and never round-tripped back. Revise the score from what
the mockup taught, then extract parts. Notation to DAW is ONE-WAY. The reverse route is only for
material that began as improvisation, and then it takes the full cleanup: quantise, strip keyswitches
and CC, restore real note lengths, re-split layered patches into single sections, respell transposing
instruments, separate divisi onto their own staves.
Adopt the Major Orchestra Librarians' Association Guidelines for Music Preparation verbatim. Final
format PDF. Instrument full names at the head of the score and abbreviations thereafter; instructions
in English, Italian, German or French; tempo marks above the top staff AND above first violins; bar
numbers restarting each movement and placed identically throughout; rehearsal letters IN ADDITION to
bar numbers, at musical landmarks; score order woodwind, brass, timpani and percussion and harp and
keyboard, vocal, strings, high to low within families; barlines continuous within a family.
Parts: computer-engraved, no handwriting; transposing parts in the transposed key; clef and key
signature at the start of every line; bar numbers on each line with ranges on multi-bar rests; cues
during long rests transposed to the reading key, audible and obvious; "Tacet until" unacceptable
except to the end of a work or act; tempo and metre changes marked even during rests; doubling kept
out of principal parts and flagged where it occurs; harp pedalling left to the player; timpani printed
separately from percussion; one part per string section with complicated divisi on separate staves and
continuous barlines; flutes 1 and 2 as separate parts.
Formatting: double-sided, odd pages right and even left; score staff at least 4 mm; part staff 7.5 mm
ideal, below 7.0 mm unacceptable, above 8.5 mm avoided; 20 mm margin; 8 to 10 staves per part page for
ledger-heavy instruments, 12 maximum; page turns at multi-bar rests, no fold-outs; 60 to 70 lb (100 to
120 gsm) matte paper; parts 9x12 to 11x14 inches (A4 minimum, B4 preferred), scores 9x12 to 11x17
inches, bound to lie flat, coil past roughly 40 sides. Front matter: per-movement and total durations,
full instrumentation, transposing keys, a percussion list with required note ranges, percussionist
count, staging diagrams, notation key.
Finally, a qualified professional proofreads score and parts — not the composer, not the copyist alone
— and raw extraction is never trusted, because anchored markings drift and collide. Elaine Gould's
Behind Bars is MOLA's cited engraving authority; keep a copy at the desk. Section 5.9 states which
deliverable this specification gates, and which it does not.
Prepare the session the way Richard Jacques prepared Headhunter at Abbey Road: mock everything up
first, carry the TEMPO MAP from the DAW into the recording rig and verify it before the date, give the
room submixed guide stems. Record SECTIONALLY — strings separately from brass and percussion — for
balance, and because those sectional passes ARE the game's adaptive stems (strings, winds, brass,
percussion, choir, solo-culture instrument), each with reverb tails decaying to silence so the runtime
can fade a layer without a truncated seam.
Per-cue fork: the game take runs to CLICK (bar-locked, loop-clean, no baked fades); the concert
performance runs WITHOUT click. Roth conducts Distant Worlds click-free on roughly two three-hour
rehearsal calls plus a ninety-minute touch-up, and that rehearsal budget is the real constraint on
difficulty: sight-readable at tempo, one page-turn crisis per part at most.
know, one score per game cue.
timeline so the programme tells the saga.
forms — a symphonic poem, a piano concerto, a symphony — with the original composer consulting
rather than arranging.
Our doctrine keeps all three reachable. The game cue is model 1 by construction; the 79-node arc gives
model 2 a spine, with PROLOGUE through Ch 13 as the first suite; the 22 threads give model 3 its
subjects later. Budget a CONCERT SUITE of four to eight minutes, through-composed with no loop seams,
as a movement of the same composition — never a stitched medley of loops.
The sourcing law binds: the score NOTATES what a culture's register contributes; it does not paste it.
and pelog are not Western scale degrees, intervals vary ensemble to ensemble (embat), and
instruments are tuned to each other within one gamelan and not across gamelans. So a
gamelan-registered cue is scored either for a guest ensemble on its own instruments with the
orchestra written around it — never doubled in unison against it — or for orchestral surrogates
evoking stratification and colotomic structure without claiming to BE gamelan. The same rule governs
Khmer near-equidistant seven-tone tuning, mbira tuning, and just-intoned bow harmonics.
mark the ornament lane performer-realised. The part is a cue and a harmonic bed, never a
transcription of a consecrated performance. Care note for the ney specifically: the secular and
classical Ottoman and Persian register is one thing, and the Mevlevi ayin repertoire is restricted
ritual practice — the card distinguishes them.
monophonic and not ornament-carried in the duduk sense: it runs a kumbengo ostinato with birimintingo
soloistic elaboration over it. Notate the kumbengo and mark the birimintingo performer-realised.
Scoring a kora as a single ornamented line misrepresents the idiom. The mbira pair and the kotekan
pair belong to this class too, and they carry the atomic-group rule of section 1.11.
to mistranslate.
4.0.1 wherever the schedule allows. Care is authenticity and weight, never omission: washing the
instrument into strings is the defect, and the register must be recognisable to someone from that
culture — which is a claim only a person from that culture can settle.
The research pass made five external human dependencies blocking conditions with no owner, no cost, no
trigger and no fallback, and never said which deliverable each one gated — so as written the
concert-hall requirements blocked the GAME build of every hero theme in a solo-dev pre-revenue
project. An unreachable definition of done is ignored in practice, which is the failure mode this
doctrine exists to prevent. It splits.
DoD-A — GATES THE GAME BUILD. Entirely solo-executable.
adaptive_role and motif_carrier_stem filled.
DoD-B — GATES THE CONCERT RELEASE ONLY. Never blocks a game build.
copyist. Owner: director. Trigger: concert-suite commissioning. Cost: per-score engraving proofread
fee.
first booked performance.
visual-verification directive: no performable claim without a reader who has looked at the part.
Owner: director. Trigger: pre-commission.
any recording session is booked. This one also gates DoD-A if a vocal capture ships in the game.
booking. This gates public release of that register's music in either lane.
NO CONCERT-PERFORMABLE CLAIM IS MADE UNTIL DoD-B CLOSES for that theme. The doctrine's own honest-tier
register applies: structure, mockup, final — and "orchestrable" is not "orchestrated and read".
without stopping playback. Additive means a foundational track runs continuously and layers fade in
on top to escalate. Interchangeable means one layer crossfades out as another crossfades in — same
material, different orchestration — so density is constant and only the colour changes.
entry and exit points. Lower memory, and the only technique that can change FORM: key, metre, tempo,
phrase structure.
they must be written to be harmonically and rhythmically compatible with every bed they can land on.
tempo change that a crossfade cannot.
The standing warning against lazy vertical work: the best vertical scores start by being an excellent
piece of music with dynamics, swells in intensity and harmonic change BEFORE it is split into layers.
Each layer must be a complete musical thought, not a fragment. Compose the piece, then cut it. Never
assemble layers and call the sum a piece.
TIER H — hero and canonical cues, composer-in-loop, PRIMARILY HORIZONTAL. Curated fixed segments of
roughly three to five minutes that hand off to a new fixed segment when the player's state changes,
with playback logic tracking precise game states. That preserves through-composed melody, real
dynamics and a written climax — everything the 30-year bar depends on. Multi-phase bosses get
film-scoring treatment: intro cue, phase-1 cue, transition or chase cue, final-arena cue that states
the hero theme.
Provenance, scoped honestly. The precedent is Ori and the Blind Forest, where Gareth Coker wrote
curated fixed tracks rather than dynamic stem remixing. Will of the Wisps is billed on its interactive
music in granular detail, which points to layered or state-driven systems in the second game, and this
doctrine does not claim otherwise on unverified ground. So HORIZONTAL-FIRST FOR TIER H IS OUR OWN
RULING, supported by the Blind Forest precedent rather than inherited from a whole-franchise claim. If
a verified Will of the Wisps reading later shows what it layers — texture or motif — that reading is
added here and section 6.5 keeps or amends its evidence accordingly.
TIER B — regional beds, ambience and variation volume, generated, PRIMARILY VERTICAL. Bulk material
where the job is presence and place, not narrative. Additive layers on an intensity parameter.
The two tiers meet at region boundaries and at combat entry. RULE: a Tier-B bed may hand off to a
Tier-H segment, but a Tier-H segment is never demoted to a Tier-B layer.
A discrete STATE decides which music; a continuous INTENSITY parameter decides how much of it. Do not
collapse them — a single float cannot express "boss phase 2" and a single enum cannot express a swell.
Ship state set, per region or chapter: explore_calm, explore_alert, traversal, stealth, combat_light,
combat_heavy, boss_p1 through boss_p3, ritual or scene_lock, victory, defeat, silence. SILENCE IS A
FIRST-CLASS STATE, not an absence — Breath of the Wild's sparse design is deliberate negative space,
and its horse-riding cue ships with no percussion because the hoofbeats are the percussion. The L1.5
ambient lane must be able to select silence and mean it.
HYSTERESIS IS MANDATORY. Rise fast, fall slow: escalate on the next bar; de-escalate only after a
combat-clear timer (6 to 10 seconds of no threat) and then only at a phrase boundary of 4 or 8 bars.
Without this, one straggling enemy makes the score flicker, which reads as broken.
MAXIMUM DE-ESCALATION LATENCY IS A DECLARED NUMBER, not an emergent one. The clear timer plus an
8-bar phrase boundary at a slow tempo can approach 30 seconds, which no rule above would catch. Each
region's grid contract states a maximum de-escalation latency in seconds; if the clear timer plus the
declared phrase quantum at the region's slowest tempo family exceeds it, the grid contract FAILS and
the phrase quantum drops to 4 bars for that region.
Everything in a sub-region must be able to stitch. Pin these before a note is written. Note that the
contract keys on the sub-region per section 4.13, not on chapter_id.
exactly; segments may differ only across a written transition segment.
that modulate do so through a transition segment, never a crossfade.
latency rule of 6.3 demands it.
audio continuing past it is the reverb and decay tail, and it overlaps the next segment's pre-entry
pickup. This dovetail is what makes a transition sound written rather than cut. Deliver tails as
real audio past the exit cue; never bake a fade into a loop file.
way out that is not a fade. In non-cadential registers this is the cycle's own arrival stroke.
The melody never lives in a layer that an intensity parameter can mute.
SEGMENT decision. Changing how dense the world sounds is a LAYER decision.
escalate by instrumental migration — Baldur's melody travels solo Hardanger fiddle, then orchestral
strings, then a brass-led statement, while Kratos's long-lined theme reduces to a six-note fanfare as
action shorthand. The melody is present at every intensity; only its carrier changes. Additive
layering used for melody produces the failure mode where quiet states have no tune at all.
adaptive_role = motif_carrier and never leaves the role vacant; motif_carrier_stem is a field on the
state-machine row, and the mute test runs once per state. A globally-stated one-stem rule would
reject the doctrine's own recommended escalation design.
group.
diminution, register and octave transfer, reharmonisation under a fixed melody, orchestration
migration, and simultaneous counterpoint of two themes at climax.
tonic and its relative only, and quantise firing to the beat; a stinger that can land on any chord
will eventually land on the wrong one.
refused it, taking the harder path of preserving iconic melody through seamless switching, including
waiting for the downbeat before cueing visuals. We take the same harder path.
One Quartz Clock per region music director. Get the Quartz subsystem, create the clock, hold the Clock
Handle as a member variable since it is the garbage-collection anchor, make the clock settings and time
signature, set beats per minute. Every layer starts with Play Quantized against that clock at a
quantization boundary — bar for layer entry, 4 or 8 bars for a state change — and scheduling is
sample-accurate across the game and audio thread boundary. Subscribe to the quantization event to drive
the state machine on musical time rather than frame time, and reset transport on the final bar of the
loop so repetition has no discontinuity.
Two honest limits. The MetaSound builder subsystem has no variables support, so graphs past roughly
three to four layers are hand-authored in-editor. And the engine has no native transition-matrix
object, so the rules in 6.4 are authored as a MusicDirector component we own. If the matrix exceeds
roughly 8 states with per-pair rules, re-evaluate a middleware alternative on evidence rather than
defending the in-engine build.
THE LAYER BUDGET RECONCILIATION, which the research pass left as three different numbers with no rule.
The generated bulk lane's own multitrack output is six or seven stems (kick, bass, snare, hats,
percussion, synth, pad) while the bulk lane's engine cap is three to four hand-authored layers, and the
hero lane defines six sectional stems.
mapping from the generator's outputs: kick plus snare plus hats plus percussion sum to percussion;
bass sums to low sustain; synth and pad sum to mid harmony; the remaining textural output becomes
colour. That mapping is recorded as a benchmark-day acceptance criterion, and it is the deterministic
post-process's job, not a mixing opinion.
keys are stable; rare keys may be ignored or shifted. Stable BPM band is roughly 60 to 180. Use the
instrumental tag plus structure tags for wordless material.
(accompany a single track) — are Base-model-only, so the quality lane and the stem lane are different
checkpoints. Expect a two-model pipeline, or generate layer by layer with lego, which conditions on
the existing bed and is therefore in key and in time by construction. Lego is the better vertical
layer generator; prefer it.
requested value, trim to an exact bar count, place the loop seam at a zero crossing, render the tail
as post-exit audio, and reject any file whose recovered tempo drifts. Generate at 48 kHz, deliver
16-bit.
NEVER ORIGINATES A MOTIF. Motif quotation in a Tier-C cue, or the motivic-residue layer of a bed, is
a COMPOSER-AUTHORED STEM laid over generated material or fed to lego as audio conditioning — never a
prompt, because the generator exposes bpm, keyscale, timesignature, duration and caption tags and has
no melodic conditioning that could honour a specific three-to-five-note cell. The schema field
motif_source (authored or generated) makes the rule checkable rather than intended.
The bar-locked grid is a Western metrical assumption, and the research pass caught that. It did not
catch the larger one: the generator is 12-TET by construction, and equal temperament is what the
register sections call the loudest inauthenticity available to us. Since generated beds carry the
ambient identity of most of the world tour, an undetected tuning failure is the highest-volume care
defect in the pipeline. Three carve-outs, all recorded on the register card and all pass-or-fail on
benchmark day.
forms, breath-timed wind lines) or uses additive and aksak metres, that region opts out of vertical
bar-quantised layering and scores horizontally instead, with segment-boundary transitions and written
tails. The generator documents 4/4, 3/4 and 6/8 with 5/4 and 7/8 as advanced with varying results, so
it will silently flatten a culture's metre toward 4/4 if permitted.
sustaining instrument, pitched bed — to the composer and sampled-instrument lane, or through a
declared post-generation retune pass whose cents offsets are recorded on the card. Untuned noise-floor
layers stay generative.
portamento, dundun speech-glide, Balinese paired detuning — route their melodic layers to the authored
lane, because vibrato, pitch bend and microtonal variation are documented weaknesses of the generator
and the ornament IS the identity.
REGISTER ELIGIBILITY IS A DECLARED VERDICT, not an inference: every card records generation_eligible as
NONE (untuned noise floor only), PARTIAL, or FULL. Across the fourteen slice nodes, section 4 records
NONE for the pitched material of CH_04, CH_05, CH_06, CH_07, CH_09, CH_10, CH_11, CH_12 and CH_13, and
PARTIAL for CH_02/03 and CH_08. THE CONSEQUENCE IS NAMED HERE RATHER THAN DISCOVERED LATER: inside the
Josh-gate slice, the generative lane's lawful share is the untuned noise floor and some non-pitched
percussion, not bulk melodic beds. The ruled lane split stands unchanged — generation owns beds,
ambience and variation volume — but the volume it can lawfully own in the first fourteen nodes is small,
and the slice's music plan is budgeted on that basis rather than on an assumed bulk lane.
BENCHMARK-DAY RUBRIC AXES, pass or fail: metre fidelity against the declared timeline; TUNING FIDELITY
against the register's declared cents table, tested by A/B; ORNAMENT PRESENCE against the register's
declared ornament class; plus the existing register-fit and hummability axes.
Someone assembles the states out of the composer's material, and that authority is assigned in writing.
The cautionary case is a shipped title whose in-game score was reassembled from edited chunks by another
hand with final soundtrack duties reassigned. For Humanity Forgotten: FOR TIER-H CUES THE
COMPOSER-IN-LOOP OWNS THE ADAPTIVE ARRANGEMENT, not just the stem delivery. Tier-B assembly is delegable
to the automated lane.
Per-cue deliverable: written score (MusicXML and MIDI plus orchestration); declared BPM, metre, tonal
centre, tuning system and loop bars; segment stems at 48 kHz with per_stem_role, adaptive_role and
motif_carrier_stem; entry and exit cue offsets; post-exit tail rendered separately; transition segments
for every modulating pair; a head-motif stinger where the cue carries a theme; and a state-machine row
per transition (source state, destination state, sync-to boundary, transition class).
THE TRANSITION ACCEPTANCE RUBRIC, which the research pass left unscoreable. A cue is not delivered until
its transitions pass, and the test is defined:
boundary).
pitch clash against the incoming tonal centre; no late melodic entry; and the motif_carrier role
occupied in both source and destination states.
(section 6.3). Exceeding it fails the cue, not the mix.
Composer-in-loop is the ruled MECHANISM for the hummability bar, not a staffing note. This section is
how the lane actually runs, in order, with what each step produces and what stops it.
No candidate is written for a node whose register card is incomplete. The card is authored from the
live T0_Chapter_Index era and region pins, filled per section 4.0, put through the interim-tier
literature pass of 4.0.1 until no PENDING-SOURCE row remains, and recorded with its
generation_eligible verdict and its hummability substitution set. Output: a ratified register card.
Blocker: any PENDING-SOURCE row.
For each cardinal or family theme: the row's dramatic job, its family and inheritance justification,
its motif_class, its reveal_gate, its allow and deny register lists with the region-as-threat check,
its plant node and diegetic surface, its payoff node, and its closure policy. Output: a filled schema
row with generation_status = brief. Blocker: an unassigned reveal_gate on a gated theme.
Three to five candidate head cells per theme, each notated with its contour string and its ONE named
deviant. Written inside the register from the first sketch — never a Western tune later dressed. Where
the register is interlock-emergent, the candidate is written as the emergent line and dealt across its
atomic group at this stage, not retrofitted later. Output: candidate MusicXML plus a mono piano or
MIDI render of the primary statement only. Cost note: this step is where the 30-year bar is won or
lost, and it is the step that must not be rushed by pipeline pressure.
Run Appendix T.1 (recall) and the section 2.3 checklist against each candidate. Binary gates first —
a candidate failing B1 does not get diagnosed, it gets replaced. Then diagnostics, register-relative,
with N/A items dropped from the denominator. Log every candidate's scores, including the losers: the
score distribution across the first ten themes is what calibrates the INFERRED thresholds, and
discarding failed candidates destroys that data.
Output: a scored candidate table in the theme row's evidence trail. Decision: the highest scorer above
the gate proceeds; if none clears the gate, the failed items are named and step 2 repeats. A theme
does not proceed on schedule pressure — the ruled bar is that it iterates until it passes.
Iteration is targeted, never general. The failed items name the fix: a failed D3 is a contour rewrite,
a failed D4 is one anomaly too many or too few, a failed D8 means the rhythm is not a cell yet, a
failed B2 means the motif is over-determined by its harmonisation. Two iteration rules:
saves it. The research finding is unambiguous and the doctrine treats it as binding.
The winning candidate goes to full score in Dorico on the real orchestra template (section 5.4), inside
the written ranges of 5.2, with breath, divisi and page density self-checked per 5.3. The concert-suite
movement of four to eight minutes is composed as part of the same work, not extracted from loops later.
Output: score PDF, parts, .musicxml and .mid archived together.
Export MIDI and MusicXML one-way to the DAW. Mock up, then record or render sectionally per 5.6. Per
state, name the motif_carrier_stem and run the per-state mute test. Declare the atomic groups. Fill the
grid contract (6.4), the state-machine rows, the transition segments for every modulating pair, and the
head-motif stinger. Deliver tails as real audio; no baked fades.
Output: stems at 48 kHz delivered 16-bit, state-machine rows, transition segments. Gate: the transition
rubric of 6.9.
The eight gates of section 1.10, then DoD-A of section 5.9. The reveal check (gate 8) runs
fresh-context and capture-backed. The register review of 4.0.1 gates public release, and DoD-B gates
any concert-performable claim. Nothing here is signed off on description: every gate produces a capture,
a score sheet, or a named reviewer with a date.
calls per the delegated care doctrine, owns every human dependency in 5.9 DoD-B and 4.0.1, and does
the final integration. The director does not volume-compose.
arrangement, not just stem delivery (6.9).
interim-tier literature pass. Read-only, evidence-carrying, never self-marking.
placement and the registry writes.
PARTIAL or FULL, which in the slice is almost nowhere (6.8).
WHAT RUNS TODAY, SOLO, WITH NO EXTERNAL BOOKING: register cards and the literature pass; theme briefs;
candidates and scoring at the interim recall tier; orchestration and score authoring; mockups, stems,
state machines and the transition rubric; every gate in DoD-A. WHAT NEEDS A BOOKING AND THEREFORE A
TRIGGER AND A BUDGET: the human recall panel (Appendix T.1), the phonetic screen, the register review,
the tradition-bearer soloist, the choir session, the professional proofread, and the player read.
The cardinal set is drafted as a COHORT, not one theme at a time, because B3 (combinability) is
evaluated at cohort close and because the family cells must exist before Tier B can inherit. Cohort
one: PROTAGONIST_THEME, HOUSE_OF_VELHEIM_THEME, FAMILIAR_BOND, HOME_AND_LOSS and JOURNEY_WORLD, plus
the Tier-B family heads for the slice registers.
After the first ten themes have run the loop, the INFERRED calibrations are re-derived from the real
score distribution and this doctrine is revised in place: the 77 percent gate, the D4 anomaly count,
the D11 tempo rule, the six-to-ten transformation budget. That revision is a named work order, not an
aspiration, and it is the first amendment this document expects to take.
The research sections produced four separate required-field lists, all targeting T0_Theme_Registry,
which could not be merged into one table: they collided with each other (pitch_center against
tonal_centre against "diatonic home", three names for two concepts) and duplicated the live columns
rather than extending them. This appendix replaces all four. Every concept is named ONCE.
THE LIVE TABLE (verified 2026-07-29) carries 26 columns: theme_id, theme_name, theme_type,
canonical_anchor, mood_tags, instrumentation_substrate, cultural_substrate, era_substrate,
recurrence_anchors, hard_line_relevance, composer_in_loop_required, generation_status,
generation_prompt_hash, last_generated_timestamp, music_cue_id_ref, cinematic_anchor_array, notes,
extensions, populated_from, recent_changes, last_updated, version, tempo_bpm, bar_length,
per_stem_role, ue_sound_asset_path. [28au UPDATE: the table now carries 28 columns — row_class and
parent_theme_id landed with the roster reconciliation — and 14 rows (12 cardinal_theme + 2
cue_variant), all composer_in_loop_required.] At Appendix-S writing: twelve rows, all
composer_in_loop_required.
SAME-AS MAPPINGS onto live columns (no new column needed):
NEW COLUMNS, as a schema revision rather than free text in extensions, because six of them are gate
inputs and a gate cannot read a prose blob:
which required-field set applies.
row_class is cardinal_theme or family_motif AND motif_class is tune; NOT required for tuned_bed,
which has no head cell, plant or payoff. The research pass's unconditional requirement became
unsatisfiable the moment beds entered the same table; making it conditional on row_class fixes it.
motif_carrier_stem, transition_class, entry_exit_offset_ms.
REGISTER-CARD COLUMNS (living on the card, referenced by register_card_ref, and absent from all four
research lists): chapter_id, sub_region, era_window, tuning_system, cents_table_ref, timeline_or_cycle
with its indexing convention, ensemble_palette, free_grammar, named_sacred_exclusion, source_citation
per row, generation_eligible, the substitution set of 2.3.1, register_reviewer, register_review_date.
Every section cites this appendix. The variant protocols that previously appeared in the leitmotif,
hummability and wordless sections (one-hour delay; no listener count; no aggregate rule) are DELETED in
favour of these.
STIMULUS: a mono piano or MIDI render of the PRIMARY STATEMENT ONLY, with no orchestration, no reverb
and no production. Transposed to a range comfortable for the panel (roughly a male or alto comfortable
tessitura) — absolute pitch is a test-protocol decision, not a compositional gate (section 2.3.3, D6).
PROCEDURE: one hearing; a 60-second standardised distractor task; sing-back captured as audio.
SCORING: correct contour of the head cell plus at least the first five pitches in correct scale-degree
or interval relation, judged against the notated cell.
AGGREGATE RULE: n = 5 listeners, PASS = 4 of 5. The research pass defined per-listener pass and never
stated the aggregate; without one the gate is unscoreable.
DELIVERY, in two tiers, because THE JOSH GATE rules that the AI QA loop is the only player until the
AAAAA slice — so a human listener panel does not exist today and, as originally written, every hero
theme would stall at item 1 forever.
second musician re-notates the cell from memory after the distractor — plus a fresh-context critic
listen-check on the captured audio. A theme may proceed through orchestration on the interim tier. No
hummability claim is made on it beyond "passed the proxy gate".
(a paid micro-task panel or a local recruit), with a consent line covering the capture. Owner:
director. Trigger: pre-slice-lock. Cost: five short sessions per theme.
Clap test (D8): the candidate's rhythm clapped with no pitch; a naive listener matches it to the right
theme out of three. First-notes test (D1): the first three to five notes played on a second hearing;
the listener names the character or place. Both run on the same panel tier as T.1 and are recorded with
it.
All judged by a fresh-context critic on CAPTURED PLAYBACK, never on a described claim.
drums carried the identity, the register is a wash. This is the test that proves a card's non-
percussion lane exists (see CH_08 at 4.7).
tonal_centre is wrong.
A harmonic-transformation identity has no tune to sing back, so it takes its own test, which must be
authored and dry-run BEFORE the vril and wonder theme starts.
discovery, a threat, a homecoming) and matches the gesture to one.
cannot recover the meaning, the gesture is colour, not a motif, and it is re-classed as timbre or
rewritten.
Items A1 through A5 of section 2.4, run on captured playback, for every motif_class = timbre and every
tuned_bed row.
those do not reconcile (written E3 sounds D2 on a major-ninth transposition), so section 5.2 records
the arithmetic-consistent pair C3 to G6 and notes the low-E-flat instrument variant.
it cannot close a 16-pulse cycle with the standard pattern's seven strokes; section 4.11a records
2-2-3-2-2-2-3 (sum 16, seven strokes) instead. The finding's twelve-pulse correction is adopted
exactly as given.
a local-name claim without a citation, and 4.11a calls the figure the standard pattern with local
names deferred to the card's cited source.
gong gede post-Majapahit court ensembles roughly 400 years wrong for the 1200 CE pin; the other asks
that semar pegulingan be pinned as CH_04's primary register to give the chapter a lawful ensemble.
The period ruling wins, and the second finding's real defect — a palette prescribed and excluded in
the same breath — is fixed by the grammar-and-repertoire split instead, which gives CH_04 a lawful
named register (selonding and gambang classes with original material) without a period violation.
The post-stratum ensembles remain available inside a declared time-fluctuation window.
could be fetched this pass, so section 3.1 records the reading as CONTESTED, bars any composer from
copying the shape until a transcription check runs, and states only what survives both readings. The
finding is adopted as a work order rather than as a correction.
verified Will of the Wisps reading) is not, because no verification was available this pass. Section
6.2 therefore states horizontal-first as OUR ruling supported by the Blind Forest precedent.
tizita minor and bati major and minor be re-checked against the same source class; that check could
not run this pass, so the whole table ships PENDING-SOURCE with a hard authoring block rather than as
a set of pinned rows.