THE SNES-JRPG SCHOOL — leitmotif economy, harmony, and the loop
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: NONE FOUND — this document cites no canon anchor anywhere in its body.
Do not build from it alone: open the spine node (docs/spine/CH_NN.md), the region page
(_source/02_Tier_2_Region_Pages/), and the T0 registry your work targets, and derive from those.
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 other half of the nostalgia class — Uematsu, Mitsuda, Shimomura and the 16-bit RPG score: one main theme parenting a corpus, the identity cell as the unit that recurs, the two named cadence families, and the loop discipline that keeps a 60-to-120-second body from turning repellent. It also carries the two TOOTH rules that bind this project's own process: grammar is learned and melodies are never reproduced, and every A/B runs the current theme as the losing control.
---
0. Provenance — where this text came from, and what was done to it
TIER: PROPOSAL. This is research, not canon. It binds nothing until ratified; the CVD and the
T1/T0 canon outrank every line of it.
SOURCE: the research return of workflow run wf_d1ba57a8-e47, agent a4edf7004db1ab405, 2026-08-05. The return was a
structured object with three keys — rules, evidence, measurable_features — and all three are
reproduced below.
EXTRACTION: VERBATIM and mechanical. The strings were read out of the run journal by script and
written here unedited — not re-worded, not re-ordered, not summarised. Only the headings, the
bullet markers and this provenance section are new. That matters because NOSTALGIA_RUBRIC.md
cites these rules BY ID, and a citation that resolves to a paraphrase of its rule is the same
WHY IT EXISTS AS A FILE: the returns lived only in the run journal, which is agent scratch and
is not canon-readable. The rubric quoted them anyway and said the mechanism was "stated verbatim
from the school grammars" — which named documents that did not exist. That is finding F3 of the
cold verification of 2026-08-05, and this file is half of its fix (the other half is the rubric's own
provenance line).
WHAT IS NOT CLAIMED: that every source below was independently re-verified. The evidence list
carries the citations the research agent gave, with its own declared gaps and negative controls
left in. Verifier finding F4 stands open against this: several named scholarly sources are bare
author-year strings, and a reader cannot resolve them from this repo alone. Treat a citation here
as a pointer to check, never as a checked fact.
---
1. The rules
LEIT-01 ONE MAIN THEME GOVERNS THE CORPUS. A single main theme is the parent of the score; canonical character/place/thread themes are DERIVED from it, not written independently. Anatone's five documented Uematsu strategies are the sanctioned derivation set and every derived theme must declare which one it used: (a) eponymous omission, (b) motivic networking, (c) thematic hybridization, (d) associative troping, (e) double main theme. MEASURE: >=70% of hero/canonical themes share a >=4-event ordered pitch-interval cell with the main theme at ANY transposition; each carries a declared strategy tag.
LEIT-02 THE IDENTITY CELL IS SHORT AND IS THE THING THAT RECURS. Every motif has a named cell of 3-7 pitches occupying <=2 bars. The cell, not the full 16-bar tune, is what is deployed across the game. Kondo built all six Ocarina songs on a repeated THREE-note motif so players could remember them without opening a menu. MEASURE: cell length <=7 pitches AND <=2 bars; cell is extractable as a standalone MIDI clip.
LEIT-03 INTERVAL SPINE IS THE INVARIANT; RHYTHM, MODE, TEMPO, TIMBRE ARE THE VARIABLES. A recurrence is valid only if the ordered interval sequence survives while at least two of {rhythm, mode, tempo, orchestration} change. FF6 Terra's theme recurs across Opening Theme, Awakening, Metamorphosis, Searching for Friends and Ending under exactly this rule. MEASURE: per motif, count distinct transformation types used across the corpus; hero motifs require >=3.
LEIT-04 DEPLOY BY ABSENCE AS WELL AS PRESENCE. A motif's disappearance is a scored narrative event: Terra's theme IS the World of Balance overworld and is REPLACED (by Dark World / Searching for Friends) in the World of Ruin. Every hero motif gets a declared silence window and a scored return. MEASURE: each hero motif has >=1 contiguous absence window of >=5 chapters logged in the motif ledger, with a named return node.
LEIT-05 CHARACTER THEMES ARE STABLE IDENTITIES, RE-CONTEXTUALIZED — NOT WAGNERIAN-DEVELOPED CUE BY CUE. In FF4-VI the character themes are mostly NON-transformative in themselves; the drama comes from what is put around and under them. Spend transformation budget at the game-arc scale, not inside a single cue. MEASURE: within any one cue, the motif's own interval spine is unmodified; change is carried by accompaniment/harmony/instrument only.
LEIT-06 THEMATIC FAMILIES, NOT A FLAT ROSTER. Group the 22 threads / 76 regions into 3-6 families; each family gets its own derived cell so cross-references are audible without every cue quoting the main theme (Atkinson's 'thematic families' reading of FF9). MEASURE: every canonical theme carries exactly one family_id; each family has >=2 members sharing a family cell distinct from the main-theme cell.
LEIT-07 QUOTE FRAGMENTS, DON'T RESTATE. At reprise points, embed 2-4 notes of the cell inside NEW material rather than replaying the tune. Mitsuda's practice is to reshape the melody substantially, quote only fragments, or write new material around it (Morning Glow, Lavos Theme, World Revolution). MEASURE: at reprise nodes, quoted-cell length is 2-4 events and quoted material is <25% of the cue's melodic note count.
ECON-01 THE MELODY MUST WORK NAKED. Render the lead alone on a flat square/sine with no harmony, no pad, no reverb. If it does not read as the theme, it is not the theme yet. This is the operational form of the hummability bar and it is a HARD gate before any orchestration spend. MEASURE: melody-only render passes blind critic recognition and blind A/B against the exemplar-anchored control.
ECON-02 FOUR-VOICE BUDGET IS THE COMPOSITION FRAME, WHATEVER THE FINAL ORCHESTRATION. Compose as LEAD + COUNTER-LINE + BASS + light percussion/arpeggio; every additional part is doubling of one of those four. Kondo's NES roster was melody / accompaniment / low end / percussion / harmonization, and that roster is what produces the register. Orchestral wash is NOT the fix for a weak theme. MEASURE: distinct non-doubling voices <=4 at any instant; a 4-voice reduction of the full arrangement must be recognizable as the same cue.
ECON-03 REGISTER SEPARATION IS MANDATORY. Lead, counter-line and bass occupy disjoint or barely-overlapping octave bands. Harmony is IMPLIED by arpeggiation and bass motion, not stated by block pads. MEASURE: lead/counter octave-band overlap <25% of sounding note-time; sustained block-chord pad <20% of cue duration.
ECON-04 SINGABLE SPAN. Hook total range <= octave + perfect 4th (17 semitones); identity cell range <= major 6th (9 semitones). MEASURE: semitone span of hook and of cell, both computed on the lead track.
ECON-05 STEPWISE MAJORITY WITH ONE SIGNATURE LEAP. Predominantly conjunct motion carrying exactly one characteristic wide interval that IS the theme's fingerprint — Uematsu centres the FF7 theme on a major 7th that recurs across the soundtrack. MEASURE: >=60% of melodic intervals are <=2 semitones; 1-2 intervals >=7 semitones per 8-bar phrase; the signature interval is declared per theme and recurs in >=2 phrases.
ECON-06 PHRASE ARCHES ANSWER THEMSELVES. Kondo's documented practice: short phrases with upward motion finish with long phrases of downward motion (Song of Storms, Saria's Song); phrases with downward motion finish upward (Song of Time, Zelda's Lullaby). MEASURE: net-contour signs of paired phrases are opposite in >=50% of phrase pairs; the answering phrase is longer than the antecedent in >=1 pair.
ECON-07 RHYTHMIC ECONOMY. The hook uses <=4 distinct note durations, and the identity cell's rhythm is memorable enough to be TAPPED without pitch. Zelda's Theme is built on a three-note basic idea with a half-quarter-half rhythmic template that is then fragmented. MEASURE: distinct durations in hook <=4; cell rhythm passes a pitchless-recognition critic pass.
ECON-08 THE COUNTER-LINE IS A SECOND MELODY, NOT A PAD. Monophonic-channel heritage forces real voice leading and counterpoint: move each voice by the smallest interval that keeps the chord identifiable, and put contrary motion against the lead at least once per phrase pair. Published analysis attributes classic-VGM memorability to the ACCOMPANIMENT as much as the tune — melody alone does not explain endurance. MEASURE: contrary-motion ratio lead-vs-counter >=0.30; counter-line passes the naked-render test as a coherent line.
ECON-09 CONSTRAINTS ARE THE CRAFT, NOT A HANDICAP TO ESCAPE. Uematsu: the more limited people are, the more ingenious they get; Mitsuda: the limited Super Famicom sound set let you focus more on melody; Kondo built six memorable songs from a FIVE-note ocarina palette. Therefore: impose an artificial palette limit on every hero theme before orchestration is allowed. MEASURE: each hero theme is authored first under a declared restricted palette (<=5-7 pitch classes for diegetic/instrument themes; <=4 voices always) and that restricted version is archived as the theme of record.
ECON-10 CLASSICAL SENTENCE STRUCTURE IS THE DEFAULT PHRASE GRAMMAR. Zelda's Theme is two 8-bar sentences plus one 8-bar sentence in binary |A B| loop, each sentence being basic-idea / repetition / continuation-by-fragmentation / cadence, with liquidation in the B section and perfect authentic cadences AVOIDED. MEASURE: hero themes parse as [BI - BI' - continuation - cadence] over 8 bars; cadence inventory contains >=1 half cadence and 0 PACs before the final loop-exit.
HARM-01 MODAL MIXTURE FROM THE PARALLEL MINOR IS THE DEFAULT COLOR SOURCE. In major-key wonder/triumph writing, bVI, bVII and bIII against a major tonic are the school's core vocabulary; Uematsu's FF7 main melody runs I - vi - I - bVI7 - bVII7 - I. MEASURE: >=1 borrowed chord per major-key cue; borrowed-chord density reported per cue.
HARM-02 TWO NAMED CADENCE FAMILIES, BOTH SANCTIONED. (a) ZELDA CADENCE I - bVII - bVI - V — heroic, forward-leaning, documented as an Andalusian-cadence adaptation turned positive/heroic. (b) MARIO CADENCE bVI - bVII - I — arrival and triumph; the bVI is the pivot that solidifies the tonal center, substituting for vi while retaining the tonic pitch, and bVII substitutes for the dominant via voice-leading rather than leading-tone pull. Kondo's own Serenade of Water lands on Bb - C - D. MEASURE: every hero/arrival/fanfare cue contains >=1 of the two families, detected on the roman-numeral reduction.
HARM-03 CHROMATIC MEDIANTS = 'DARK WITHOUT SAD'. Major chords a third away (notably bIII) darken without signalling grief — the register for ancient / otherworld / mystery. Never use them as a sadness cue. MEASURE: chromatic-mediant count per cue, cross-checked against the cue's declared affect tag (mystery/ancient cues >=1; grief cues may be 0).
HARM-04 NON-FUNCTIONAL HARMONY IS THE WONDER ENGINE. Chords chosen for color rather than function — planing/parallel triads, progressions that do not resolve — produce the hazy dreamlike register (Mitsuda, Secret of the Forest, described as balancing melodic writing against atmospheric suggestion). MEASURE: fraction of chord changes not explainable by functional roman-numeral syntax; wonder/exploration cues target >=0.30, hero/arrival cues <=0.15.
HARM-05 EXTENSIONS LIVE IN THE ACCOMPANIMENT, NEVER IN THE TUNE. The Wise ambient register runs on min9 and maj7#11 sonorities (Aquatic Ambience is documented as Cm9 to Abmaj7#11) while the melody above stays plain and singable. MEASURE: chords carrying 7ths/9ths/#11 >=0.40 of chord events in ambient cues; melodic notes landing on the extension tone <=0.20 of hook notes.
HARM-06 OSTINATO-OVER-CHANGES. A fixed melodic/rhythmic figure holds while the harmony moves underneath; the figure does not change meaning, the harmony changes ITS meaning. In the JRPG form survey 8 of 10 loops carry an ostinato and 6 hold it basically throughout; ostinato is also what makes a loop point survivable without breaking immersion. MEASURE: longest repeated >=2-bar figure covers >=60% of exploration/ambient loop duration; >=2 distinct harmonies occur beneath the unchanged figure.
HARM-07 PEDAL POINTS AND CHROMATIC BASS ARE THE TWO STANDING BASS DEVICES. Static bass under moving upper harmony = wonder/threshold; chromatically descending bass under a near-static melody = pastoral/nostalgic (Lon Lon Ranch). MEASURE: pedal-bar percentage and chromatic-descent detection reported per cue; every wonder or pastoral cue fires one of the two.
HARM-08 DECEPTIVE CADENCE IS THE IN-CUE MAJOR-TO-MINOR PIVOT; BORROWED-CHORD PIVOTS ARE THE KEY-CHANGE DEVICE. Uematsu uses V-vi to set up the listener for V-I and land on the minor instead, and modulates by common TONE (a shared pitch, e.g. B natural bridging E major to G major) and common CHORD — advancing the technique by using BORROWED chords as the pivot. MEASURE: cues with an internal mode flip contain >=1 V-vi; modulations are annotated with pivot type (tone / chord / borrowed-chord).
HARM-09 MODE ASSIGNMENT IS PART OF THE BRIEF, AND PER-CULTURE INSTRUMENTATION RIDES ON TOP OF IT — IT DOES NOT REPLACE IT. Dorian (raised 6th) = adventurous/pastoral-with-hope; Aeolian = elegy and time passing; Ionian-with-mixture = wonder and triumph; harmonic-minor color = exotic/other (Gerudo Valley); Mixolydian/bVII = folk-heroic; Serenade of Water is D Dorian in section 1 shifting to D Aeolian and D Ionian in section 2. MEASURE: every cue declares mode(s) per section; mode-change events counted; declared mode matches the region/affect table.
AFF-01 THE SCHOOL'S SIGNATURE AFFECT IS MELANCHOLIC NOSTALGIA — THE BALANCE BETWEEN HAPPY AND SAD — AND IT IS BUILT BY MODAL AMBIGUITY PLUS VOICE LEADING, NOT BY PRODUCTION VALUES. Kondo's progressions are described as difficult to pin as clearly major or minor, flourishing from voice leading rather than a chord chart. MEASURE: major/minor classifier confidence on the hook should be LOW (target 0.45-0.65, i.e. ambiguous) for nostalgia-register cues; a cue that classifies as unambiguously major or minor has probably missed the register.
AFF-02 WONDER = ASCENDING LINE AGAINST DESCENDING ACCOMPANIMENT. In Serenade of Water the melody rises slowly and steadily in constant quarter notes AGAINST cascading arpeggios; contrary motion plus modal coloration is the wonder engine, not reverb or string size. MEASURE: wonder-tagged cues show sustained opposite contour between lead and accompaniment over >=4 consecutive bars.
AFF-03 MELANCHOLY IS HARMONIC, NOT ATMOSPHERIC. Build it by putting borrowed-minor harmony under a major-key melody, or by restating the same melody a mode lower. Slowing down and adding reverb is not the device. MEASURE: melancholy-tagged cues must fire >=1 of {borrowed-minor reharmonization of an unchanged melody, mode-lowered restatement}; reverb/tempo deltas alone fail the check.
AFF-04 COMPOSED TEMPO CHANGE IS A LEGITIMATE STRUCTURAL DEVICE FOR TIME PASSING. Serenade of Water carries a structural ritardando 90 to 75 bpm that slows perceived time. Write it into the score; do not leave it to the mix. MEASURE: tempo map contains composed rit./accel. events where the cue's brief calls for temporal shift.
AFF-05 TIMBRE MARKS PLACE; IT NEVER RESCUES A THEME. Wise deliberately does NOT keep a fixed instrument set per game — he picks instruments that match the environment; Mitsuda brings non-Western instruments in for mysticism (Corridors of Time, Schala's Theme) and rare orchestration touches like divided high violins for unsettled ethereality (Wind Scene). But the theme must already work on a piano. MEASURE: piano-reduction test passes BEFORE instrument assignment; instrument family assignment checked against the culture/period table.
AFF-06 TWO-TONE MAIN THEME. The standard main-theme form in this school is a heroic/driving A-section against a sweeping/melancholic B-section, mirroring the game's dual tone — the Chrono Trigger main theme is built exactly this way. MEASURE: the main theme parses into >=2 sections with opposed affect tags and a measurable contrast in tempo-feel, density or mode.
LOOP-01 LOOP LENGTHS ARE PRESCRIBED BY CUE CLASS. Hook-bearing theme loops 30-60s; ambient/exploration 1:30-2:30; combat 15-30s. Half the analyzed JRPG loops sit under 1:00 and half at 1:30-2:30. MEASURE: loop duration in seconds and bars, checked against cue class.
LOOP-02 THE CONTINUOUS THEMATIC CHUNK IS 30-50 SECONDS — roughly 16 bars at moderate 4/4. That is the unit players actually remember; everything else is frame. MEASURE: duration of the contiguous span containing the hook; target 30-50s.
LOOP-03 SANCTIONED LOOP ARCHITECTURES: AAB, AABA', or four equal sections. Two-section loops are permitted ONLY when phrase-level construction is measurably more complex to compensate. MEASURE: section map emitted per cue; form label from the allowed set, or a documented complexity justification.
LOOP-04 VARY BETWEEN REPEATS WITH THE CHEAP FOUR, NEVER WITH NEW HARMONY. Add/remove counter-melody, shift octave, change instrumentation, change register. The tune and the changes stay put. MEASURE: >=1 of the four variation types fires between repeats; harmony delta between repeats = 0.
LOOP-05 EVERY LOOP CARRIES A BREAK. A tag or a breakdown to rhythmic hits gives relief before the loop point — all but two of the ten analyzed JRPG loops have one. MEASURE: a bar range of reduced texture or rhythmic-hit writing exists in the last third of the loop.
LOOP-06 THE SEAM MUST BE INAUDIBLE. Section entrances land square on the beat; no ramp-in, no pickup that reads as a stutter across the loop point. MEASURE: section starts quantized to downbeats; pitch/velocity/RMS discontinuity at the seam below threshold; blind seam-detection test fails to locate the loop point.
LOOP-07 CONSCIOUS-LOOP THRESHOLD. The ear starts registering unchanged repetition consciously at roughly 90-120 seconds; any single unchanged block must sit under that, and beyond it LOOP-04 variation or block shuffle must fire. MEASURE: longest unchanged block duration <90s.
LOOP-08 BLOCK-SHUFFLE FOR EXPLORATION MUSIC (THE KONDO METHOD). Build the field cue as ~8-bar / ~15-second blocks over RELATED chord progressions and let an algorithm order them; state-shift on player behaviour — calmer when the player stands still, heroic when an enemy appears. MEASURE: exploration cues ship as >=6 interchangeable blocks with a compatibility matrix and >=2 player-state variants.
LOOP-09 INTERACTIVE IDEAS ARE PART OF THE COMPOSITION, NOT THE IMPLEMENTATION. Kondo's bar is musical ideas that could only be thought of in a video game — tempo accelerating in the last 100 seconds, the bass subtly changing deep underground, a different cue underwater. Each cue brief must name at least one such idea. MEASURE: every cue carries >=1 declared interactive musical hook, authored in the score.
LOOP-10 DO NOT 'FIX' THE NOSTALGIA REGISTER BY GOING THROUGH-COMPOSED. Across a 21,391-piece VGM corpus, repetition structure stayed essentially CONSTANT across platforms and console generations; what actually improved with hardware was song length and instrument-track count. Repetition is the idiom, not a hardware artifact to outgrow. MEASURE: repetition ratio of candidate cues sits inside the exemplar-class band, not below it.
TOOTH-01 GRAMMAR IS LEARNED; MELODIES ARE NEVER REPRODUCED. Cadence formulas (bVI-bVII-I, I-bVII-bVI-V), modes, ostinato figures, sentence structure and loop forms are NOT protectable and ARE the intended borrowings. Actual tunes are never copied. MEASURE: max contiguous match of (pitch-interval, duration-ratio) pairs against the exemplar corpus must be <6 events, transposition-invariant; any hit >=6 is an automatic reject.
TOOTH-02 EVERY A/B RUNS THE CURRENT THEME AS THE LOSING CONTROL PLUS AN EXEMPLAR-CLASS ANCHOR. A candidate that beats only the incumbent has not cleared the bar; it must be scored against the exemplar-anchored rubric with the school's grammar rules as the scoring axes. No music artifact reports success on proxy metrics alone. MEASURE: per candidate, win-rate vs incumbent control AND vs exemplar anchor, with per-rule rubric scores attached.
---
2. The evidence
Anatone, Richard. 'Leitmotivic Strategies in Nobuo Uematsu's Final Fantasy Soundtracks.' Music Theory Spectrum 45/2 (Fall 2023): 257-283. Names the five techniques connecting a game's main theme to character themes — eponymous omission, motivic networking, thematic hybridization, associative troping, double main theme — and shows FF4-VI character themes are mostly NON-transformative individually while forming inter-related thematic families around the main theme. https://academic.oup.com/mts/article-abstract/45/2/257/7225384
Anatone, Richard (ed.). The Music of Nobuo Uematsu in the Final Fantasy Series. Intellect Books, 2022, 372pp, ISBN 978-1-789-38554-0. Four parts x three chapters; includes Sean Atkinson, 'That Tune Really Holds the Game Together: Thematic Families in Final Fantasy IX'; Ross Mitchell on the diatonic rhythms of Uematsu's battle music; Anatone on the FF6 opening sequence. Source for the thematic-family construct (LEIT-06). https://www.intellectbooks.com/the-music-of-nobuo-uematsu-in-the-final-fantasy-series
USM masters thesis, 'An Analysis of Nobuo Uematsu's Linear Structures' (Univ. of Southern Mississippi, aquila.usm.edu). Finds linear progressions guiding the music, with the most important linear structures supporting a movement's tonal identity and their END POINTS COINCIDING WITH JUNCTURES IN THE GAME'S PROGRAMMING — i.e. tonal design is written to the loop/state boundary. Grounds LOOP-06. https://aquila.usm.edu/cgi/viewcontent.cgi?article=1882&context=masters_theses
Video Game Music Shrine, 'Inside The Score — Koji Kondo (Ocarina of Time OST).' Documents the ZELDA CADENCE I-bVII-bVI-V (C-Bb-Ab-G) as an Andalusian-cadence adaptation made 'more positive and heroic' via modal mixture borrowing two chords from the parallel minor, and the MARIO CADENCE bVI-bVII-I (Ab-Bb-C) for triumph; describes Kondo's harmony as hard to pin as major or minor, flourishing 'from voice leading than any music theory chord chart', and the NES channel roster (melody / accompaniment / low end / percussion / harmonization) forcing voice leading and counterpoint. Grounds HARM-01/02, AFF-01, ECON-02/08. https://videogamemusicshrine.com/inside-the-score-the-legend-of-zelda-ocarina-of-time-original-soundtrack-koji-kondo/
Video Game Music Shrine, 'Inside The Score — Serenade of Water.' D Dorian in section 1 shifting to D Aeolian and D Ionian in section 2; the Dm-to-G i-IV move; the Mario cadence Bb-C-D; melody rising in constant quarter notes AGAINST cascading arpeggios; Dorian's raised 6th producing major/minor ambiguity as the wonder device; a structural ritardando 90 to 75 bpm slowing perceived time; and two distinct melodies drawn from the same progression under the ocarina's five-note limit. Grounds HARM-09, AFF-02, AFF-04, ECON-09. https://videogamemusicshrine.com/inside-the-score-the-legend-of-zelda-ocarina-of-time-serenade-of-water/
Koji Kondo, 2001 composer interview (shmuplations translation). Melodies 'easy to remember' inside structures with 'neat twists and turns'; the interactivity bar — 'musical ideas which enhance the interactivity of the game, ideas which could only be thought of in a video game' (tempo speeding in the last 100 seconds of SMB; the bass subtly changing deep underground and different music underwater in SM64); Ocarina field music built as short 8-bar blocks played randomly, calmer when Link rests and shifting to a heroic melody when an enemy appears; and 'the most important thing is feeling like the song really matches what's happening in the game.' Grounds LOOP-08, LOOP-09. https://shmuplations.com/kojikondo/
Nintendo/Zelda Universe feature on Kondo's Ocarina sound design. All six ocarina songs (Zelda's Lullaby, Epona's, Saria's, Sun's, Song of Time, Song of Storms) are built on a repeated THREE-NOTE motif — two short three-note phrases resolving into a longer phrase — explicitly so players could remember them without opening the song menu; the initial motif is reused inside the longer phrase for cohesion; the whole ocarina system is written from a five-note palette (D, F, A, B, D). Grounds LEIT-02, ECON-09. https://zeldauniverse.net/features/the-brilliance-of-koji-kondos-sound-design-in-ocarina-of-time/
'Zelda's Theme: Sentence Structure' (High Score music analysis blog, 2020). Binary |A B| on continuous loop; A = 16 bars as two 8-bar sentences, B = 8 bars as one sentence; each sentence is [basic idea - repetition - continuation - cadence] with the basic idea a three-note motif on a half-quarter-half rhythmic template, continuation by FRAGMENTATION, and liquidation in B; prevalent dominant harmony with perfect authentic cadences AVOIDED; A disjunct, B conjunct; fourths and fifths recycled as the unifying interval. Grounds ECON-10, ECON-07. https://highscore.music.blog/2020/03/19/zeldas-theme-analysis/
Koji Kondo, documented melodic-arch practice (via Ocarina analyses): short phrases with UPWARD motion finish with long phrases of DOWNWARD motion (Song of Storms, Saria's Song), while phrases with downward motion finish upward (Song of Time, Zelda's Lullaby). Grounds ECON-06. https://zeldauniverse.net/features/the-brilliance-of-koji-kondos-sound-design-in-ocarina-of-time/
'The Mario Cadence' (Game Music Theory blog, 2011). Formula bVI-bVII-I, expanded in SM64 'Dire Dire Docks' to I | bVII | I | bVII | bVI | bVII | I; bVI substitutes for vi while retaining the tonic pitch and enabling a descending-third progression from the lowered third degree; when bVI enters, 'the sense of cadence, or at least tonal center, becomes solidified'; bVII substitutes for the dominant, creating momentum through voice-leading rather than leading-tone motion (citing Walter Everett on major-mode systems with modal-degree mixture). Attested across SMB, SMB2, SMB3, SM64, Galaxy and multiple Ocarina pieces. Grounds HARM-02. http://gamemusictheory.blogspot.com/2011/04/mario-cadence.html
'7 Music Theory Lessons from Final Fantasy VII' (Game Developer). Uematsu's melody balances stepwise and skipping motion around a MAJOR 7th interval that recurs across the soundtrack; deceptive cadence V-vi to flip major to minor; ostinatos anchoring harmony and key changes, 'essential for limited sound channels'; borrowed chords from the parallel minor with the main melody running I - vi - I - bVI7 - bVII7 - I; common-TONE modulation E major to G major anchored on B natural; and common-CHORD modulation advanced by using borrowed chords as the pivot. Grounds ECON-05, HARM-01, HARM-06, HARM-08. https://www.gamedeveloper.com/audio/7-music-theory-lessons-from-final-fantasy-vii
'13 Essential Chord Progressions from Game Music' (Composer Code). Documents IV-V-bVI-bVII in all major chords taking 'the triumph feeling to 11' (SMB victory); bIII chromatic mediant as 'dark without sounding sad' (Metroid Brinstar); Cm9 to Abmaj7#11 as the harmonic core of DKC 'Aquatic Ambience'; Dorian i-VII-VI-VII as 'less melancholic' fantasy/medieval color (Oblivion); chromatically descending bass for pastoral sweetness (Lon Lon Ranch); harmonic-minor color in Gerudo Valley; minor ii-V-i in Astral Observatory. Grounds HARM-03, HARM-05, HARM-07, HARM-09. https://composercode.com/13-essential-chord-progressions-from-game-music/
Uematsu, Nobuo — Red Bull Music Academy interview (2014). No formal training ('I've never studied music... I simply wanted to make music by myself'); constraints as a creative engine — 'I think that the more limited people are, the more ingenious they begin to get'; NES-era technique of layering slightly detuned frequencies and shaping fade timing for emotional depth; place themes drawn from his rural Shikoku childhood plus Celtic/Tolkien RPG conventions; a stated preference for love themes and 'a magnificent main theme' over battle music. Grounds ECON-09. https://daily.redbullmusicacademy.com/2014/10/nobuo-uematsu-interview/
Mitsuda, Yasunori — composer interviews (shmuplations 2003/2000; corroborated in Music of Chrono Trigger). The Super Famicom's limited sound set 'meant you could focus more on melody' and fit his strengths; process is to reread the scenario repeatedly until the music the game is asking for arrives; the Chrono Trigger main theme's A-section (heroic, driving) versus B-section (sweeping, melancholic) is a deliberate mirror of the game's dual tone. Grounds AFF-06, ECON-09. https://shmuplations.com/yasunorimitsuda/
Greatest Game Music, Chrono Trigger soundtrack analysis. 'Secret of the Forest' uses NON-FUNCTIONAL HARMONY to conjure a hazy, dreamlike atmosphere while balancing melodic writing against atmospheric suggestion; 'Wind Scene' uses high divided violins, a rare 16-bit orchestrational touch, for an ethereal unsettled mood; 'Corridors of Time' and 'Schala's Theme' bring non-Western instruments in for mysticism; and Mitsuda's thematic method is to substantially reshape a melody, quote only FRAGMENTS of it, or write new material around it (Morning Glow, Lavos Theme, World Revolution). Grounds HARM-04, LEIT-07, AFF-05. https://www.greatestgamemusic.com/soundtracks/chrono-trigger-soundtrack/
Video Game Music Daily, 'What Makes it Memorable? — Corridors of Time.' The first half is built from easy notes returning often to the same two high and low pitches within a low range; the ostinato is 'mystic, ethereal bells playing four notes in a series of four bars, each beginning with a slightly higher note than the next'; longer repeated bass notes create emphasis; the second half swaps guitar for a female choral line for contrast in instrument, scale and pacing. Hooktheory reports the track as F# minor, 112 BPM, 4/4, chord complexity 87 (inverted, extended, seventh and parallel-mode borrowed chords). Grounds LEIT-02, HARM-06, ECON-04. https://vgmdaily.wordpress.com/2010/11/21/what-makes-it-memorable-chrono-trigger-corridors-of-time-yasunori-mitsuda/
Uematsu, FF6 leitmotif deployment (Final Fantasy Wiki / TV Tropes leitmotif inventory, corroborated by VGMPF). Terra's Theme is simultaneously her leitmotif and the World of Balance overworld theme, worked into Opening Theme, Metamorphosis, Save Them!, and Ending Theme, with a variation in the third movement of Omen, a softer derivative in Awakening, and an up-beat remix as Searching for Friends — and it is REPLACED on the overworld in the World of Ruin (by Dark World, then Searching for Friends). Grounds LEIT-03, LEIT-04. https://finalfantasy.fandom.com/wiki/Terra%27s_Theme
David Wise — VGMO interview on Donkey Kong Country. Composed within the SNES's 64k memory limit, inspired by the Korg Wavestation's WAVE SEQUENCING (short, often single-cycle waveforms arranged in a harmonically pleasing order); no MIDI available, so HEX sub-routines were typed by hand to order the waveforms; DKC2 extended this by sampling a Roland Juno-2 at multiple filter cutoff/resonance settings; truncating single-cycle waveforms to save memory introduced desirable subtle distortion and harmonics. 'Aquatic Ambience' was his first attempt at replicating wave sequences and set his technique for years. Grounds ECON-09. https://vgmonline.net/davidwiseinterview/
'Aquatic Ambience' (Wikipedia, citing Wise and critic Ignatiy Vishnevetsky). Built from eight waveforms sequenced on a Korg Wavestation — 'just taking eight waveforms and played them in sequence'; five weeks to complete; described as using a sophisticated palette of synthesized instruments and futuristic sound effects to create a mood of calm, markedly unlike typical platform themes, and called the 'Eleanor Rigby' of video game music. Grounds HARM-05, AFF-05. https://en.wikipedia.org/wiki/Aquatic_Ambience
David Wise's instrumentation doctrine (Automaton West interview; Mario Wiki composer profile). Wise mixes natural environmental sounds with prominent melodic and percussive accompaniment, and rather than fixing an instrument set per game he selects instruments that MATCH THE ENVIRONMENT the player is in; his stated DKC2 motivation was the challenge of making good-sounding music inside a very limited memory budget — 'make the SNES sound as pretty as possible' (Bramble Blast). Grounds AFF-05, ECON-09. https://automaton-media.com/en/news/20230516-18913/
'JRPG Song Forms' (Drum Chant, 2021) — form survey of 10 tracks by Uematsu and Hiroki Kikuta. Optional intro (4 of 10) plus a looping section; half the loops under 1:00, half 1:30-2:30; hierarchy of beats/riffs (1-4 bars), phrases (2-6 bars), sections; AAB ternary in six songs and four-equal-section forms elsewhere; two-section loops compensating with more complex phrase construction; a continuous thematic melody chunk of ~30-50 seconds (16 bars at moderate 4/4); ostinato in 8 of 10, held basically throughout in 6; a tag or breakdown to rhythmic hits in all but two; variation by add/remove countermelody, octave shift and instrumentation change; and 'entrances and starts of sections are almost all square and on the beat.' Grounds LOOP-01 through LOOP-06, HARM-06. https://drumchant.wordpress.com/2021/02/14/jrpg-song-forms/
Hunt, Samuel J. 'An Analysis of Repetition in Video Game Music.' CSMC/MuMe 2020 (AI Music Creativity), UWE Bristol. Corpus of 21,391 pieces across consoles. The hypothesis that early games carry above-average repetition because of hardware limits was MOSTLY DISPROVEN — repetition structure stayed more-or-less consistent across platforms and generations; what improved was song length and instrument-track count. Grounds LOOP-10. https://boblsturm.github.io/aimusic2020/papers/CSMC__MuMe_2020_paper_10.pdf
Loop-fatigue practice literature (A Sound Effect immersion guide; Game Developer 'Rethinking the audio loop in games'; DiVA overview of dynamic-music research). The ear starts consciously registering unchanged repetition at roughly 90-120 seconds; practical loop targets are 30-60s for background and 15-30s for combat; loop length should meet or exceed typical per-session gameplay time in that state; linear ABA' loop structures and cohesively resolved sectional transitions disguise the repeat by 'taking the listener on a journey'. Grounds LOOP-01, LOOP-07. https://www.asoundeffect.com/game-audio-immersion/
'Video Game Music: Ostinato and Repetition' (pillowmath). Ostinato defined as repetition of a musical motif, the smallest recognizable form of musical idea; memory limits forced short loops, and self-similar repetition-heavy writing prevents the player noticing the loop point and breaking immersion; four functions identified — driving force, emotional tone, atmospheric immersion, structural transition — with textural treatments (rhythmic, legato, staccato) producing distinct effects. Named cases include Wise's legato synth ostinato in DKC2 'Forest Interlude' for trance-like transitional ambience. Grounds HARM-06, LOOP-06. https://pillowmath.github.io/musings/ostinato-repetition.html
'Why is classic video game music better?' (Signifying Sound and Fury, 2015). Older composers had effectively one tool — melody — so they HAD to write melodies, whereas modern scoring can retreat into atmosphere; but the piece's central corrective is that memorability came from the ACCOMPANIMENT as much as the tune: had the composers written only the melody without orchestrated support, the themes would not be as well remembered. Deliberate choices about melodic RANGE, short quick notes suited to the synths, and timbral choices where available are named as the craft. Grounds ECON-01, ECON-08.
---
3. The measurable features
hook_range_semitones — semitone span of the full hook on the lead track. Target <=17 (octave + P4). Fails ECON-04.
cell_range_semitones — semitone span of the declared identity cell. Target <=9 (M6). Fails ECON-04.
cell_length_events / cell_length_bars — pitch count and bar span of the identity cell. Target <=7 pitches and <=2 bars. Fails LEIT-02.
conjunct_ratio — fraction of melodic intervals <=2 semitones on the lead. Target >=0.60. Fails ECON-05.
signature_leap_count_per_8bar — count of melodic intervals >=7 semitones per 8 bars, plus the declared signature interval and how many phrases it recurs in. Target 1-2 leaps; signature recurs in >=2 phrases. Fails ECON-05.
distinct_durations_in_hook — count of distinct note-duration classes in the hook. Target <=4. Fails ECON-07.
contour_answer_ratio — fraction of adjacent phrase pairs whose net contour signs are OPPOSITE, plus whether the answering phrase is longer. Target >=0.50 opposite, >=1 longer-answer pair. Fails ECON-06.
sentence_parse — does the theme parse as [basic idea / repetition / continuation-by-fragmentation / cadence] over 8 bars; emit the parse tree. Fails ECON-10.
cadence_inventory — counted cadence types across the cue: half cadences, deceptive (V-vi), PACs. Target >=1 HC and 0 PAC before loop exit. Fails ECON-10, HARM-08.
naked_melody_recognition — blind critic recognition rate on a melody-only flat-timbre render, with no harmony or reverb. Hard gate; must pass before orchestration spend. Fails ECON-01.
piano_reduction_pass — does the cue survive reduction to single-staff melody + bass and remain recognizable, BEFORE instrument assignment. Fails AFF-05.
four_voice_reduction_recognition — is the 4-voice reduction of the full arrangement still recognizable as the same cue. Fails ECON-02.
register_overlap_ratio — octave-band overlap between lead and counter as a fraction of sounding note-time. Target <0.25. Fails ECON-03.
block_pad_ratio — fraction of cue duration carrying sustained block-chord pad (the orchestral-wash detector). Target <0.20. Fails ECON-03.
contrary_motion_ratio — fraction of lead/counter simultaneous motion events moving in opposite directions. Target >=0.30 overall; >=4 consecutive bars sustained for wonder-tagged cues. Fails ECON-08, AFF-02.
restricted_palette_compliance — was the theme of record authored under a declared restricted palette (<=5-7 pitch classes for diegetic themes, <=4 voices always) and archived. Fails ECON-09.
borrowed_chord_density — count and per-bar rate of bVI, bVII, bIII (and other parallel-minor borrowings) in major-key cues. Target >=1 per cue. Fails HARM-01.
cadence_family_detector — regex over the roman-numeral reduction for I-bVII-bVI-V (Zelda) and bVI-bVII-I (Mario). Target >=1 hit in every hero/arrival/fanfare cue. Fails HARM-02.
chromatic_mediant_count — count of chromatic-mediant chord moves, cross-checked against the cue's affect tag (mystery/ancient >=1; grief may be 0). Fails HARM-03.
non_functional_harmony_ratio — fraction of chord changes not explainable by functional roman-numeral syntax (planing/parallel/color moves). Target >=0.30 for wonder/exploration; <=0.15 for hero/arrival. Fails HARM-04.
chord_extension_ratio — fraction of chord events carrying 7ths/9ths/#11. Target >=0.40 in ambient cues. Paired with melody_on_extension_ratio (hook notes landing on the extension tone) — target <=0.20. Fails HARM-05.
ostinato_coverage_ratio / ostinato_harmony_count — longest repeated >=2-bar figure as a fraction of loop duration, and the number of distinct harmonies occurring beneath it unchanged. Target >=0.60 coverage and >=2 harmonies for exploration/ambient. Fails HARM-06.
bass_device_flags — pedal-bar percentage and chromatic-descent detection on the bass track. Every wonder or pastoral cue must fire one. Fails HARM-07.
modulation_ledger — count of key changes with pivot type annotated (common-tone / common-chord / borrowed-chord pivot). Fails HARM-08.
mode_map — per-section mode classification (Dorian / Aeolian / Ionian-with-mixture / harmonic-minor color / Mixolydian) plus mode-change event count, checked against the region and affect tables. Fails HARM-09.
major_minor_ambiguity_score — classifier confidence that the hook is major or minor. Target 0.45-0.65 (deliberately AMBIGUOUS) for nostalgia-register cues; high confidence is the failure mode. Fails AFF-01.
melancholy_device_flags — presence of {borrowed-minor reharmonization over an unchanged melody, mode-lowered restatement}. Target >=1 for melancholy-tagged cues; reverb/tempo deltas alone do not count. Fails AFF-03.
composed_tempo_events — count of written rit./accel. events in the tempo map (mix-side tempo changes do not count). Fails AFF-04.
instrument_culture_match — instrument-family assignment checked against the per-culture/period table, and confirmation that assignment happened AFTER piano_reduction_pass. Fails AFF-05.
main_theme_duality — does the main theme parse into >=2 sections with opposed affect tags plus a measurable contrast in tempo-feel, density or mode. Fails AFF-06.
motif_share_rate — fraction of hero/canonical themes sharing a >=4-event ordered pitch-interval cell with the main theme at any transposition. Target >=0.70. Fails LEIT-01.
derivation_strategy_tag — every derived theme carries one of the five sanctioned strategies (eponymous omission / motivic networking / thematic hybridization / associative troping / double main theme). Missing tag fails LEIT-01.
family_id_coverage — every canonical theme has exactly one family_id; each of the 3-6 families has >=2 members sharing a family cell distinct from the main-theme cell. Fails LEIT-06.
transformation_type_count — per motif, count of distinct transformation axes used across the corpus from {rhythm, mode, tempo, orchestration} while the interval spine is preserved. Target >=3 for hero motifs. Fails LEIT-03.
intra_cue_spine_stability — within a single cue, is the motif's interval spine unmodified (change carried only by accompaniment/harmony/instrument). Fails LEIT-05.
silence_window_ledger — per hero motif, longest contiguous absence in chapters and the named return node. Target >=1 window of >=5 chapters with a scored return. Fails LEIT-04.
reprise_quote_length / reprise_quote_share — at reprise nodes, quoted-cell length in events (target 2-4) and quoted material as a share of the cue's melodic note count (target <0.25). Fails LEIT-07.
thematic_chunk_duration_s — duration of the contiguous span containing the hook. Target 30-50s. Fails LOOP-02.
form_label — emitted section map reduced to a form label; must be AAB, AABA' or four-equal-sections, else a documented phrase-complexity justification. Fails LOOP-03.
variation_delta_types / harmony_delta_between_repeats — which of {add-remove counter, octave shift, instrumentation change, register change} fired between repeats (target >=1) and the harmonic difference between repeats (target 0). Fails LOOP-04.
break_present — a bar range of reduced texture or rhythmic-hit writing located in the last third of the loop. Fails LOOP-05.
seam_metrics — section starts quantized to downbeats (boolean), pitch/velocity/RMS discontinuity magnitude at the loop point, and blind seam-detection failure rate (higher is better). Fails LOOP-06.
longest_unchanged_block_s — longest span with no LOOP-04 variation event. Target <90s. Fails LOOP-07.
exploration_block_inventory — for field cues: count of interchangeable ~8-bar / ~15s blocks (target >=6), presence of a chord-compatibility matrix, and count of player-state variants (target >=2, e.g. idle-calm and enemy-heroic). Fails LOOP-08.
interactive_hook_declared — each cue names >=1 interactive musical idea authored into the score (timed tempo shift, depth-linked bass, underwater variant, state-shifted melody). Fails LOOP-09.
repetition_ratio_vs_exemplar_band — self-similarity/repetition ratio of the candidate compared against the exemplar-class band; being BELOW the band (over-through-composed) is a failure, not a virtue. Fails LOOP-10.
max_exemplar_ngram_match — longest contiguous transposition-invariant match of (pitch-interval, duration-ratio) pairs against the exemplar corpus. Hard reject at >=6 events. Cadence formulas, modes, ostinato figures and forms are excluded from the match set. Fails TOOTH-01.
ab_win_rate_vs_incumbent and ab_win_rate_vs_exemplar_anchor — per candidate, blind A/B win rates against the current site theme (the losing control) AND against an exemplar-class anchor, with per-rule rubric scores attached. Proxy-only scores (distinctness, motif recurrence, hummability proxy) may not be reported as success on their own. Fails TOOTH-02.
Generated by harness/site/structure_site.py — the URL path is the repo path. review root