37 rules

Multiple voices on a shared stave

How do two independent voices share one staff without becoming unreadable?

Once two voices share a stave, stem direction stops being a function of pitch and becomes a function of identity: the upper voice stems up wherever it sits. From that follow the rules for beams that must not mix voices, rests that displace by role, and noteheads a second apart that must offset rather than overlap.

Breaks the readingCheckable from the score alone
A beam group must never contain notes from two different voices, even when they share a stave and align rhythmically
A beam is a single voice's rhythmic grouping. Two voices that happen to move in identical rhythm on a shared stave (common in homorhythmic chorale writing) are still beamed SEPARATELY — each voice gets its own beam object, even if the two beams end up visually parallel and close together.
Breaks the readingCheckable from the score alone
A beam must never cross a barline, even when the underlying beat-group run is otherwise legitimate
Beat grouping (which notes get beamed together within a bar) follows the time signature; no beam may extend across a barline into the next measure, regardless of how the notes group rhythmically.
Breaks the readingNeeds the rendered page
A down-stem voice's beam sits BELOW its noteheads even when those noteheads are high on the staff — the direct visual consequence of the stem-role fixation
Beams attach at the stem tip, so a voice locked to down-stems draws its beam below the notehead row for that group, regardless of how high those notes actually sit on the staff. A reader must not interpret 'beam below' as 'these are low notes' on a shared stave — it means 'this is the down-stem voice,' which can be sounding above the up-stem voice at that moment (see voice-crossing-definition-and-tolerance).
Breaks the readingCheckable from the score alone
A forced down-stem voice keeps its down stem even when its notes rise to the middle line or above it — this is the whole point of the rule, and it is not optional at that boundary
Single-voice notation switches stem direction at the middle line (notes on or above the middle line stem down by default, notes below stem up). Shared-stave writing OVERRIDES this: once a voice is assigned the down-stem role for a passage, its notes keep stemming down even when they climb to or above the middle line — exactly the region where an unforced note would normally flip to up-stems. The forced role wins over the pitch default in every case; there is no pitch height at which the role assignment lapses.
Breaks the readingCheckable from the score alone
Beam.generateBeams must always receive `groups` (meter-correct beat grouping) and, for any voice with a forced stem, `maintain_stem_directions`
A renderer-invariant guard, not a per-lesson authoring rule: the call site that generates beams must always pass the meter's beat groups (or beaming falls back to 4/4 in every time signature) and must preserve a forced stem direction (or a shared-stave voice's up/down role assignment is silently overwritten).
Breaks the readingCheckable from the score alone
Cramming three or four independent voices onto a SINGLE stave (not the SATB two-and-two split) is almost never correct engraving
A single stave holding 3-4 concurrently-notated voices — as opposed to the SATB convention of splitting them 2-and-2 across two staves — produces stacked opposing-stem chords with no reliable way to tell which notehead belongs to which voice, since VexFlow/MusicXML stem-direction and beam conventions only cleanly disambiguate TWO roles (up/down) per stave. Gould states short-score (2-voices-per-stave) writing as the practical ceiling for a single stave; anything beyond that belongs on separate staves.
Breaks the readingCheckable from the score alone
Dynamic markings are hard-coded to render below the staff for every voice, in BOTH renderers — wrong for the upper voice of a shared pair
On a stave shared by two voices, a dynamic (or expression) marking belonging to the upper (stem-up) voice is conventionally placed ABOVE the staff, and one belonging to the lower (stem-down) voice is placed BELOW — the opposite of single-voice writing, where dynamics always go below regardless of stem direction. This positional convention is what tells the reader which voice a shared-stave dynamic belongs to.
Breaks the readingCheckable from the score alone
Identical simultaneous rests across two voices sharing a stave must render as ONE shared rest, not two displaced ones
When two voices sharing a staff have a rest of the same duration starting at the same beat, engraving convention draws a single rest (centered, default position) shared by both voices rather than two separately-displaced rest glyphs.
Breaks the readingNeeds both
Noteheads a step (second) apart on a shared stave must be horizontally offset — the HIGHER-PITCHED note of the pair moves right, the lower stays put, and Gradus verifies the grouping but not the side
When two voices sharing a staff sound pitches a second apart at the same onset, the noteheads must be offset left/right so they don't overlap or read as a single notehead with two stems. The offset is decided by momentary PITCH order, not by fixed voice role: the higher-sounding of the two notes shifts to the right (away from the shared stem column), the lower-sounding note keeps the normal (left) position. Because the JSX renderer's forced stem role is fixed to voice identity while pitch order can invert during a voice crossing (see voice-crossing-definition-and-tolerance), the side that shifts right can belong to EITHER the up-stem or the down-stem voice depending on the instant — it tracks pitch, not role.
Breaks the readingCheckable from the score alone
Public-API lyrics carry no verse/voice distinction — two voices with different text on a shared stave will collide into the same lyric line
When two voices sharing a stave each carry their own lyric text (e.g. a hymn's melody and a descant, or two chorale parts each texted independently), MusicXML distinguishes them with a `<lyric number="N">` verse attribute, which Verovio (and every other MusicXML reader) uses to stack multiple simultaneous lyric lines vertically below the stave, one row per verse number, so each voice's text stays legible and distinct from the other's.
Breaks the readingCheckable from the score alone
The pitch-based (unforced) stem-direction fallback has TWO confirmed bugs: the treble middle-line threshold is a third too low, and chord stem direction uses only the highest note instead of whichever extreme is farthest from the middle line
Two independent conventions govern stem direction when a voice is NOT role-forced (a lone voice on its own stave — open score, stacked, grand-staff, or the odd voice in a 3-voice SATB texture — falls back to this path per `resolveVoices`'s `stem: null`). First: single-note stem direction switches exactly at the staff's middle line — notes AT or ABOVE the middle line stem down, notes BELOW stem up. The middle line is B4 in treble clef (`pitchToDiatonic('B4') = 34`) and D3 in bass clef (`pitchToDiatonic('D3') = 22`). Second: for a CHORD, stem direction is decided by whichever notehead is FARTHEST from the middle line (the extreme that would need the longer stem), not simply the chord's highest note — a chord with a note just above the line and another note far below it stems by the far note, not the near one.
Breaks the readingCheckable from the score alone
The public Notation API lets a caller put all 4 voices on ONE instrument/stave with no auto-split and no warning — the exact 'three or four voices on one stave' fault this chapter says is almost always wrong
Gradus's own JSX renderer never lets this happen: `resolveVoices()` structurally caps any 3-4-voice block at 2 voices per stave via `staveIdxOf(i) = i < 2 ? 0 : 1`. The public Notation API's `InstrumentInput.voices: VoiceLine[]` has no equivalent — an agent (or Sean, hand-authoring a request) can supply `voices: [{voice:1,...},{voice:2,...},{voice:3,...},{voice:4,...}]` under ONE `instruments[]` entry, which is one staff, and jsonToScore.ts will happily parse, distribute, and merge all four voice lines onto that single staff with no split, no warning, and no error.
Breaks the readingNeeds both
Three or four genuinely independent beamed voices cannot be engraved on two staves — open score is structurally required
An SATB grand staff assumes at most two voices per staff and works because chorale voices are homorhythmic. Fugal writing with 3-4 independent, continuously beamed voices cannot be split 2-and-2 onto two staves without at least one of the two pairs hitting the lattice problem, because with 3-4 voices, at least one pair is forced to share a stave.
Breaks the readingCheckable from the score alone
Unison notes of differing duration or dot-count across two voices must NOT be allowed to render at the same position
When two voices sharing a stave sound the identical pitch (true unison) but with different note durations or different dotting, the two noteheads must be kept visually distinct (offset), never stacked directly on top of each other.
Breaks the readingCheckable from the score alone
Whole notes have no stem at all — a same-pitch, same-duration whole-note unison across two voices is the one case where the notehead offset is the reader's ONLY cue that two voices exist, and Gradus does not apply it
For any beamed or stemmed duration, an opposing-stem unison (see opposed-stem-unison-same-duration) still gives the reader a fallback cue even when noteheads coincide exactly, because the two stems point in different directions and can be told apart by that alone. A whole note (`w`, `w.`, `w..`) has NO stem in traditional notation — VexFlow renders it as a bare open oval. If two voices share a stave and both sound the identical pitch as a whole note in the same bar, there is no stem, no flag, and (per opposed-stem-unison-same-duration's current, unfixed state) no offset either — nothing at all distinguishes 'two voices happen to coincide' from 'this is simply one note.' The reader has no way to know a second voice is even present.
Misleads the readerCheckable from the score alone
A rest must not cross a beat boundary unless it starts on one and fills whole beats (or is a whole-bar rest)
Rest durations and placement must keep the metric beat visible — a rest that starts mid-beat and continues past the next beat boundary hides where beat 2, 3, etc. fall.
Misleads the readerNeeds the rendered page
A whole/half notehead a second apart from the opposite voice needs a WIDER horizontal offset than a quarter notehead, because the open notehead glyph is wider and has no stem to visually break up the collision
Because whole and half noteheads (open, no stem for whole notes) are physically wider glyphs than filled (quarter/eighth) noteheads, and a whole note additionally has no stem line to help the eye separate two close voices, the horizontal clearance needed between two seconds-apart noteheads scales with notehead width, not a single fixed offset tuned for the more common filled-notehead case.
Misleads the readerNeeds both
Beamed-note lattice on a shared stave: diatonic (not chromatic) distance with a persistence gate
Two voices sharing a stave, both running beamed (sub-quarter) notes with opposing stems, in a close enough tessitura and for long enough, produce a visual lattice where the upper voice's beam and the lower voice's beam interleave through the middle of the staff — a collision that can occur even with ZERO simultaneous pitch crossings, because it is a product of the two voices' PITCH RANGES over a stretch of time, not their pitch order at any one instant.
Misleads the readerNeeds both
Non-identical rests on a shared stave are displaced vertically by voice role, by roughly one staff space, clear of the opposite voice's noteheads
When two voices sharing a staff do NOT have identical simultaneous rests, each voice's rest is displaced away from the shared middle-line default position — the upper (up-stem) voice's rest raised by about one staff space, the lower (down-stem) voice's lowered by about one staff space — enough to clear the other voice's noteheads and stems at that beat. This one-space default is the convention shared across major engraving programs' 'two-voice rest offset' tables (Finale and Sibelius both default to a one-space shift for quarter, eighth, and sixteenth rests in two-voice contexts; whole and half rests shift a comparable amount), not one specific numeric figure asserted by a single treatise.
Misleads the readerNeeds the rendered page
Opposing stems on a shared stave must not visually collide or overlap when their notes sit close together — stem length may need extending, not just fixed direction
When two voices' noteheads sit close in pitch (a third or closer) with opposing stem directions, the default stem length (typically one octave, i.e. 3.5 staff spaces) can leave the two stems' free ends nearly touching or the beams/flags at those ends overlapping. Standard engraving practice extends the stem of one or both notes slightly beyond the default length so the two stems, and anything attached to their ends (flags, beams, articulation marks), keep clear of each other.
Misleads the readerCheckable from the score alone
Public-API hairpins, pedal marks, and octave-shifts share the dynamic-placement defect — all hard-coded to one side regardless of voice
Like dynamics, a crescendo/decrescendo hairpin, piano pedal mark, or octave-shift bracket belonging to the upper voice of a shared pair should render on the side that disambiguates it from the lower voice's marks (hairpins conventionally track the dynamic they're paired with; pedal marks are a keyboard-specific convention that arguably belongs below regardless of voice, but should still be reviewed per case).
Misleads the readerNeeds both
SATB short score (voices 1+2 on treble, 3+4 on bass) is valid exactly when both same-stave pairs stay homorhythmic and non-lattice — stated as one composite boundary
Gradus's SATB grand-staff default (voices 1+2 share the treble stave with opposed stems, voices 3+4 share the bass stave with opposed stems) is the correct convention for chorale-style four-part harmony, and stops being correct exactly at the boundary defined by the preceding rules taken together.
Misleads the readerCheckable from the score alone
Shared-stave stem-role inversion becomes an engraving problem only when SUSTAINED, not on a single occurrence
An occasional voice crossing on a shared stave (the up-stem voice briefly sounding below the down-stem voice) is normal and doesn't need remediation. SUSTAINED or frequent role inversion makes the fixed up/down stem convention actively misleading — the reader starts to see stems pointing the 'wrong way' relative to pitch often enough that the shared-stave stem convention stops doing its job.
Misleads the readerCheckable from the score alone
Slur curvature is hard-coded (fixed control points), not derived from stem direction — a latent defect, not yet triggered because slurs are voice-1-only
A slur on a shared stave is conventionally curved on the side AWAY from the stems (curving below the noteheads when stems point up, above when stems point down) so it doesn't collide with the stem or beam.
Misleads the readerNeeds both
The public Notation API never emits a `<stem>` element — shared-stave stem-role correctness depends entirely on Verovio's own MusicXML-import defaults, unmeasured and unverified by Gradus
MusicXML's `<stem>up</stem>` / `<stem>down</stem>` element lets an exporter explicitly fix stem direction per note, exactly what the JSX/VexFlow renderer does via `stem_direction` on every forced-role note. Verovio, like any MusicXML consumer, falls back to its own internal default (voice/layer-based, similar in spirit to the JSX convention but not guaranteed identical in every edge case — e.g. a 3-4 voice single-staff instrument, which Verovio was never told to auto-split) when no `<stem>` is present.
Misleads the readerNeeds the rendered page
Tie curvature (StaveTie), unlike slurs, IS built per-voice today — but relies entirely on VexFlow's own default direction, never confirmed against a real shared-stave two-tie case
Like slurs, a tie's curvature on a shared stave should curve away from the stem side so it doesn't collide with the opposite voice's stem or notehead. Unlike the slur code (explicitly voice-1-only), Gradus's tie construction runs for EVERY voice, so this is a live path today, not a latent one.
Falls short of best practiceCheckable from the score alone
A SUSTAINED passage where two voices move in identical rhythm and pitch is marked 'a2' / 'unis.' and reduced toward a single notated line, not repeated as two coincidentally-matching voices bar after bar
The per-onset opposed-stem unison rule above governs an incidental unison — two voices happen to land on the same note for a beat or two. A SUSTAINED passage (a full phrase or more) where two voices move together in lockstep pitch and rhythm is a different situation: standard orchestral/choral practice marks the passage 'a2' (or 'unis.' in choral writing) above the stave and, in many house styles, drops to a single notated line with the marking rather than continuing to duplicate every note in two voices for the passage's duration.
Falls short of best practiceCheckable from the score alone
Accidentals are authored explicitly per note per voice, with no shared carry-forward state — this makes the classic cross-voice accidental-ambiguity fault structurally impossible
Under traditional engraving, an accidental holds for the remainder of the bar for that voice/octave only — it does NOT apply to a different voice sharing the same stave, even at the identical pitch and bar. The classic engraving fault this convention guards against is a reader misreading a first voice's accidental as also applying to a second voice's identical-looking note.
Falls short of best practiceNeeds the rendered page
Articulation placement (staccato, accent, etc.) relies on VexFlow's automatic stem-relative default and should be confirmed, not assumed, per shared-stave voice
Articulations are conventionally placed on the side of the notehead opposite the stem, so two voices sharing a stave with opposing stems naturally get their articulations on opposite sides without explicit authoring.
Falls short of best practiceCheckable from the score alone
Gradus has no cross-staff voice notation — a voice cannot borrow the neighboring stave for a phrase, and this gap must be documented rather than silently produce a wrong-clef or wrong-register rendering
Real engraving occasionally lets a voice's notes render temporarily on the OTHER stave of a grand staff or open score (a beam or stem crossing the gap between staves) when a keyboard line's register briefly crosses over — standard practice per Gould's chapter on cross-staff notation, used to keep a single musical line visually continuous instead of forcing awkward register-driven clef choices.
Falls short of best practiceCheckable from the score alone
Rhythmic independence between two voices is measurable and should gate open-score recommendations, not just pitch proximity
Homorhythmic writing (chorale-style, both voices attacking together most of the time) tolerates close tessitura and even brief beam proximity because the ear and eye track the two lines as one texture. Independent counterpoint (fugue, invention) does not get that benefit, so much lower thresholds of proximity or crossing already read as confusing.
Falls short of best practiceNeeds the rendered page
Two simultaneous, differently-altered pitches a second apart need their accidentals vertically staggered, not just their noteheads horizontally offset
When the seconds-apart offset above is applied, each notehead may still carry its own accidental. If both notes need an accidental at that onset, the two accidental glyphs must be staggered vertically (one further left/higher than the other) so neither one visually reads as belonging to the wrong notehead — a plain left-of-notehead placement for both, with the noteheads already horizontally offset, would put one accidental directly beside the OTHER voice's shifted notehead.
Falls short of best practiceNeeds the rendered page
Two voices sharing a stave need a minimum vertical clearance between their combined note-plus-stem extents, independent of any single collision check
Beyond the specific notehead/rest/stem collision rules above, engraving practice maintains an overall minimum vertical buffer between the LOWEST extent of the upper voice's material (notehead, ledger line, or downward-extending element at that beat) and the HIGHEST extent of the lower voice's material at the same beat — roughly one staff space of daylight in normal engraving, tightening only when the specific offset rules above have already resolved a direct collision.
Falls short of best practiceCheckable from the score alone
Two voices sounding the same pitch but spelled with different letter names at the same onset can read as a printing error
When two voices sharing a stave sound the same actual pitch but are spelled differently (e.g. F#4 in one voice, Gb4 in the other) at the same instant, a reader can mistake the mismatch for a copying error rather than a deliberate enharmonic choice, unless the harmonic context makes the differing spelling clearly purposeful.
Falls short of best practiceNeeds both
Unison notes of IDENTICAL duration across two voices: the majority default is ONE shared notehead with opposed stems — Gradus should deliberately override that default, not accidentally fall short of it
When two voices sharing a stave sound the identical pitch with the identical duration, the convention used by Gould as the primary recommendation, and implemented as the automatic default in Dorico, Finale, and Sibelius alike, is a SINGLE shared notehead carrying two stems, one up and one down ('double-stemmed unison') — the two voices are understood to coincide for that instant and the reader is not asked to parse two overlapping heads. A secondary, less common convention — used by some house styles, and by Ross as an alternative worth naming, specifically WHEN the engraver wants the two independent parts to stay visually traceable even through a momentary unison (e.g. a species-counterpoint or chorale exercise where the point of the exercise is seeing two lines) — keeps two separate, very slightly offset (canted) noteheads instead.
Falls short of best practiceCheckable from the score alone
Voice crossing (simultaneous pitch-order inversion between adjacent voices) is normal in moderation and must not be flagged as an error on sight
Voice crossing occurs when two adjacent voices (e.g. Alto and Tenor) invert their expected pitch order WITHIN a single simultaneity — the lower-numbered voice sounds below the higher-numbered one at that instant. This is a real, accepted compositional device and is distinct from a shared-stave stem-role inversion, which is the engraving-visual symptom of the same event.
Falls short of best practiceCheckable from the score alone
Voice overlap (a voice moving beyond an adjacent voice's PREVIOUS position) is a voice-leading concept, not a shared-stave engraving hazard
Voice overlap occurs when a voice moves to a pitch beyond where an ADJACENT voice was on the immediately preceding onset (not simultaneous with it) — e.g. Alto's new note is higher than Soprano's prior note, even though Alto and Soprano don't cross at any single instant. This differs from crossing (a within-simultaneity phenomenon) and has no direct rendering consequence on a shared stave — it doesn't itself cause a notehead, stem, or beam collision.

Loading…

A beam group must never contain notes from two different voices, even when they share a stave and align rhythmically

A beam must never cross a barline, even when the underlying beat-group run is otherwise legitimate

A down-stem voice's beam sits BELOW its noteheads even when those noteheads are high on the staff — the direct visual consequence of the stem-role fixation

A forced down-stem voice keeps its down stem even when its notes rise to the middle line or above it — this is the whole point of the rule, and it is not optional at that boundary

Beam.generateBeams must always receive `groups` (meter-correct beat grouping) and, for any voice with a forced stem, `maintain_stem_directions`

Cramming three or four independent voices onto a SINGLE stave (not the SATB two-and-two split) is almost never correct engraving

Dynamic markings are hard-coded to render below the staff for every voice, in BOTH renderers — wrong for the upper voice of a shared pair

Identical simultaneous rests across two voices sharing a stave must render as ONE shared rest, not two displaced ones

Noteheads a step (second) apart on a shared stave must be horizontally offset — the HIGHER-PITCHED note of the pair moves right, the lower stays put, and Gradus verifies the grouping but not the side

Public-API lyrics carry no verse/voice distinction — two voices with different text on a shared stave will collide into the same lyric line

The pitch-based (unforced) stem-direction fallback has TWO confirmed bugs: the treble middle-line threshold is a third too low, and chord stem direction uses only the highest note instead of whichever extreme is farthest from the middle line

The public Notation API lets a caller put all 4 voices on ONE instrument/stave with no auto-split and no warning — the exact 'three or four voices on one stave' fault this chapter says is almost always wrong

Three or four genuinely independent beamed voices cannot be engraved on two staves — open score is structurally required

Unison notes of differing duration or dot-count across two voices must NOT be allowed to render at the same position

Whole notes have no stem at all — a same-pitch, same-duration whole-note unison across two voices is the one case where the notehead offset is the reader's ONLY cue that two voices exist, and Gradus does not apply it

A rest must not cross a beat boundary unless it starts on one and fills whole beats (or is a whole-bar rest)

A whole/half notehead a second apart from the opposite voice needs a WIDER horizontal offset than a quarter notehead, because the open notehead glyph is wider and has no stem to visually break up the collision

Beamed-note lattice on a shared stave: diatonic (not chromatic) distance with a persistence gate

Non-identical rests on a shared stave are displaced vertically by voice role, by roughly one staff space, clear of the opposite voice's noteheads

Opposing stems on a shared stave must not visually collide or overlap when their notes sit close together — stem length may need extending, not just fixed direction

Public-API hairpins, pedal marks, and octave-shifts share the dynamic-placement defect — all hard-coded to one side regardless of voice

SATB short score (voices 1+2 on treble, 3+4 on bass) is valid exactly when both same-stave pairs stay homorhythmic and non-lattice — stated as one composite boundary

Shared-stave stem-role inversion becomes an engraving problem only when SUSTAINED, not on a single occurrence

Slur curvature is hard-coded (fixed control points), not derived from stem direction — a latent defect, not yet triggered because slurs are voice-1-only

The public Notation API never emits a `<stem>` element — shared-stave stem-role correctness depends entirely on Verovio's own MusicXML-import defaults, unmeasured and unverified by Gradus

Tie curvature (StaveTie), unlike slurs, IS built per-voice today — but relies entirely on VexFlow's own default direction, never confirmed against a real shared-stave two-tie case

A SUSTAINED passage where two voices move in identical rhythm and pitch is marked 'a2' / 'unis.' and reduced toward a single notated line, not repeated as two coincidentally-matching voices bar after bar

Accidentals are authored explicitly per note per voice, with no shared carry-forward state — this makes the classic cross-voice accidental-ambiguity fault structurally impossible

Articulation placement (staccato, accent, etc.) relies on VexFlow's automatic stem-relative default and should be confirmed, not assumed, per shared-stave voice

Gradus has no cross-staff voice notation — a voice cannot borrow the neighboring stave for a phrase, and this gap must be documented rather than silently produce a wrong-clef or wrong-register rendering

Rhythmic independence between two voices is measurable and should gate open-score recommendations, not just pitch proximity

Two simultaneous, differently-altered pitches a second apart need their accidentals vertically staggered, not just their noteheads horizontally offset

Two voices sharing a stave need a minimum vertical clearance between their combined note-plus-stem extents, independent of any single collision check

Two voices sounding the same pitch but spelled with different letter names at the same onset can read as a printing error

Unison notes of IDENTICAL duration across two voices: the majority default is ONE shared notehead with opposed stems — Gradus should deliberately override that default, not accidentally fall short of it

Voice crossing (simultaneous pitch-order inversion between adjacent voices) is normal in moderation and must not be flagged as an error on sight

Voice overlap (a voice moving beyond an adjacent voice's PREVIOUS position) is a voice-leading concept, not a shared-stave engraving hazard