49 rules

Expression marks and attachments

Where do dynamics, hairpins, articulations and ornaments belong on the page?

Everything attached to a note rather than sounded by it: dynamics and their vocabulary, hairpins and their minimum span, articulations and their relationship to stem direction, ornaments, and repeat structures. Placement here is not decoration — it decides which performer a mark applies to and which note it modifies.

Breaks the readingCheckable from the score alone
A fermata over a rest must render — Model A currently drops it silently
A fermata over a rest (holding a silence, very common at the end of a phrase or before a final chord) is standard notation.
Breaks the readingCheckable from the score alone
A hairpin must span at least two distinct notes in the correct direction
A crescendo/diminuendo wedge needs a start point and a later end point; a wedge with the end at or before the start is not a hairpin.
Breaks the readingNeeds both
A hairpin, slur, or tuplet must not be silently dropped across a system break — replace or split it
A hairpin that would need to cross a system break should either be replaced with the text convention ("cresc."/"dim." plus continuation dashes) or, if drawn as a wedge, both system-line halves must be drawn (the wedge stops at the right margin of the first system and resumes at the left margin of the next).
Breaks the readingCheckable from the score alone
A tie or slur must not cross a repeat barline
A tie or slur written across a repeat barline is ambiguous: on the first pass it connects to the material after the repeat sign, but on the repeated pass it would need to connect back to the start of the repeated section — a different note. Standard practice re-articulates instead (no tie) or uses a different device (a footnote, or simply not tying).
Breaks the readingCheckable from the score alone
D.C./D.S./coda/segno flags must be internally consistent and point at a real target
Da Capo (return to the beginning) and Dal Segno (return to the segno sign) are mutually exclusive instructions at a single point; "To Coda" requires an actual coda sign to jump to; Dal Segno requires an actual segno sign earlier in the piece.
Breaks the readingCheckable from the score alone
Dynamic and articulations must not be duplicated onto a bar-crossing tie's second half
A tied note that crosses a bar line is one sustained sound; the dynamic and any articulation belong to the note's single onset (the first half) — the tied continuation carries neither.
Breaks the readingCheckable from the score alone
Dynamic text validated against a closed vocabulary on ALL THREE surfaces
A dynamic marking is one of a closed set of conventional abbreviations (ppp…fff, fp, sf, sfz, fz, rfz, sffz per Gould) rendered in bold italic serif type — never arbitrary prose.
Breaks the readingCheckable from the score alone
Every pedal-down must be matched by a later pedal-up before the next pedal-down or the piece's end
A `Ped.` (pedal down) mark must be released with a `✱` (pedal up) mark before a new pedal-down begins, and every pedal engaged must eventually be released.
Breaks the readingCheckable from the score alone
Grace notes must never consume rhythmic/tick budget of their own
Grace notes take no time from the notated meter — they are performed by "borrowing" a sliver of the neighboring note's duration, not by occupying their own beat slot.
Breaks the readingCheckable from the score alone
Model A's tempoMarking prop is declared but never rendered — a second dead field alongside fingering
A tempo indication (e.g. "Andante", "♩ = 96") is placed above the topmost stave of the first system, in bold plain (non-italic) text, distinct from expressive text.
Breaks the readingCheckable from the score alone
No fresh-attack articulation on the receiving note of a tie
A tie represents one sustained sound with a single attack; the destination note of a tie has no re-attack, so it cannot carry staccato/accent/tenuto/marcato (which all describe how a note is attacked or detached). A fermata IS allowed on a tie's final note, since it marks how long the sound is held, not how it is attacked.
Breaks the readingCheckable from the score alone
No two hairpins overlap or nest in the same voice without resolving in between
A voice has one active hairpin at a time; a second hairpin cannot begin before the first has ended (no overlapping wedges, no nested wedges) — a change of direction mid-swell (cresc. then immediately dim.) is written as two back-to-back, non-overlapping hairpins meeting at a shared point, or as a single "messa di voce" shape which Gradus's schema has no dedicated symbol for.
Breaks the readingCheckable from the score alone
Only fermata is valid on a rest; other articulations are meaningless there, and Model A drops all of them silently
A rest has no attack, so staccato/accent/tenuto/marcato do not apply to it; a fermata over a rest, however, is a standard and common marking (holding a silence).
Breaks the readingCheckable from the score alone
Rehearsal marks must be strictly sequential with no repeats or reversals
Rehearsal letters/numbers (A, B, C… or 1, 2, 3…) increase monotonically through the score with no duplicate and no skipped-then-reused letter.
Breaks the readingCheckable from the score alone
The fingering field is declared but never rendered anywhere
Fingering numerals are a standard attachment placed directly above or below the notehead (above the staff for the right hand/treble convention, below for the left hand/bass, or immediately adjacent for string instruments), used throughout instructional keyboard and string literature — exactly the audience Gradus serves.
Breaks the readingCheckable from the score alone
Volta (first/second ending) brackets must be well-formed and sequentially numbered
A volta bracket that starts must close before the next repeat-end (or before the piece ends); within one repeat group, endings are numbered sequentially starting at 1 ("1.", "2.", …), not skipping or repeating a number.
Misleads the readerCheckable from the score alone
A fermata in an ensemble score must be aligned across every simultaneously-active stave
When one instrument holds a fermata, every other instrument sounding at that same beat holds one too, so the ensemble stops and resumes together; an un-mirrored fermata on only one stave is a coordination error.
Misleads the readerCheckable from the score alone
A fermata over a barline (rather than over a note) marks an ensemble grand pause, not extra length on one note
A fermata drawn directly above a barline (not above any specific notehead) is a distinct, standard marking indicating a brief silent pause affecting the whole ensemble at that point, independent of any single note's duration — different from a fermata over a note or rest, which extends that specific sound. This is also the convention used in printed Bach chorale collections at internal phrase endings, where the fermata over the final note of a phrase does NOT lengthen the note — it marks the phrase boundary while the underlying pulse (tactus) continues, a convention specific to the chorale-harmonization genre Gradus teaches directly.
Misleads the readerCheckable from the score alone
A hairpin should arrive at a stated dynamic
A crescendo/diminuendo wedge is conventionally followed immediately by (or ends exactly at) the dynamic level it was building toward; an open-ended hairpin with no destination dynamic is incomplete.
Misleads the readerCheckable from the score alone
A trill lasting multiple notes gets one tr sign plus a wavy extension line, not a repeated stamp
When a single sustained pitch is trilled across several written note values (tied together), engraving shows the `tr` sign once at the first note and a wavy line (trillo/tr~~~~~) continuing for the duration — never the `tr` glyph repeated on each tied segment.
Misleads the readerCheckable from the score alone
A trill needing a chromatic auxiliary must show the accidental above the trill sign
When the diatonic upper auxiliary of a trill needs raising or lowering to match the harmony (routine in Bach), a small accidental is printed directly above the `tr` sign.
Misleads the readerCheckable from the score alone
Accent and marcato together on one note is non-standard
Accent (`>`) and marcato (`^`) are both single "strong attack" marks of increasing intensity; no authority documents stacking both on the same notehead as a sanctioned combination the way staccato+tenuto=portato is sanctioned.
Misleads the readerNeeds the rendered page
Articulation placement follows stem direction (opposite the stem side)
A staccato dot, accent, tenuto line, or marcato roof sits on the side of the notehead away from the stem — above the notehead when the stem points down, below when the stem points up (so the mark never collides with the stem or beam).
Misleads the readerCheckable from the score alone
Breath marks and caesura have zero representation on any of the three Gradus authoring surfaces
A breath mark (comma-shaped tick above the staff) indicates a brief lift/breath, with the borrowed time typically taken from the preceding note; a caesura ("railroad tracks", //) indicates a brief but essentially complete interruption of sound across the ensemble, more definitive than a breath mark. Both are standard phrasing devices distinct from a rest, a fermata, or a tie.
Misleads the readerCheckable from the score alone
Dynamic placement: below for instrumental, above for vocal-with-text
Dynamics sit below the staff for instrumental parts; for a vocal line with underlaid text, dynamics go above the staff so they are not confused with syllables placed below.
Misleads the readerCheckable from the score alone
Grace note(s) should be slurred to their principal note — currently drawn on neither surface
A grace note or grace group is conventionally connected to its principal note with a small slur, showing the ornament leads melodically into the main note.
Misleads the readerCheckable from the score alone
Model A cannot author an acciaccatura (slashed grace note) — only appoggiatura-style
A slashed grace note (acciaccatura) is played as fast as possible before the beat and takes no perceptible time from the main note; an unslashed grace note (appoggiatura) takes a defined, often substantial, rhythmic value from the main note. The two look different (slash through the stem) and are notationally distinct, not just performance nuance.
Misleads the readerCheckable from the score alone
Model A cannot author an inverted mordent; Model C can
The mordent (lower-neighbor alternation) and the inverted/upper mordent (upper-neighbor alternation) are two distinct, commonly-needed ornaments.
Misleads the readerCheckable from the score alone
Model A cannot author ANY repeat structure beyond a single whole-block repeat pair
Full repeat-structure notation (mid-piece repeat barlines, first/second endings, D.C./D.S./coda/segno) is standard for fugue expositions, Baroque dance forms, and sectional chorale settings — all genres Gradus's curriculum explicitly teaches.
Misleads the readerCheckable from the score alone
Model A cannot author combination articulations (e.g. portato)
Portato (a sustained-but-separated touch) is notated as a staccato dot AND a tenuto line together on one notehead — a legitimate, named combination, distinct from either mark alone.
Misleads the readerCheckable from the score alone
Model B/C has no freeform per-note expression-text field at all — an agent cannot author dolce/espress./cantabile
Expressive text markings (dolce, espressivo, cantabile, semplice, etc.) are a standard, open-ended vocabulary distinct from the closed dynamic and technique vocabularies — a composer/arranger needs to attach arbitrary short italic text to a note.
Misleads the readerNeeds both
Multiple grace notes in one group must be beamed together with an explicit boundary, not left to auto-beam heuristics
When 2+ grace notes precede one principal note, they are beamed together as a single figure; the beam must not extend into or merge with a neighboring grace group belonging to a DIFFERENT principal note.
Misleads the readerCheckable from the score alone
One dynamic level per performer at a given onset
Two voices sharing one staff and one performer (soprano+alto on the treble stave of a chorale reduction, or two hands of one pianist) receive one dynamic at a given beat, not two contradicting ones, unless the passage is explicitly marked divisi/con espressione for independent parts.
Misleads the readerCheckable from the score alone
Pedal marks are completely unreachable through Model A and the agent-facing Model B API
Sustain-pedal notation is a routine, frequently-needed mark for keyboard repertoire — directly relevant to any Gradus keyboard-writing content (chorale realizations at the keyboard, Baroque keyboard fugue).
Misleads the readerCheckable from the score alone
Sudden-accent dynamics (fp, sf, sfz, fz, rfz, sffz) do not redefine the prevailing sustained level
fp/sf/sfz/fz/rfz/sffz are momentary accents on a single note; the underlying sustained dynamic before the accent is what continues afterward (an `fp` conventionally implies the level drops to piano immediately after the loud attack, not that the piece is now "at fp level").
Misleads the readerCheckable from the score alone
The final measure of a piece must end with a final barline, never a plain single barline or a dangling repeat sign
The very last barline of a complete piece (or a complete movement) is always a final/closing double barline (thin-thick), never a plain single barline and never a forward-only repeat sign left unresolved.
Falls short of best practiceNeeds the rendered page
A dynamic under a chord must be centered on the chord's rendered width, not the lowest notehead
A dynamic mark (or any note-attached annotation) is horizontally centered under the full visual width of the notehead(s) it governs — for a chord, centered on the chord's stem/notehead cluster, not left-aligned to whichever pitch the code happens to attach the annotation to.
Falls short of best practiceCheckable from the score alone
A hairpin should begin from an established dynamic, not from silence
A crescendo/diminuendo wedge is read relative to the dynamic level it starts FROM; a hairpin with no established dynamic anywhere before its start point has no baseline to swell from or taper to, which is ambiguous rather than merely informal (contrast with expr-hairpin-no-terminal-dynamic, which covers the destination end).
Falls short of best practiceCheckable from the score alone
Articulation on a slurred (but not tied) note is always valid — do not conflate slur with tie
A slur connects notes into one smooth phrase for BOWING/BREATH purposes, but each note under a slur is still its own attack (a different pitch, or the same pitch re-struck) and may carry its own staccato/accent/tenuto exactly as an unslurred note would — this is the standard "slurred staccato" figure (soft, separated notes in one bow/breath), a distinct and legitimate combination from portato. This is the OPPOSITE convention from a tie (expr-artic-tied-continuation), where the destination note genuinely has no new attack.
Falls short of best practiceCheckable from the score alone
Do not stack a hairpin wedge and the literal word cresc./dim. over the identical span
Use either the graphic wedge or the word (with continuation dashes for a long span) — not both simultaneously over the same notes, which is redundant and visually busy.
Falls short of best practiceCheckable from the score alone
Hairpin direction vocabulary must be identical across authoring surfaces
There are exactly two hairpin directions (crescendo, decrescendo/diminuendo); an audit tool spanning multiple content surfaces needs one canonical vocabulary to compare across them.
Falls short of best practiceNeeds both
Hairpin minimum length
A hairpin needs enough horizontal room to read as a wedge rather than an accent mark (`>`). CORRECTED FROM DRAFT: neither Gould nor Ross states a specific numeric minimum span in beats — Ross's guidance (The Art of Music Engraving, ch. Dynamic Markings) is qualitative ("a hairpin must be long enough to be legible as a gradual change, not confused with an accent"), not a beat count. The 1.0-beat floor below is a Gradus engineering heuristic modeled on that qualitative guidance, not a cited numeric authority rule, and should be labeled as such in any user-facing message.
Falls short of best practiceCheckable from the score alone
Hairpins, slurs, and tuplets are voice-1-only in the JSX renderer — flag any attempt to imply otherwise
In genuine polyphonic/SATB writing, a crescendo can apply to any individual voice, not only the top line.
Falls short of best practiceCheckable from the score alone
No redundant restatement of an unchanged dynamic
A dynamic mark is only written where the level changes (or after a structural break like a rest of significant length or a new system/movement); repeating the identical mark on consecutive notes with nothing intervening is visual clutter.
Falls short of best practiceNeeds the rendered page
Ornament glyphs must clear beams and slurs, not collide with them
A trill/turn/mordent sign sits above the staff clear of any beam passing below it; when a slur also arcs over the same passage, the ornament is placed OUTSIDE (above) the slur's arc, never crossed by it, so the two marks read as visually distinct layers.
Falls short of best practiceCheckable from the score alone
Tempo text is bold plain type; expression text is italic — the two must not be visually or semantically conflated
A tempo marking (Andante, ♩=96) is set in bold, non-italic (roman) type above the top stave; an expression marking (dolce, espress., cantabile) is set in italic type, conventionally above the staff near the note it modifies. The two serve different functions and must be visually distinguishable.
Falls short of best practiceCheckable from the score alone
The model cannot distinguish before-the-beat vs on-the-beat grace-note realization
Historically and in modern practice, some grace-note figures are realized BEFORE the beat (borrowing time from the PRECEDING note/silence, common in some 18th-century keyboard "flick" ornaments and in many modern acciaccatura realizations) while others are realized ON the beat (borrowing from the FOLLOWING principal note, the standard Classical appoggiatura realization). These are different rhythmic placements, not merely different notated shapes.
Falls short of best practiceCheckable from the score alone
Turn placement (over the note vs. inserted after it) is not authorable
A turn can sit directly over a notehead (ornamenting that note itself) or be placed in the space AFTER a note, between it and the next, indicating the turn's figure is inserted between the two written pitches — a real distinction with different rhythmic execution, common in Classical-era ornamentation.
Falls short of best practiceNeeds the rendered page
When multiple mark categories stack on one note, they nest in a fixed radial order outward from the notehead
When a note carries more than one kind of attached mark (articulation, dynamic, ornament, fingering), the conventional stacking order from the notehead outward is: articulation (staccato dot/accent/tenuto — closest, since it modifies the attack itself), then fingering (if present), then dynamic (further out, it is read independently of the attack), with ornaments (trill/turn/mordent) placed above the staff regardless of what's below. Marks must never render at the same vertical offset such that they overlap each other.

Loading…

A fermata over a rest must render — Model A currently drops it silently

A hairpin must span at least two distinct notes in the correct direction

A hairpin, slur, or tuplet must not be silently dropped across a system break — replace or split it

A tie or slur must not cross a repeat barline

D.C./D.S./coda/segno flags must be internally consistent and point at a real target

Dynamic and articulations must not be duplicated onto a bar-crossing tie's second half

Dynamic text validated against a closed vocabulary on ALL THREE surfaces

Every pedal-down must be matched by a later pedal-up before the next pedal-down or the piece's end

Grace notes must never consume rhythmic/tick budget of their own

Model A's tempoMarking prop is declared but never rendered — a second dead field alongside fingering

No fresh-attack articulation on the receiving note of a tie

No two hairpins overlap or nest in the same voice without resolving in between

Only fermata is valid on a rest; other articulations are meaningless there, and Model A drops all of them silently

Rehearsal marks must be strictly sequential with no repeats or reversals

The fingering field is declared but never rendered anywhere

Volta (first/second ending) brackets must be well-formed and sequentially numbered

A fermata in an ensemble score must be aligned across every simultaneously-active stave

A fermata over a barline (rather than over a note) marks an ensemble grand pause, not extra length on one note

A hairpin should arrive at a stated dynamic

A trill lasting multiple notes gets one tr sign plus a wavy extension line, not a repeated stamp

A trill needing a chromatic auxiliary must show the accidental above the trill sign

Accent and marcato together on one note is non-standard

Articulation placement follows stem direction (opposite the stem side)

Breath marks and caesura have zero representation on any of the three Gradus authoring surfaces

Dynamic placement: below for instrumental, above for vocal-with-text

Grace note(s) should be slurred to their principal note — currently drawn on neither surface

Model A cannot author an acciaccatura (slashed grace note) — only appoggiatura-style

Model A cannot author an inverted mordent; Model C can

Model A cannot author ANY repeat structure beyond a single whole-block repeat pair

Model A cannot author combination articulations (e.g. portato)

Model B/C has no freeform per-note expression-text field at all — an agent cannot author dolce/espress./cantabile

Multiple grace notes in one group must be beamed together with an explicit boundary, not left to auto-beam heuristics

One dynamic level per performer at a given onset

Pedal marks are completely unreachable through Model A and the agent-facing Model B API

Sudden-accent dynamics (fp, sf, sfz, fz, rfz, sffz) do not redefine the prevailing sustained level

The final measure of a piece must end with a final barline, never a plain single barline or a dangling repeat sign

A dynamic under a chord must be centered on the chord's rendered width, not the lowest notehead

A hairpin should begin from an established dynamic, not from silence

Articulation on a slurred (but not tied) note is always valid — do not conflate slur with tie

Do not stack a hairpin wedge and the literal word cresc./dim. over the identical span

Hairpin direction vocabulary must be identical across authoring surfaces

Hairpin minimum length

Hairpins, slurs, and tuplets are voice-1-only in the JSX renderer — flag any attempt to imply otherwise

No redundant restatement of an unchanged dynamic

Ornament glyphs must clear beams and slurs, not collide with them

Tempo text is bold plain type; expression text is italic — the two must not be visually or semantically conflated

The model cannot distinguish before-the-beat vs on-the-beat grace-note realization

Turn placement (over the note vs. inserted after it) is not authorable

When multiple mark categories stack on one note, they nest in a fixed radial order outward from the notehead