midiviz

01 — How pianists choose fingerings: rules, traditions, ergonomics

Research brief for designing a fingering algorithm’s cost model and test cases. Finger numbers: 1 = thumb … 5 = little finger, both hands. “RH”/”LH” = right/left hand. Spans are in semitones (st) unless stated. Source keys like [P97] point to the list at the end. Verification: [V] = checked against a fetched source this session; [T] = standard pedagogy (scale books, widely repeated) not individually re-fetched; [?] = plausible, needs checking.


1. Core principles that run through the whole tradition

  1. Next-note-next-finger. Adjacent scale steps go to adjacent fingers. Parncutt et al. trace this to Sancta Maria (1565), C.P.E. Bach (1753) and Türk (1789) [P97][V].
  2. Few hand-position changes. Play as many notes as possible from one position. Documented since Diruta and Sancta Maria in the Renaissance [P97][V].
  3. Short fingers (1, 5) on white keys, long fingers (2, 3, 4) on black. C.P.E. Bach and Türk avoid 1 and to a lesser extent 5 on black keys [P97][V]. Chopin turned this around and treated the long-fingers-on-black shape as the hand’s natural home (§2.6).
  4. Thumb passes are easiest between a finger on a black key and the thumb on a white key. White→white is next, then black→black. Non-thumb on white to thumb on black is “by far the most difficult” and turns up only in virtuoso repertoire (Chopin Op. 25/11) [P97][V].
  5. Weak side of the hand. Finger 4 is the least independent, and 3-4-5 runs are hard. Ortmann (1929): “no amount of practice can overcome entirely this physiological limitation” [P97][V]. In sight-reading data, the 4-3 transition occurs 45% more often than 3-4 [P97][V].
  6. Same fingering for the same pattern. Sequences, repeats and transposed figures get identical fingering so the motor pattern can be reused (Hummel, Czerny; Busoni and von Bülow had pupils practise a piece in other keys without changing the fingering [P97][V]).
  7. The music has the last word. Phrasing, articulation, voicing and tone colour can override ergonomics. Neuhaus held that the best fingering is the one that best serves the music, even if it is not the most convenient [N][T]. Chopin used fingering for expression (§2.6).
  8. Fingering is personal. In the PIG dataset, two pianists agree on only 60–80% of notes, and a human annotator matches the other annotators’ fingerings 71.4% of the time (Mgen) [NSY19][V]. There is no single right answer. Evaluate against several reference fingerings, or score “acceptable” rather than “identical”.

2. Historical treatises (what each contributed)

2.1 François Couperin — L’art de toucher le clavecin (Paris 1716, rev. 1717) [C][T]

2.2 C.P.E. Bach — Versuch über die wahre Art das Clavier zu spielen (Berlin 1753) [CPE][V]

2.3 Daniel Gottlob Türk — Klavierschule (1789) [T]

2.4 Muzio Clementi — Introduction to the Art of Playing on the Piano Forte (1801) [T]

2.5 J.N. Hummel (1828) and Carl Czerny (Op. 500, c. 1839) [H][T]

2.6 Frédéric Chopin — Projet de méthode (unfinished), as reported in Eigeldinger [E][V]

2.7 Liszt, Tausig (and later Busoni) [T]

2.8 Theodor Leschetizky (via Malwine Brée, Die Grundlage der Methode Leschetizky, 1902) [T]

2.9 Modern pedagogy [T]


3. Concrete rules and standard fingerings

3.1 Scale fingerings, one octave ascending (descend in reverse) [WB][T, ABRSM/Clementi standard]

In a multi-octave run, the top-note finger in brackets is replaced by the continuing finger. RH continuity: (5)→1 in the white-key group. LH continuity: the bottom note’s 5 becomes 1 when passing over.

Majors

Key RH LH
C, G, D, A, E 1 2 3 1 2 3 4 (5) 5 4 3 2 1 3 2 1
B 1 2 3 1 2 3 4 (5) 4 3 2 1 4 3 2 1
F 1 2 3 4 1 2 3 (4) 5 4 3 2 1 3 2 1
F♯ / G♭ 2 3 4 1 2 3 1 2 4 3 2 1 3 2 1 4
D♭ / C♯ 2 3 1 2 3 4 1 2 3 2 1 4 3 2 1 3
A♭ 3 4 1 2 3 1 2 3 3 2 1 4 3 2 1 3
E♭ 3 1 2 3 4 1 2 3 3 2 1 4 3 2 1 3
B♭ 2 1 2 3 1 2 3 4 (start 2, continuing finger 4) 3 2 1 4 3 2 1 3

Harmonic minors (melodic minors mostly follow these. F♯, C♯ and B♭ melodic have variants; check an ABRSM scale book.)

Key RH LH
a, e, d, g, c 1 2 3 1 2 3 4 (5) 5 4 3 2 1 3 2 1
b 1 2 3 1 2 3 4 (5) 4 3 2 1 4 3 2 1
f 1 2 3 4 1 2 3 (4) 5 4 3 2 1 3 2 1
f♯ 3 4 1 2 3 1 2 3 4 3 2 1 3 2 1 4
c♯ 3 4 1 2 3 1 2 3 3 2 1 4 3 2 1 3
g♯ 3 4 1 2 3 1 2 3 3 2 1 4 3 2 1 3
e♭ / d♯ 3 1 2 3 4 1 2 3 2 1 4 3 2 1 3 2
b♭ 2 1 2 3 1 2 3 4 2 1 3 2 1 4 3 2

Rules that generate the table (useful as a test oracle) [MIA][V]:

3.2 Chromatic scale [T]

3.3 Arpeggios, two octaves [MIA][V][T]

3.4 Broken chords, accompaniment figures [T]

3.5 Octaves [LP][PTP][W][SE][V]

3.6 Double thirds and sixths [SE][DPH][T]

3.7 Repeated notes [SE][T]

3.8 Trills [T][SE]

3.9 Crossings: which are allowed [P97][T]

Parncutt Table 1 implies (for the RH, with the LH mirrored):

Crossing Allowed range (st, rising = thumb under going up) Use
1 under 2 (2→1 up) up to 5 very common (2-1-2 pivots)
1 under 3 (3→1 up) up to 4 the scale pass
1 under 4 (4→1 up) up to 3 common (scales, 4-note groups)
1 under 5 not in table (MinPrac 1-5 = –1) essentially never
2/3/4 over 1 (descending) same distances standard descending pass
3 over 4, 4 over 5, 2 over 3 not modelled (MinPrac > 0) rare: legato double notes, Chopin chromatic, Couperin-style 3-4 pairs
5 over anything — avoid

3.10 Thumb and 5 on black keys [P97][V]

3.11 Long fingers on black keys [P97][PTP][V]

Fingers 2, 3 and 4 reach black keys without moving the hand forwards. Chopin’s E-F♯-G♯-A♯-B position is the ideal. “4 on a black key after 3 on a white key” is an extra difficulty (Four-on-Black rule) [P97][V].

3.12 Finger substitution [WIKI-FS][V][NSY19][V]

3.13 Hand position, extension and contraction (“position principle”) [P97][S][T]

3.14 Spans per finger pair (Parncutt et al. 1997, Table 1, RH, semitones) [P97][V]

Positive = rising from the first finger to the second. Negative = thumb under or fingers over.

Pair MinPrac MinComf MinRel MaxRel MaxComf MaxPrac
1-2 -5 -3 1 5 8 10
1-3 -4 -2 3 7 10 12
1-4 -3 -1 5 9 12 14
1-5 -1 1 7 10 13 15
2-3 1 1 1 2 3 5
2-4 1 1 3 4 5 7
2-5 2 2 5 6 8 10
3-4 1 1 1 2 2 4
3-5 1 1 3 4 5 7
4-5 1 1 1 2 3 5

3.15 Hand size [PASK][V]

1-5 span Octave (12 st) 9th (13–14 st) 10th (15–16 st) 11th (17 st)
17.0 cm / 6.7” on the edge — — —
19.3 cm / 7.6” minimal comfort on the edge — —
21.6 cm / 8.5” relaxed minimal comfort on the edge —
23.9 cm / 9.4” relaxed relaxed minimal comfort on the edge

3.16 Redistribution between hands, crossing hands, glissando [T]

3.17 Phrasing and articulation [E][S][T]

3.18 Editorial conventions [T][HOOK][V]


4. Ergonomic facts

Fact Value / statement Source
White key width ≈ 23.5 mm at the front; octave 164–165 mm [MK][SE-K][V]
Average semitone 13.7 mm (but irregular: E-F, B-C = one white key) [P97][V]
Black key ≈ 12 mm wide (15/32” at base, 7/16” at top), ≈ 92–100 mm long, top about 6 mm above the white key level in one technician spec (other sources put it at ≈ 12 mm) [SE-K][V][?]
Black keys sit further back the hand must move into the keys (distally) to reach them with 1 or 5, so the thumb on black moves the whole hand inward [P97][F][V]
Key leverage black and white keys have offset fulcra; they sound evenly only if each is played at its optimal point, which is another reason to keep 1 and 5 off black keys (most acute in the bass) [P97][V]
Finger independence thumb, index and little finger are the most individuated; middle and ring the least (Häger-Ross & Schieber 2000) [HRS][V]
3-4-5 coupling juncturae tendinum link the extensor tendons of 3-4-5. Lifting 4 while 3 is down is hard, hence 3→4 is harder than 4→3 [P97][HRS][T]
Finger length 3 > 4 ≈ 2 > 5 > 1 (the thumb is much shorter and works sideways from its carpometacarpal joint, which is what makes thumb-under possible) [T]
Asymmetric transitions (Czerny sight-reading) 4-3 > 3-4 (+45%); also 3-1 > 1-3, 1-4 > 4-1, 3-2 > 2-3, 2-4 > 4-2, 5-4 > 4-5 [P97][V]
Contraction vs extension finger adduction is easier than abduction, so tight spans (below MinRel) are easier than wide ones [P97][V]

5. Existing computational models (for calibration)


6. Implications for an algorithm

Design: dynamic programming / Viterbi over per-note (or per-chord) finger states, separately per hand, with costs = Σ weighted rules. Hand size enters as a scale factor on all span limits. Tempo and articulation scale context-dependent costs. Parameters below are starting points: tune them on PIG (agreement with any annotator, Msoft/Mrec-style), then sanity-check against the oracles.

6.1 Hard constraints (infinite cost)

| ID | Rule | Parameter | |—|—|—| | H1 | Span between consecutive fingers outside [MinPrac, MaxPrac] × handScale | Table 3.14; handScale 0.85 (small, 19 cm) / 1.0 (22 cm) / 1.12 (24 cm) | | H2 | Same finger on two different simultaneous notes | — | | H3 | Chord: fingers must be in pitch order (RH ascending 1→5, LH 5→1) | allow 1 on two adjacent white keys only as a special state (Chopin) | | H4 | Chord span 1-5 > MaxPrac(1-5) | 15 st default; 12 st small hand; 17 st large | | H5 | 5 crossing under/over any finger, 1 under 5 | — |

6.2 Soft costs between consecutive notes (monophonic line)

| ID | Rule | Suggested cost | |—|—|—| | S1 | Stretch beyond MaxComf / below MinComf | 2 per st (Parncutt) | | S2 | Below MinRel | 1 per st (thumb pair), 2 (other) | | S3 | Above MaxRel | 1 per st (thumb pair), 2 (other) | | S4 | Thumb under 2 / 3 / 4 (rising RH, falling LH) | base 1 / 0.5 / 1, plus S1–S3 | | S5 | Pass level: finger black → thumb white / white-white / black-black / finger white → thumb black | 0 / 1 / 1.5 / 3 (+ thumb-on-black cost) | | S6 | Non-thumb crossing 3-over-4, 4-over-5, 2-over-3 | 6–8 (allowed only in legato double notes / slow passages); 4-over-5 8+ | | S7 | Same finger, different keys (glide/jump) | 0 if a rest/staccato precedes (gap ≥ ½ beat) or slow tempo; 2 for a black→white slide; 6 for legato otherwise | | S8 | Same finger on a repeated note | 0 below ~6 notes/s; then rising linearly to 3 at 10 notes/s; prefer descending 4-3-2-1 / 3-2-1 (cost 0) | | S9 | Weak finger use (4, 5) | 1 per use (Jacobs-style: 0.5 for 5, 1 for 4) | | S10 | Three-Four-Five consecutive | 1 | | S11 | 3→4 transition | 1; 4→3 0 | | S12 | 3/4 adjacent with 3 white, 4 black | 1 | | S13 | Thumb on black | 1 base, +2 per white neighbour; ×0.5 when tempo is slow or legato is pedalled; ×1.5 in the bass (below C3) | | S14 | 5 on black | 0 base, +2 per white neighbour | | S15 | Long finger (2, 3, 4) on black / short finger (1, 5) on white | bonus –0.25 (implements Chopin’s natural position) | | S16 | Asymmetry bonuses (observed preferences) | 3-1 vs 1-3, 1-4 vs 4-1, 3-2, 2-4, 5-4: –0.2 each |

6.3 Three-note / position costs

| ID | Rule | Suggested cost | |—|—|—| | P1 | Full position change (thumb pivot) | 2 | | P2 | Half change (rotation / insertion / jump) | 1 | | P3 | Size of the change beyond MaxComf / MinComf | 1 per st | | P4 | Position change at a phrase/slur/beat boundary or after a rest | ×0.3 discount | | P5 | Position change in the middle of a slur on short notes | ×1.5 | | P6 | Thumb pass immediately before a large leap | +1 (the hand cannot prepare) |

6.4 Chords, double notes, polyphony

| ID | Rule | Suggested cost | |—|—|—| | C1 | Gap between adjacent chord fingers vs relaxed span | use S2/S3 with Table 3.14 within the chord | | C2 | Largest gap not between 1-2 (RH chords) | 1 | | C3 | Octave: 1-5 white / 1-4 black-top / 1-3 | 0 / 0 (1-4 on black also allowed on white: 0.5) / 2 (×0 for large hands in legato) | | C4 | Legato octave line: upper finger alternation 5-4(-3) | –0.5 bonus vs 5-5 under slur | | C5 | Double thirds: standard 13/24/35 pattern | 0; other pairs +1 | | C6 | Held note (sustained voice) blocks its finger; other fingers must fit the span | hard constraint + spans from held finger | | C7 | Silent substitution on a held note | 3 (only when it unlocks a legato that is otherwise impossible; PIG shows 0.34% use) | | C8 | Voicing: melody (top voice RH) on 4 or 5 in chords | 0 (preferred), 3 or 2 on top +0.5 |

6.5 Global / pattern costs

| ID | Rule | Suggested cost | |—|—|—| | G1 | Pattern consistency: repeated or transposed figure (same interval sequence) fingered differently | 1 per differing note (look-back window of 2 bars or 16 notes) | | G2 | Scale template match: if a run matches a diatonic scale ≥ 6 notes, prefer the table 3.1 fingering | –0.5 per matching note | | G3 | Tempo scaling | multiply S5, S6, S13, P1, P2 by (1 + 0.1·(notes/s − 4)) clamped to [0.6, 2.0] | | G4 | Hand assignment (when the hand is not given) | cost 4 for a note taken by the “wrong” hand (pitch side); redistribution allowed when the other hand is idle for ≥ the note duration | | G5 | Pedal active | ×0.5 on legato-only costs (S7, S6, P5) |

6.6 Test cases (oracles)

  1. All 24 scales in §3.1, RH and LH, 2 octaves up and down. Expect an exact match. Also: 4 once per octave and no thumb on black.
  2. Chromatic scale from every pitch, RH and LH. Expect 1 on white, 3 on black, 1-2 at E-F and B-C.
  3. Root-position arpeggios C, G, F, A, E, D, B majors and a, d, e, c minors (RH 1231235; LH 5421421 or 5321321 per §3.3). Dominant 7th: RH 1234 1235.
  4. Five-finger positions C-D-E-F-G (RH 1-5) and the Chopin position E-F♯-G♯-A♯-B (RH 1-5, LH 5-1). The algorithm must not insert a thumb pass.
  5. Alberti bass C-G-E-G LH → 5-1-3-1.
  6. Repeated note ×8 at 12 notes/s → 4321 4321 or 321 321 (accept either). At 2 notes/s → the same finger is acceptable.
  7. Trill C-D ×16 → 1-3, 2-3 or 1-2; never 3-4 or 4-5. Trill E-F♯ → 2-3 (accept 1-3).
  8. Octave scale in C major → 1-5 throughout. Octaves on F♯-G♯-A♯ → 1-4 accepted.
  9. Wide leaps: C4 → G5 RH (19 st > 1-5 MaxPrac) must be a position change, not a stretch.
  10. Pentatonic black-key run (F♯ G♯ A♯ C♯ D♯ ×2): the thumb on black is acceptable. Check that the cost doesn’t explode (compare with Chopin’s practice).
  11. A sequence (e.g. Hanon No. 1 across an octave): the fingering repeats identically (G1).
  12. Parncutt’s examples: D4→C5 played 2-5 = 4 stretch points. C♯4→E4 played 1-2 = 3 points thumb-on-black. A♯3-C♯4-E4 played 2-1-2 = +1 thumb-pass. Use these to unit-test the rule implementations [P97][V].
  13. PIG dataset: report Mgen, Mhigh and Msoft against all annotators. The target is close to the 2nd-order HMM (≈64% Mgen, ≈85% Msoft); the human ceiling is ≈71% / 91% [NSY19][V].

Sources