WHEN VOICES INTERLEAVE: TIMING DEVIATIONS IN SIX
PERFORMANCES OF TELEMANN’S FANTASIAS FOR SOLO FLUTE
Pa ice Thibaud Ma hieu Gi aud Yann Tey au
Uni . Lille, CNRS, Cen ale Lille, UMR 9189 CRIS AL, F-59000 Lille, F ance
[email p o ec ed]
ABSTRACT
Pe o me s con ey musical meaning no only h ough pi ch
and dynamics bu also h ough mic o- iming de ia ions.
This s udy examines pe o mance analysis and iming in
Telemann’s 12 Fan asias o Solo Flu e, ocusing on how
musical elemen s, such as implied polyphony, onse po-
si ions, and me e , in luence musical pe o mance. We
elease a co pus wi h anno a ions on in e lea ed oices
ga he ing 11 musicological sou ces. We i s e alua ed
how simple ules may de ec such in e lea ed oices om
he sco es. We hen analyzed six comple e eco dings o
he an asias, compa ing hei iming de ia ions agains a
me onomic in e p e a ion. Resul s e eal signi ican im-
ing de ia ions in luenced no only by no e posi ion wi hin
hy hmic g oupings bu also by he p esence o in e lea ed
melodic oices, in pa icula when hese in e lea ed oices
a e no a ed wi h opposing s ems.
1. INTRODUCTION
1.1 Pe o mance Analysis and Timing
Musical pe o mance is a p ocess ha ex ends beyond he
me e ep oduc ion o a sco e. As e iewed by [1], key pe -
o mance pa ame e s include empo and iming, dynam-
ics, pi ch, and imb e. Addi ionally, o he aspec s, such as
emo ional impac [2], can also be conside ed. In his s udy,
we ocus on iming, examining i s ole in shaping musical
in e p e a ion and exp essi i y. Speci ically, iming de i-
a ions e e o di e ences in he pe o med no es’ onse s
and du a ions compa ed o a heo e ical, me onomic pe -
o mance a a cons an empo. De ia ions be ween ac ual
imings and me onomic pe o mances a e no andom bu
ecu en [3]. Among exis ing measu es, he In e -Onse -
In e al (IOI) is o en used in musical cogni ion o cha -
ac e ize local a ia ions du ing a pe o mance a di e en
le els (no e, bea , ba , and ph ase) [4].
© P. Thibaud, M. Gi aud, Y. Tey au . Licensed unde a C e-
a i e Commons A ibu ion 4.0 In e na ional License (CC BY 4.0). A -
ibu ion: P. Thibaud, M. Gi aud, Y. Tey au , “When Voices In e lea e:
Timing De ia ions in Six Pe o mances o Telemann’s Fan asias o Solo
Flu e”, in P oc. o he 26 h In . Socie y o Music In o ma ion Re ie al
Con ., Daejeon, Sou h Ko ea, 2025.
1.2 Timing and Musical Elemen s
Only a ew pe o mance s udies ha e a emp ed o ela e
iming de ia ions o speci ic s uc u al o composi ional as-
pec s o he music. Repp [5] examined piano pe o mances
o “T äume ei” by Schumann ac oss di e en in e p e a-
ions and measu ed simila i y in he ela i e du a ions in
consecu i e g ace no es. He also emphasized ex ending he
ini ial downbea as a iming s a egy o pe o ming i a ds
a he ends o melodic ges u es in he opening measu es
o Chopin’s “E ude in E Majo ” [6]. Palme showed he
“melody lead” o co esponding no es in cho ds played by
expe imen ed pianis s [7] and s udied he in e nal s uc u e
o a ill played in Moza sona a in e p e a ions [8]. Cla ke
[9] examined lis ene s’ abili y o pe cei e iming a ia ions
in exp essi e musical sequences, in bo h onal and a onal
exce p s. The s udy e ealed ha a 20 ms leng hening is
pe cep ible wi hin no es anging om 100 o 400ms, and
ha he de ec abili y o a iming de ia ion a ies acco ding
o i s posi ion wi hin he sequence.
Such iming de ia ions include he ba oque p ac ice o
playing no es inégales, wo consecu i e no es o he same
hy hmical alue wi h di e en du a ions [10]. F om i s
i s men ion by Loys Bou geois in 1550 o he la e 18 h
cen u y, such a p ac ice was ini ially es ic ed o one-
qua e o he ac us in duple me e [11]. The s ong
bea s o he i s pa o each bea we e o en empha-
sized compa ed o he ollowing no es, c ea ing pai s o
no es wi h s ong and weak accen s. Such an (unequal)
pe o mance was gene ally no men ioned on he sco es
bu heo ized in musical ea ises and lu e me hods om
his pe iod [12,13]. Cy ga he ed a ious ba oque sou ces
(Quan z, Ma heson) desc ibing whe e he inequali y oc-
cu s in ba oque dances [14]. Moelan s s udied he iming
a io as de ined by É ienne Loulié in 1696 [15], be ween
he longes and he sho es no es when ha psicho dis s and
iolinis s play ga o e [16]. He highligh ed he in luence
o in e al size, bea posi ion, and empo on he no es in-
égales. Honing showed ha his iming a io dec eases
when empo inc eases [17].
1.3 Con ibu ions
How may he pe o me s de ia e om he no a ed du a-
ions? Do hese de ia ions ollow musical elemen s, and
a e hey consis en ac oss pe o me s? We aim o in es-
iga e how some music elemen s in luence a pe o mance.
Focusing on a ba oque co pus – he 12 an asias o solo
363
Figu e 1. The second episode o he Vi ace uga o mo e-
men o Fan asia No.2 (TWV 40:3) begins a mm.38. I
ea u es wo in e lea ed oices, a descending bass line and
an uppe oice. In he au og aph sco e, Telemann occa-
sionally highligh s he bass oice in hese measu es by us-
ing s ems poin ing in opposi e di ec ions wi hin he same
beam, as he e on no es 5, 7, and 9.
lu e by Telemann –, we in es iga e how hey may empha-
size (o no ) some no es, including implied polyphony wi h
in e lea ed oices. We also in es iga e whe he such im-
plied polyphony can be de ec ed om he sco es.
In he ollowing, we in oduce he co pus and he an-
no a ions o in e lea ed oices (Sec ion 2) and a me hod
o p edic ing hose anno a ions (Sec ion 3). We p opose
a me hodology o measu e iming de ia ions in pe o -
mances and conduc an analysis on six pe o mances, in-
cluding a ocus on in e lea ed oices (Sec ions 4 and 5).
Finally, we discuss code and da a dis ibu ion, and he e-
sul s and pe spec i es in pe o mance analysis and implied
polyphony (Sec ions 6 and 7).
2. THE TELEMANN FLUTE FANTASIAS CORPUS
The 12 an asias o solo lu e by Geo g Philipp Telemann
(TWV 40:2-13), edi ed by himsel in Hambu g, p obably
by 1722-1723, a e one o he a e “signi ican wo ks o
unaccompanied lu e be o e he wen ie h cen u y” [18].
Each an asia is a ee- o m composi ion, wi h a sequence
o hemes e oking imp o isa ion. Telemann’s music e-
lec s he mixed as e, “a blending o he F ench, I alian,
English, and Polish na ional s yles” [18]. The an asias
he e include di e en mo emen s which may e lec hese
in luences. Some passages ea u e implied polyphony
h ough in e lea ing o wo o mo e melodic lines, wha
Pis on desc ibes as a “compound melody” [19]. Such
oices [20] may be no a ed on he Telemann au og aph by
a beamed g oup o no es whe e some s ems poin down-
wa d and o he s upwa d (Figu e 1).
We buil he co pus om MusicXML iles ansc ibed
by Á pád Zol án Szabó 1, and amended hese ansc ip-
ions by ca e ully checking he only au hen ica ed copy
o he an asias [21]. Collec ing upon 11 musicological
sou ces (Table 1), we ga he ed anno a ions abou local
keys, mo emen s names 2, implied cho ds and such in e -
1h ps://imslp.o g/wiki/Special:Re e seLookup/
236786 (C ea i e Commons A ibu ion-Sha eAlike 4.0)
2Each an asia is di ided in o a leas h ee sho mo emen s, and up o
se en, wi h he inal one always being a dance. The di ision and naming
Sou ce mo .
names
oices
anno a ions
ha mony
Telemann [21] only
empi
s ems
indica ions
–
B own [23] pa ial – –
By d [24] – 7 –
DeB ee [25] – 7,9-12 –
Eppinge [26] all 4-5,7,10-12 –
Hun [27] – 7 –
Kuijken [28] pa ial – –
Min [22] all uga os only LK (all), some CP
Po e [29] all – –
Sil a [20] 1-3 1-3 LK (all), CP ( 1-3)
Zohn [18] all – –
Table 1. The musicological sou ces collec ed in his
da ase analyze some o all o he 12 Telemann an asias.
Ha mony anno a ions include local keys (LK) and cho ds
p og ession (CP).
oices bubu BuBu uuuu ubbu ubub Bubu
%22.5 15.3 10.8 8.1 6.3 4.5
oices buub bmmm Buuu bbuu Bumm mmmm
% 4.5 3.6 3.6 3.6 2.7 2.7
Table 2. Dis ibu ion o main oice anno a ions in g oups
o ou six een hs in bina y me e mo emen s o he co -
pus. Bold alues show al e na ing uppe /bass oices.
lea ed oices. We also encoded he ha monic anno a ions
on wo an asias by [20,22].
The oice anno a ions accoun o 2,324 no es (28.2%
o he no es o he co pus) and mos ly deno e wo oices.
A middle oice appea s in only h ee mo emen s: he Alle-
g o uga o o Fan asia No.7 [27], he Co en e o Fan asia
No.10, and he Vi ace uga o o Fan asia No.11 [25, 26].
Voice anno a ions a e encoded by u(uppe oice, 48% o
hese anno a ions), m(middle oice, 4.6%), b(bass oice,
21.6%), B(bass oice, no a ed wi h opposi e s em-no es
by Telemann, 25.9%). The sou ces we e la gely consis-
en 3. No es wi hou anno a ions o a e ambiguous cases
a e agged by x. These elemen s we e labeled using he
open-sou ce web pla o m Dez ann [30] (Figu e 2).
Focusing on ull anno a ed g oups (e.g., g oups wi h an-
no a ion on all no es) o ou six een h no es wi hin a bea
in bina y me e , 49.6% o he six een h a e in e lea ed, al-
e na ing be ween an uppe oice and a bass oice (bubu,
ubub, and all combina ions wi h also B). In con as , only
13.5% consis o a single oice, wi h mos being he uppe
oice and 2.7% he middle oice (Table 2). These g oups
o al e na ing no es a e mainly (81%) ound in he uga os
mo emen s (11 o o he 34 mo emen s in he co pus).
o hese sec ions we e no p o ided by Telemann bu a e de i ed om
consensus o majo i y ag eemen among he consul ed sou ces.
3Fo example, six opposi e s em-no es in he co pus (0.3%) a e in e -
p e ed as uppe o middle oice by some sou ces. We kep hem as bass
oices, as in he au og aph.
P oceedings o he 26 h ISMIR Con e ence, Daejeon, Ko ea, Sep embe 21-25, 2025
364
Figu e 2. The Vi ace uga o om Fan asia No.1 (TWV
40:2) mm.11-16 wi h ha monic labels and iden i ica ion o
he uga o o m a ailable on Dez ann .
3. DETECTION OF INTERLEAVED VOICES
IN SCORES
In his sec ion, we s udy whe he , s a ing om he sco e,
a simple algo i hm can de ec he in e lea ed oices as an-
no a ed in he co pus. Whe eas oice sepa a ion in poly-
phonic music is well s udied [31,32], i is no he case o
implied polyphony. Da is [33] p oposed a model based
on music pe cep ion elemen s o iden i y whe e a oice
change occu s. Th ee c i e ia a e applied o analyze all
in e als in a piece: (i) he size o he dia onic in e al
[34, 35]; (ii) he s udy o he change o con ou di ec-
ion (ascending and descending oices a ound he in e -
al) [36,37]; and (iii) he conjunc mo ion o no es on bo h
sides o he in e al [38]. Machine lea ning models could
p edic oice changes, bu gi en ou small, speci ic co pus,
we p e e o ocus on simple ules.
3.1 Me hodology
Fo each in e al Ibe ween wo consecu i e no es, se e al
weigh s a e compu ed:
•w1(Pi ch dis ance). Numbe o dia onic s ep inc e-
men s in I, abo e he hi d ;
•w2(Change o con ou di ec ion). Numbe o
changes o di ec ion o he wo in e als su ounded
Icompa ing o i (0, 1, o 2);
•w3(S e ch o conjunc no es). Sum o he leng hs
o he s e ches o conjunc no es be o e and a e I;
•w4(Al e na ion in in e als). Looking be o e and
a e he no es o ming I, numbe o changes o in-
e al di ec ion un il his al e na ion s opped.
Da is p oposed he i s h ee weigh s, w1, w2, w3, and
he ule w1>0and w1+w2+w3≥4 o p edic a change
o oices [33]. To add ess he limi a ions she iden i ied
ega ding hese c i e ia, we in oduce a new weigh w4and
s udy mo e combina ions o hese weigh s.
3.2 Resul s
We implemen ed wi h music21 [39] hese ules and e al-
ua ed he esul s agains he exis ing anno a ions used as
e e ences o all passages wi h in e lea ed oices. The
o iginal ule achie ed a p ecision o 78% (Table 3), hough
wi h a sligh ly lowe ecall (64%).
ules p ecision ecall F1-sco e
w1>0and w1+w2+w3≥4[33] 0.78 0.64 0.70
w1≥40.77 0.30 0.43
w1>0and w1+w2≥40.81 0.57 0.67
w1>0and w1+w3≥40.75 0.39 0.51
w1+w4≥40.80 0.72 0.76
Table 3. Classi ica ion esul s on he de ec ion o in e -
lea ed oices using se e al se s o ules.
We achie ed be e esul s while conside ing w4, he
highes ecall eaching 72% ac oss he co pus (and
69% when conside ing only uga o mo emen s, da a no
shown). This model is highly selec i e. The ule w1
plays a decisi e ole. Mo emen s wi h la ge in e als be-
ween he in e lea ed oices, such as he Fan asia No. 4
wi h six hs and en hs, pe o m be e (mo e han 90% p e-
cision). Con e sely, he mos di icul cases a ise when
oices a e closely spaced, such as in some pedal passages
like he opening measu es o Fan asia No. 5, whe e a C
pedal is played on o bea s.
The emainde o his pape , ocusing on iming de ia-
ions, elies on he e e ence oice anno a ions.
4. PERFORMANCE ANALYSIS METHODOLOGY
4.1 Reco dings
The e a e o e 50 eco dings o he comple e se o he an-
asias [40]. We ocused he e on six pe o mances (Table 4),
balancing be ween wo “his o ical” eco dings (Rampal
and Kuijken, ha we e al eady s udied by [40], who e al-
ua ed hem as wo e e ence in e p e a ions), and mo e e-
cen ones, as well as balancing be ween (ba oque) a e so
and (mode n) lu es. As in [41], da a we e collec ed o e-
sea ch pu poses using spo DL 4linking Spo i y iden i ie s
o YouTube iden i ie s.
pe o me yea ins umen hype links
Jean-Pie e Rampal 1973 lu e YT / MBID
Ba hold Kuijken 1978 a e so YT/ MBID
Amy Po e 2008 lu e YT
F ançois Laza e i ch 2016 a e so YT / MBID
Emmanuel Pahud 2018 lu e YT / MBID
Ka ja Pi elina 2020 a e so YT
Table 4. The six comple e audio eco dings o he 12 Tele-
mann an asias analyzed in his s udy. Reco ded du ing he
COVID-19 lockdown, he Pi elana eco ding is he only
non-comme cial one and also con ains ideos. The las
column epo s, when a ailable, links o YouTube playlis s
(YT) and MusicB ainz iden i ie s (MBID).
4.2 Expanded Sco e P epa a ion
Some pe o me s, pa icula ly Kuijken, do no s ic ly ad-
he e o all he epea s no ed in he sco e o 7 o he 12 an-
4h ps://gi hub.com/spo DL/spo i y-downloade
P oceedings o he 26 h ISMIR Con e ence, Daejeon, Ko ea, Sep embe 21-25, 2025
365
10 20 30 40 50 60 70 80
Measu e
0
50
100
150
200
Tempo (bpm)
Tocca a Vi ace Fuga o Cadence Passepied
Kuijken
Laza e i ch
Pahud
Pi elina
Po e
Rampal
Figu e 3. Tempo by measu e in bpm, compu ed on h ee
consecu i e measu es, o he six eco dings o Fan a-
sia No.1 (TWV 40:2). All pe o me s unde line he ou
mo emen s o his Fan asia.
asias. Fo example, in Fan asia No.2 and No.5, he does no
play he inal epea o he las dance mo emen (a Gigue
and a Cana y, espec i ely). On he o he side, Rampal
and Pahud, play he Fan asia No.4 wi h an addi ional (non-
w i en) epea in he A ia da Capo ( he hi d mo emen )
AABABA, ins ead o AABA. We lis ed all hese epea s
and gene a ed, o each pe o me , an expanded symbolic
sco e ollowing hei pe o mance.
Mo eo e , pe o ming such ba oque music o en im-
plies adding o namen a ion. Telemann no a ed only 16
g ace no es h oughou his en i e co pus, mos o hem
in Fan asia No.5, and 64 emblemen s. Pe o me s may
add o he o namen s, pa icula ly in slow mo emen s, such
as he in oduc o y Sa abande o Fan asia No.9, as played
by Laza e i ch. Repea s a e o en o namen ed he second
ime, as abundan ly demons a ed by Rampal ( o exam-
ple he Passepied, las mo emen , o Fan asia No.1). We
ha e chosen no o include he w i en o namen s in he ex-
panded sco es, he alignmen p ocedu e desc ibed below
manages o align hem wi h o namen ed pe o mances.
4.3 Alignmen be ween Audio and Expanded Sco e
The symbolic sco es a e aligned o he audio eco dings
wi h a wo- old alignmen p ocedu e.
We ook in o accoun uning di e ences be ween he
a e so and mode n lu es wi h a uning es ima o [42].
Audio spec og ams we e compu ed using 22.05 kHz sam-
pling a e, window size 1024, and hop size 512. Since
we deal wi h monophonic an asias, we selec ed he un-
damen al equency 0as he key audio ea u e and
gene a ed, o each no e, disc e e ha monic lu e em-
pla es as Gaussians cen e ed a {n 0}1≤n≤10 wi h (1/np)-
dec easing ampli ude [43]. The obse a ion model com-
pu es simila i ies be ween he audio and he symbolic se-
quences (no e pi ches and es s). We u he applied a
so max unc ion u ning hese simila i ies in o p oba-
bili ies, exp essed in log-p obabili y space. The ew ies
(22) we e handled by me ging he wo no es in o a single
one wi h summed du a ion.
Second, we use a decoding module elying on Dynamic
Time Wa ping (DTW) [44,45] o e ie e he op imal pa h
(i.e., alignmen ) be ween he symbolic sequences om he
(a)
2
1
0
1
2
3
4
5
6
o
(%)
b
u
b
u
B
u
B
u
Bu
1 2 3 4 5 6 7 8 9 10
no e
6
4
2
0
2
4
IOI
(%)
(b)
2
1
0
1
2
3
4
5
6
o
(%)
b
u
b
uB
u
B
uB
u
1 2 3 4 5 6 7 8 9 10
no e
6
4
2
0
2
4
IOI
(%)
Figu e 4. The ∆IOI and ∆o alues o he passage shown
in he Figu e 1 a e compa ed ac oss wo pe o mances on
a mode n lu e. In Po e ’s pe o mance (a) , he a e age
o e absolu e alues o ∆IOI is 2.9%. In con as , Rampal
(b) , who plays his mo emen a a as e empo, has less
de ia ions (1.9%).
expanded sco e and he ou pu o he obse a ion model.
As we do no ha e g ound u h alignmen o he an asias,
alignmen was no e alua ed he e bu only used as a sup-
po o u he exploi a ions, as done in [46] o oice.
Once alignmen is comple e, each no e nhas a pe o -
mance onse op(n), om which we de i e a pe o mance
IOI, deno ed as IOIp(n). This also allows us o obse e
empo a ia ions along each an asia (Figu e 3).
4.4 Me ics o Timing De ia ions
We wan o compa e op(n)and IOI p(n) o heo e i-
cal coun e pa s, assuming no iming de ia ions. Na -
u ally, pe o mances exhibi global empo a ia ions (as
illus a ed in Figu e 3), including g adual slowdowns o
speedups. Howe e , ou ocus he e is on mic o-de ia ions
a a local scale, such as wi hin a g oup o six een h no es.
To accoun o his, we use each measu e as a e e ence
poin o he empo. Fo each no e n, we hus es ima e i s
me onomic onse om(n), assuming ha he en i e measu e
M(n)encompassing i is played a a cons an empo:
om(n) = om(n1) + σ(n)om(n2)−om(n1)
whe e n1and n2a e he i s bea s o he measu es
M(n)and M(n) + 1, and σ(n)∈[0,1[ ep esen s he
p ecise symbolic posi ion a io o nwi hin his measu e.
We also compu e he me onomic IOI as IOI m(n). The
de ia ions in IOI and onse iming a e hen compu ed:
∆IOI (n) = IOI p(n)−IOI m(n)
du (M(n))
∆o(n) = op(n)−om(n)
du (M(n))
P oceedings o he 26 h ISMIR Con e ence, Daejeon, Ko ea, Sep embe 21-25, 2025
366
15
10
5
0
5
10
15
IOI
(%)
1s TM 2nd TM 3 d TM
Kuijken Laza e i ch Pahud Pi elina Po e Rampal
Figu e 5.∆IOI (%) o he 1s , 2nd, and 3 d eigh h no es
wi hin a bea in g oups o h ee, ac oss all mo emen s in a
e na y me e and o all pe o me s. The boxes span om
he i s o he hi d qua ile, while he whiske s ex end o
he mos ex eme da a poin s wi hin 1.5 imes he in e qua -
ile ange (IQR). The median is indica ed by a ho izon al
line, and he mean by a ’+’ sign.
whe e du (M(n)) is he du a ion o he enclosing measu e
M(n). Fo example, in he measu e 38-39 on he Fan asia
No.2 (Figu e. 1), Po e plays he second eigh no e 50ms
la e han i s me onomic posi ion (Figu e 4a). This delay
co esponds o ∆IOI (n1) = ∆o(n2) = 3.3% o he mea-
su e du a ion (abou 20% o an eigh h no e). No e ha , by
de ini ion, ∆o(n1) = 0 on he i s no e (s ong bea ). The
inequali y a io is he e 1.30.
5. TIMING DEVIATION RESULTS
We ocused on eigh h and six een h no es. These c i e ia
we e analyzed: on-bea /o -bea posi ion, in e lea ed oice
anno a ion, me e , posi ion o a no e wi hin a bea , o a
combina ion o hese il e s wi h he no es du a ions.
Means and s anda d de ia ions o ∆IOI a e p esen ed
in Table 5. These dispe sed dis ibu ions de ia e om no -
mali y due o a ious ac o s, including human a iabili y
in pe o mance, he accen ua ion o s ong bea s, and he
g ouping e ec s o speci ic hy hms. This is con i med us-
ing a Shapi o-Wilk es . Because o his non-no mali y, we
use he Mann-Whi ney U es [47] o compa e wo selec ed
popula ions (Table 6) while simul aneously assessing he
e ec size. An e ec size g ea e han 0.5 is conside ed
as la ge. We i s ocus on gene ic s a is ics (Sec ion 5.1),
hen analyze he impac o in e lea ed oices (Sec ion 5.2).
5.1 G oup o No es in a Same Bea
In e na y me e , when conside ing g oups o h ee con-
secu i e eigh h no es wi hin a bea (
ZZZ
), he second
eigh h no e is, on a e age, played sho e by all pe o m-
e s (Fig. 5). Rema kably, Pahud plays he i s eigh h no e
o hese g oups ∆IOI 1= 5.23% longe han he second
ones (s ong e ec size o 0.79), leading o a ∆o2= 2.35%
de ia ion on he onse o he second no es. Acco ding o
Ga ison, “[Pahud’s] app oach o a icula ion is simila o
ha o a e so playe s he slu s less o en and a ies his
onguing s yle” [40]. Fo example, in he las mo emen
o Fan asia No.3 (Gigue in 6/8 ), he emphasizes he i s
eigh h no e o each g oup, pa icula ly on he i s bea .
The hi d no es a e played sligh ly longe compa ed
o he second ones (up o 4.27% o Pi elina), al hough
he e ec is less p onounced (0.29-0.43). These de ia-
ions align wi h pe o mance p ac ices in Ba oque e na y
dances, whe e “S ong hy hmic a icula ion, some imes
lead o exp essi e placemen o no es on he second o hi d
subdi ision o he bea ” [48].
In bina y me e , when conside ing g oups o ou six-
een h no es wi hin he same bea (
©
©
©
©
), he i s no e is,
in a e age, longe . These di e ences a e signi ican o all
pe o me s (excep o Rampal). Kuijken plays i 1.78%
longe , wi h a s ong e ec size o 0.69, as in he i s Vi-
ace uga o o Fan asia No.3 .
5.2 In e lea ed and Opposi e-S em No es
In e lea ed passages mos commonly occu in uga o
mo emen s, pa icula ly in bina y me e s wi hin his co -
pus, such as he wo Vi ace uga o sec ions o Fan asia No.
3, sepa a ed by ou measu es o a La go.
We compu e he e he “
CC
in e lea ed” me ics o all
bea s in bina y me e whe e a pai o eigh h no es ap-
pea s, each wi h a di e en oice anno a ion (excluding
cases wi h a single oice anno a ion o an eigh h es ). Fo
ou pe o me s, he e is a signi ican di e ence in im-
ing be ween he i s and second eigh h no es, wi h s ong
e ec sizes (0.57-0.67). Rema kably, o h ee pe o m-
e s (Kuijken, Pahud and Po e ), no signi ican di e ence
is obse ed when compa ing he same si ua ion in non-
in e lea ed passages. Laza e i ch’s in e p e a ion is pa -
icula ly in e es ing, as he plays he i s eigh h no e 3.14%
longe in in e lea ed passages and 2.64% longe in non-
in e lea ed ones. This could be explained by his consis en
app oach o inequali y in iming wi hin his Ba oque co -
pus, and some imes a sligh ly low empo like in he i s
mo emen o Fan asia No.6 (Adagio), which he plays in
3’45", while he as es pe o me s played i in less han
2’30".
Fo ou pe o me s (Kuijken, Laza e i ch, Pahud,
Pi elina), he e ec s a e mode a e o s ong (0.41-0.60)
when one ocus on he opposi e s emmed no es no a ed in
Telemann au og aph, as shown on he B/xline on Table 6.
The speci ic e ec o hese no es agains he bass no es
no anno a ed by Telemann (B/b) is ema kably obse ed
(0.44-0.56) in he same eco dings.
The speci ic e ec o in e lea ed can also be seen on
anno a ed six een h-no e g oups (
©
©
©
©
), 9.86% o hem
ea u ing in e lea ed oices (Table 2). As in he non-
in e lea ed case, he i s six een h ∆IOI is qui e di e -
en om he o he s, bu , no ably, he e is o Rampal and
Pi elina a signi ican di e ence (p- alue <10−3), be ween
he second and he hi d six een hs wi h s ong e ec sizes
o 0.83 and 0.74, whe eas he e is no s a is ical di e -
ence in g oups wi hou in e lea ed oices (da a no shown).
S udies could u he in es iga e hese cases, in pa icula
by sepa a ing hem acco ding o hei in e lea ing pa e ns,
beginning by he uppe oice o by he bass oice.
P oceedings o he 26 h ISMIR Con e ence, Daejeon, Ko ea, Sep embe 21-25, 2025
367
anno a ed oices on bea o bea
pe o me x b B b +B u m all
all
Rampal -0.27 ± 9.64 -0.47 ± 6.95 -0.14 ± 5.85 -0.29 ± 6.36 0.26 ± 7.15 -0.60 ± 7.30 -0.53 ± 10.63 -0.28 ± 9.47 0.05 ± 7.46 0.26 ± 9.26
Kuijken 0.07 ± 9.23 -0.17 ± 4.59 0.85 ± 5.46 0.40 ± 5.12 0.52 ± 7.06 -0.68 ± 7.57 0.69 ± 10.40 1.13 ± 9.19 -0.29 ± 6.75 0.09 ± 8.49
Po e -0.09 ± 9.78 0.50 ± 5.26 -0.09 ± 5.65 0.17 ± 5.49 0.20 ± 7.06 -0.57 ± 8.22 0.01 ± 10.90 0.88 ± 9.49 -0.07 ± 7.25 0.28 ± 9.00
Laza e i ch -0.19 ± 10.76 -0.66 ± 5.43 0.26 ± 6.31 -0.14 ± 5.95 1.28 ± 8.54 -1.45 ± 8.57 0.79 ± 12.02 1.55 ± 10.90 -0.73 ± 8.03 -0.90 ± 9.71
Pahud -0.14 ± 9.15 0.05 ± 4.70 1.20 ± 5.30 0.69 ± 5.07 0.35 ± 7.42 -1.26 ± 7.66 0.73 ± 10.43 1.29 ± 8.81 -0.62 ± 6.59 -0.06 ± 8.20
Pi elina 0.00 ± 9.80 -1.23 ± 4.87 -2.32 ± 6.55 -1.84 ± 5.89 1.96 ± 8.58 -0.28 ± 7.61 -0.26 ± 11.20 0.45 ± 10.09 0.23 ± 7.38 0.62 ± 9.31
CC
in bina y me e
©
©
©
©
in bina y me e
ZZZ
in e na y me e (Fig. 5)
pe o me 1s
in e lea ed 2nd
in e lea ed 1s
x 2nd
x 1s
©
2nd
©
3 d
©
4 h
©
1s
2nd
3 d
Rampal -0.71 ± 6.28 0.26 ± 7.82 -0.38 ± 10.55 -0.10 ± 10.72 -0.07 ± 3.42 -0.02 ± 3.65 -0.54 ± 4.26 0.24 ± 4.10 -0.36 ± 9.22 -0.91 ± 7.69 0.61 ± 7.03
Kuijken 0.56 ± 4.22 -0.29 ± 4.74 0.76 ± 10.95 -0.42 ± 9.56 1.23 ± 3.84 -0.55 ± 2.99 -0.60 ± 3.49 -0.15 ± 3.68 0.71 ± 6.10 -0.74 ± 7.40 0.78 ± 9.27
Po e 1.21 ± 5.35 -0.63 ± 5.35 0.13 ± 10.49 0.43 ± 10.49 0.52 ± 3.31 -0.53 ± 2.94 -1.14 ± 2.79 0.51 ± 3.41 -0.03 ± 6.52 -0.43 ± 8.02 0.49 ± 7.12
Laza e i ch 1.72 ± 6.62 -1.42 ± 6.08 1.12 ± 11.79 -1.52 ± 11.00 1.45 ± 4.05 -1.01 ± 3.68 -0.16 ± 4.55 -0.65 ± 4.97 2.16 ± 10.09 -2.99 ± 8.78 0.75 ± 9.31
Pahud 0.29 ± 4.29 -0.33 ± 4.07 0.01 ± 9.86 0.02 ± 10.24 1.58 ± 4.19 -1.06 ± 2.62 -0.90 ± 2.66 -0.34 ± 3.20 2.28 ± 6.45 -2.95 ± 4.99 0.55 ± 7.17
Pi elina 1.21 ± 8.73 -0.75 ± 6.91 0.03 ± 9.91 1.17 ± 11.12 0.56 ± 3.84 -0.43 ± 3.29 -0.36 ± 4.10 0.67 ± 4.70 0.97 ± 9.12 -2.50 ± 7.44 1.77 ± 7.06
Table 5. Mean and s anda d de ia ion o he ∆IOI (n)o no es (excluding es s), in pe cen age on he leng h o each
measu e, o he six pe o me s on all Telemann Fan asias. Values abo es 1% a e in bold. (Top). The non-anno a ed no es
(column x) a e played wi h small de ia ions in con as o he uppe and lowe oices (columns u,b+B). (Bo om). Focus
on g oups o no es wi hin he same bea . Se e al pe o me s de ia e anno a ed no es w i en wi h in e lea ed oices mo e
han when hey a e no . They also emphasize he i s no e o he g oup.
Rampal
Kuijken
Po e
Laza e i ch
Pahud
Pi elina
all he co pus
b+B/x– 0.55 0.54 0.54 0.58 0.44
B/x– 0.57 – 0.56 0.60 0.41
u/x0.54 0.53•– 0.57 0.52•0.58
B/b– 0.56•– 0.55•0.55 0.44
u/b+B0.53•– 0.46 0.53•0.43 0.65
all he co pus
all: on bea / o bea – 0.57 0.54 0.60 0.61 –
: on bea / o bea – 0.55 0.54 0.61 0.58 –
bina y me e
CC
in e l.: 1s / 2nd – 0.58 0.64 0.67 0.57•–
CC
x: 1s / 2nd – – – 0.60 – –
©
©
©
©
: 1s / 2nd – 0.69 0.64 0.72 0.75 0.61
©
©
©
©
: 2nd / 3 d 0.59 – 0.59 0.44•– –
©
©
©
©
: 3 d / 4 h 0.40 0.45•0.28 – 0.43 0.39
e na y me e
ZZZ
: 1s / 2nd – 0.58•– 0.69 0.79 0.63
ZZZ
: 2nd / 3 d 0.42•– 0.43•0.36 0.31 0.29
Table 6. E ec size o Mann-Whi ney U es measu ing
he magni ude o he di e ence be ween wo g oups on he
me ic ∆IOI (n), ocusing on cases whe e he p- alue is
<10−2(•) o <10−3(all o he s).
6. DATA AND CODE AVAILABILITY
We p o ide, unde open licenses, cu a ed sco es, an-
no a ions on s uc u e (mo emen names, epea s), and
anno a ions on implied polyphony and ha mony, unde
he eche che.da a.gou . long- e m a chi e a h ps:
//doi.o g/10.57745/MSLVWS. We u he dis-
ibu e hese ma e ials, oge he wi h code enabling
he ep oduc ion o esul s and igu es, ia he gi
eposi o y a h ps://gi lab.com/algomus. /
elemann- oices-in e lea e. The da ase also
includes no e-le el synch oniza ion wi h each eco ding,
along wi h compu ed me onomic onse s and iming de i-
a ions. These da a can u he be b owsed, and que ied on
a web applica ion buil wi h S eamli 5.
5h ps:// elemann oicesin e lea e.s eamli .app
7. CONCLUSIONS AND PERSPECTIVES
Whe he hey play mode n lu e o a e so, musicians sig-
ni ican ly play wi h iming de ia ions. Analyzing six pe -
o mances o he Telemann an asias, we showed ha he
ini ial no es o g oups o h ee eigh hs in a e na y me e
o ou six een hs in a bina y me e a e emphasized, being
longe (up o 5.23% ela i e o he measu e du a ion) han
he o he no es in he g oup.
In e lea ed oices a e a signi ican pa o he Telemann
an asias – 28.2% o he no es om ou musicological
sou ces. We showed ha simple ules allow o de ec hese
oices om he sco e wi h a F1-sco e o 76%. No es iné-
gales a e a s anda d Ba oque p ac ice, bu we showed he e
ha hey a e qui e signi ican wi hin in e lea ed oices.
Th ee o he six pe o me s play wi h signi ican iming de-
ia ions in such eigh h no es, whe eas hey do no in non-
in e lea ed passages. No es w i en wi h opposi e s ems in
Telemann’s au og aph sco e ecei e pa icula emphasis.
Any s udy on pe o mance depends on alignmen qual-
i y. Many audio- o-sco e aligne s exis oday, elying
ei he on audio ea u es (e.g., ch omas) o mo e ad-
anced obse a ion models (e.g., deep lea ning-based ap-
p oaches). A na u al ex ension o ou wo k would be o
sys ema ically e alua e such me hods on anno a ed g ound
u h and assess hei pe o mance compa ed o ou sim-
ple, 0-based model. Explici ly conside ing o namen a-
ions may also be e cap u e he pe o me s’ in e p e a-
ions. The a iabili y o epea s, handled manually he e,
p esen s ano he esea ch oppo uni y.
The cu en s udy was limi ed on six eco dings and
may p esen biais speci ic o hese a is s (and his co pus).
Gi en he la ge numbe o exis ing eco dings o hese an-
asias, u he s udies could conside addi ional audio da a.
Finally, he pe o mances could be analyzed on o he mu-
sic elemen s, such as pi ches, in e als, o pa e ns, o o he
passages wi h implied polyphony, such as a peggios.
P oceedings o he 26 h ISMIR Con e ence, Daejeon, Ko ea, Sep embe 21-25, 2025
368
8. ACKNOWLEDGMENTS
The au ho s since ely hank Flo ence Le é and Ken
Dégue nel om he Algomus eam o hei insigh ul
commen s, as well as Emmanuel Leguy o his suppo
wi h he Dez ann pla o m. We a e also g a e ul o he con-
e ence e iewe s o hei cons uc i e eedback.
9. ETHICS STATEMENT
The da ase does no con ain any lis ene -speci ic o pe -
sonally iden i iable in o ma ion, as i is based solely on
p e ious academic s udies.
10. REFERENCES
[1] A. Le ch, C. A hu , A. Pa i, and S. Gu u ani, “Mu-
sic Pe o mance Analysis: A Su ey,” in ISMIR 2019,
Del , Ne he lands, 2019.
[2] S. B. Dias and L. J. Hadjileon iadis, “Telemann’s Fan-
asia No 2 à T a e s Sans Basse: Composi ional In-
en ions, Pe o mance Cha ac e is ics and Audio Foo -
p in Signal P ocessing Analysis,” in Pe spec i es on
Music, Sound and Musicology: Resea ch, Educa ion
and P ac ice, se . Cu en Resea ch in Sys ema ic Mu-
sicology, L. Co eia Cas ilho, R. Dias, and J. F. Pinho,
Eds. Cham: Sp inge In e na ional Publishing, 2021,
pp. 31–47.
[3] Gab ielsson, Al , “S udying emo ional exp ession in
music pe o mance,” Bulle in o he Council o Re-
sea ch in Music Educa ion, pp. 47–53, 1999.
[4] A. F ibe g and G. Ba el, “S uc u al Communica ion,”
in The Science & Psychology o Music Pe o mance:
C ea i e S a egies o Teaching and Lea ning, Ap .
2002.
[5] B. H. Repp, “Di e si y and commonali y in music pe -
o mance: An analysis o iming mic os uc u e in
Schumann’s "T äume ei.",” Jou nal o he Acous ical
Socie y o Ame ica, ol. 92, no. 5, pp. 2546–2568,
1992.
[6] ——, “A mic ocosm o musical exp ession. I. Quan-
i a i e analysis o pianis s’ iming in he ini ial mea-
su es o Chopin’s E ude in E majo ,” The Jou nal o
he Acous ical Socie y o Ame ica, ol. 104, no. 2 P 1,
pp. 1085–1100, Aug. 1998.
[7] C. Palme , “On he Assignmen o S uc u e in Music
Pe o mance,” Music Pe cep ion: An In e disciplina y
Jou nal, ol. 14, no. 1, pp. 23–56, 1996.
[8] ——, “Ana omy o a Pe o mance: Sou ces o Musical
Exp ession,” Music Pe cep ion: An In e disciplina y
Jou nal, ol. 13, no. 3, pp. 433–453, 1996.
[9] E. F. Cla ke, “The pe cep ion o exp essi e iming in
music,” Psychological Resea ch, ol. 51, no. 1, pp. 2–
9, Jun. 1989.
[10] D. Fulle , No es Inégales, se . The New G o e Dic-
iona y o Music and Musicians. Ox o d Uni e si y
P ess, 2001, ol. 1.
[11] G. Houle, “V- A icula ion o Quan a i e Me e ,” in
Me e in Music, 1600-1800: Pe o mance, Pe cep-
ion, and No a ion, se . Music–Schola ship and Pe o -
mance. Blooming on: Indiana Uni e si y P ess, 1987.
[12] J. D. Boland, Me hod o he One-Keyed Flu e,
Ba oque and Classical. Be keley: Uni e si y o Cal-
i o nia P ess, 1998.
[13] C. A. Fon ijn, “Quan z’s unegal: Implica ions o he
pe o mance o 18 h-cen u y music,” Ea ly Music, ol.
XXIII, no. 1, pp. 54–64, Feb. 1995.
[14] M. Cy , Pe o ming Ba oque Music. London: Rou -
ledge, 2016.
[15] E. Loulié, Elemen s Ou P incipes de Musique, Mis
Dans Un Nou el O d e. Pa is, F ance: Ch is ophe
Balla d, 1696.
[16] D. Moelan s, “The Pe o mance o No es Inégales: The
In luence o Tempo, Musical S uc u e, and Indi idual
Pe o mance S yle on Exp essi e Timing,” Music Pe -
cep ion: An In e disciplina y Jou nal, ol. 28, no. 5,
pp. 449–460, 2011.
[17] H. Honing, “E idence o empo-speci ic iming in mu-
sic using a web-based expe imen al se up,” Jou nal o
Expe imen al Psychology: Human Pe cep ion and Pe -
o mance, ol. 32, no. 3, pp. 780–786, 2006.
[18] S. D. Zohn, Music o a Mixed Tas e: S yle, Gen e and
Meaning in Telemann’s Ins umen al Wo ks. Ox o d:
Ox o d Uni e si y p ess, 2008.
[19] W. Pis on, Coun e poin . New Yo k London: W. W.
No on & Company, 1947.
[20] P. da Sil a and A. Ca los, “A pe o mance guide o
h ee o Telemann’s 12 an asias o lu e,” Ph.D. dis-
se a ion, Uni e si y o Alabama Lib a ies, 2012.
[21] G.P. Telemann, 12 Fan asias o Flu e wi hou Bass,
TWV 40:2-13, se . Royal Conse a oi e o B ussels, Li-
b a y (B-Bc). Hambu g: Telemann, 1732-1733, no. T
5823 W.
[22] Kyoung Joo Min, “Analyses o he wel e an asies o
solo lu e(1732-1733) and he wel e an asies o solo
iolin(1735) o G. Ph. Telemann :,” Ph.D. disse a ion,
Uni e si y o Washing on, 1998.
[23] R. B own, “Telemann Fan asias: A ea o ingenui y
and inspi a ion,” 2015.
[24] D. By d and T. C aw o d, “P oblems o music in o -
ma ion e ie al in he eal wo ld,” In o ma ion P o-
cessing & Managemen , ol. 38, no. 2, pp. 249–272,
Ma . 2002.
P oceedings o he 26 h ISMIR Con e ence, Daejeon, Ko ea, Sep embe 21-25, 2025
369
[25] R. de B ee, “Telemann’s Fan asias: A Fan as ic Sou ce
o Inspi a ion. How can I imp o e my imp o isa o y
skills , using Telemann’s Fan asias ?” 2020.
[26] Eppinge , S., “Geo g Philipp Telemann : 12 Fan asien
ü Flö e solo, Teil I,” TIBIA, ol. 2, pp. 86–99, 1984.
[27] L. Hun , “A Ho n T ansc ip ion o he wel e an asias
o lu e wi hou bass, TWV 40:5-8, by Geo g Philipp
Telemann,” Ph.D. disse a ion, Uni e si y o Geo gia,
2016.
[28] Ba hol Kuijken, “12 Fan asias TWV 40:2-13 (wi h
Facsimile),” 1987.
[29] A. Po e , “Telemann: 12 Fan asias o Flu e wi hou
Bass – A S udy Guide DVD,” 2008.
[30] C. Balles e , B. Baco , L. Bigo, V. N. Bo san,
L. Cou u ie , K. Dégue nel, Q. Dinel, L. Feis haue ,
K. F iele , M. Go ham, R. G oul , J. Hen schel,
A. d’Hooge, D.-V.-T. Le, F. Le é, F. Macca ini, I. Ma -
iˇ
ci´
c, G. Micchi, M. Mülle , A. S ama iadis, T. Ta in,
P. Thibaud, C. Weiß, R. Yang, E. Leguy, and M. Gi-
aud, “In e ac ing wi h Anno a ed and Synch onized
Music Co po a on he Dez ann Web Pla o m,” T ans-
ac ions o he In e na ional Socie y o Music In o ma-
ion Re ie al, ol. 8, no. 1, May 2025.
[31] E. Chew and X. Wu, “Sepa a ing Voices in Polyphonic
Music: A Con ig Mapping App oach,” in Compu e
Music Modeling and Re ie al: Second In e na ional
Symposium, CMMR 2004, Esble g, Denma k, May
2004, pp. 1–20.
[32] E. Cambou opoulos, “Voice And S eam: Pe cep ual
And Compu a ional Modeling O Voice Sepa a ion,”
Music Pe cep ion - MUSIC PERCEPT, ol. 26, pp. 75–
94, Sep. 2008.
[33] S. Da is, “Implied Polyphony in he Solo S ing Wo ks
o J. S. Bach: A Case o he Pe cep ual Rele ance
o S uc u al Exp ession,” Music Pe cep ion, ol. 23,
no. 5, pp. 423–446, Jul. 2006.
[34] G. A. Mille and G. A. Heise, “The T ill Th esh-
old,” The Jou nal o he Acous ical Socie y o Ame ica,
ol. 22, no. 5, pp. 637–638, Sep. 1950.
[35] E. Na mou , The Analysis and Cogni ion o Ba-
sic Melodic S uc u es: The Implica ion-Realiza ion
Model. Chicago: Uni e si y o Chicago P ess, 1990.
[36] G. A. Heise and G. A. Mille , “An Expe imen al S udy
o Audi o y Pa e ns,” The Ame ican Jou nal o Psy-
chology, ol. 64, no. 1, pp. 68–77, 1951.
[37] D. Hu on and M. Royal, “Wha Is Melodic Accen ?
Con e ging E idence om Musical P ac ice,” Mu-
sic Pe cep ion: An In e disciplina y Jou nal, ol. 13,
no. 4, pp. 489–516, 1996.
[38] W. J. Dowling, “The pe cep ion o in e lea ed
melodies,” Cogni i e Psychology, ol. 5, no. 3, pp.
322–337, No . 1973.
[39] M. S. Cu hbe and C. A iza, “Music21: A oolki o
compu e -aided musicology and symbolic music da a.”
in ISMIR, J. S. Downie and R. C. Vel kamp, Eds. In-
e na ional Socie y o Music In o ma ion Re ie al,
2010, pp. 637–642.
[40] L. Ga ison, “Telemann’s Twel e Fan asias: A Su ey
o Edi ions and Reco dings,” Flu is Qua e ly, ol. 45,
no. 5, pp. 24–31, 2020.
[41] D. Edwa ds, S. Dixon, and E. Bene os, “PiJAMA: Pi-
ano Jazz wi h Au oma ic MIDI Anno a ions,” T ansac-
ions o he In e na ional Socie y o Music In o ma ion
Re ie al, ol. 6, no. 1, Sep. 2023.
[42] B. McFee, C. Ra el, D. Liang, D. Ellis, M. McVica ,
E. Ba enbe g, and O. Nie o, “Lib osa: Audio and Mu-
sic Signal Analysis in Py hon,” in Py hon in Science
Con e ence, Aus in, Texas, 2015, pp. 18–24.
[43] A. Con , “A Coupled Du a ion-Focused A chi ec-
u e o Real-Time Music- o-Sco e Alignmen ,” IEEE
T ansac ions on Pa e n Analysis and Machine In elli-
gence, ol. 32, no. 6, pp. 974–987, Jun. 2010.
[44] S. Dixon and G. Widme , “MATCH: A Music Align-
men Tool Ches ,” in P oceedings o he 6 h In e -
na ional Con e ence on Music In o ma ion Re ie al,
London, Uni ed Kingdom, 2005.
[45] M. Mülle , “Dynamic Time Wa ping,” in In o ma ion
Re ie al o Music and Mo ion. Sp inge , Be lin,
Heidelbe g, 2007, pp. 69–84.
[46] L. Salais, P. A ias, C. L. Moine, V. Rosi, Y. Tey au ,
N. Obin, and A. Roebel, “P oduc ion S a egies o Vo-
cal A i udes,” in In e speech 2022. Incheon, Ko ea:
ISCA, Sep. 2022, p. 4985.
[47] H. B. Mann and D. R. Whi ney, “On a Tes o Whe he
one o Two Random Va iables is S ochas ically La ge
han he O he ,” The Annals o Ma hema ical S a is ics,
ol. 18, no. 1, pp. 50–60, 1947.
[48] M. Li le and N. Jenne, Dance and he Music o J. S.
Bach, Expanded Edi ion, se . Music: Schola ship and
Pe o mance. Blooming on, IN: Indiana Uni e si y
P ess, 2001.
P oceedings o he 26 h ISMIR Con e ence, Daejeon, Ko ea, Sep embe 21-25, 2025
370
