ORIGINAL PAPER
Animal Cogni ion (2025) 28:26
h ps://doi.o g/10.1007/s10071-025-01940-7
es s on a laye ed assemblage o di e en social cogni i e
skills (e.g., join a en ion, common g ound, collabo a ion,
and easoning abou communica i e in en ; Cla k 1996),
and in ol es ace- o- ace in e ac ion, equen employmen
o mu ual gaze, and u n- aking, o ming he “in e ac ion
engine” (Le inson 2006). In pa icula , u n- aking, a coop-
e a i e en e p ise (Sacks e al. 1974), has ecen ly ecei ed
much esea ch a en ion because i has been sugges ed as an
ancien unde pinning o he language sys em, wi h p ecu -
so s al eady p esen ac oss he p ima e lineage (Le inson
2016). Tu n- aking e e s o ecip ocal exchanges o al e -
na ing sho and lexible u ns be ween wo o mo e in e -
ac an s and is go e ned by dis inc ea u es: one in e ac an
speaks a a ime, he o ganisa ion o alloca ed u ns be ween
in e ac an s a e a ied, empo al gaps a e minimised
In oduc ion
Language, a uniquely human ai and coope a i e laye ed
sys em, has been a opic o in e es ac oss cen u ies and
esea ch disciplines because i s eme gence emains a sci-
en i ic puzzle (Ch is iansen and Ki by 2003; Fi ch 2010;
Hause e al. 2014; Knigh e al. 2000). Recen ly, i has been
p oposed ha he special capaci y o social in e ac ion
among humans acili a ed he e olu ion o language (Le in-
son 2006, 2016; see also Vygo sky 1962). This capaci y
Kayla Kol
kayla.w[email p o ec ed]
1 Compa a i e BioCogni ion, Ins i u e o Cogni i e Science,
Osnab ück Uni e si y, Osnab ück, Ge many
Abs ac
Coope a i e u n- aking, a undamen al cha ac e is ic o human social in e ac ion, has been pos ula ed as a c ucial mecha-
nism o language eme gence and is obse ed ac oss he p ima e lineage. Howe e , ela i ely li le is known abou he
in luence o demog aphic and social ac o s on u n- aking. As acco ding o he sociolinguis ic Communica ion Accommo-
da ion Theo y, indi iduals adap hei communica ion acco ding o hei ecipien cha ac e is ics, which may shape u n-
aking. Thus, we aimed o gain insigh s in o he ac o s (age, ela edness, dominance ank, and social bonds) in ela ion o
he u n- aking in as uc u e o one o ou closes li ing ela i es, he chimpanzee (Pan oglody es schwein u hii) o he
Ngogo popula ion in Kibale Na ional Pa k, Uganda. We speci ically ocused on a coope a i e con ex , g ooming, including
ges u es and ac ions, and collec ed da a o e nine mon hs (Sep embe 2021 o June 2022) in ol ing 42 male chimpanzees.
We analysed 311 g ooming in e ac ions among 157 dyads conce ning he ole o demog aphic and social ac o s in u n
ansi ion in as uc u e, u n ansi ion ypes, and empo al ela ionships. Ou indings demons a ed ha u n ansi ions
and ypes we e in luenced by age and dominance ank, whe eas social bonds and ela edness did no exhibi e ec s. Spe-
ci ically, he p obabili y o u n ansi ions was highe o olde ini ia o s and lowe - anking o younge ecipien s. These
e ec s a ied ac oss u n ansi ion ypes, whe e ini ia o ’s dominance ank and ela edness showed no e ec s on any
ype. In addi ion, no e ec was ound o he empo al ela ionships. Al hough he social dynamics o u n- aking emain
la gely unexplo ed ac oss bo h human and non-human s udies, ou indings sugges ha u n- aking can occu selec i ely
be ween ce ain indi iduals, in line wi h he Communica ion Accommoda ion Theo y, unde sco ing he need o g ea e
ocus on in es iga ing how demog aphic and social ac o s shape u n- aking.
Keywo ds Language e olu ion · Coo dina ion · G ooming · Chimpanzees · Social accommoda ion · Age ·
Rela edness · Dominance ank · Social bonds
Recei ed: 13 Sep embe 2024 / Re ised: 13 Feb ua y 2025 / Accep ed: 14 Feb ua y 2025 / Published online: 24 Ma ch 2025
© The Au ho (s) 2025
Tu n- aking in g ooming in e ac ions o chimpanzees (Pan oglody es
schwein u hii) in he wild: he ole o demog aphic and social ac o s
KaylaKol 1· SimonePika1
1 3
Animal Cogni ion (2025) 28:26
(app oxima ely 200 ms; S i e s e al. 2009), and o e laps
a e a oided (Sacks e al. 1974).
Tu n- aking is sugges ed o be a uni e sal en e p ise
ound ac oss human languages and cul u es (S i e s e al.
2009); howe e , he social iden i ies o he in e ac an s may
also in luence u n- aking, in line wi h he sociolinguis ic
Communica ion Accommoda ion Theo y (Giles 2008; Giles
and Powesland 1975). This heo y posi s ha indi iduals
adap hei communica ion s yles, o align o “accommo-
da e” wi h hei in ended ecipien based on ac o s like
social bonds (e.g., Fa ley e al. 2013). Fo ins ance, mo he s
ha e been shown o shape u n- aking in e ac ions wi h hei
in an s (e.g., G a ie e al. 2015). The e o e, in acco dance
wi h Communica ion Accommoda ion Theo y, u n- aking
may be in luenced by and adap ed o he social con ex and
ela ionships be ween in e ac an s and may ex end beyond
human communica ion.
Some schola s ha e p oposed ha u n- aking may p o-
ide an e olu iona y “missing link” in communica ion
be ween non-human species and humans (Le inson 2016),
whe e a mo e sophis ica ed pu ely signal in as uc u e ep-
esen s u n- aking in human con e sa ions (see Rossano
2018). Recen compa a i e s udies ha e shown ha some
o ms o u n- aking a e also p esen in non-human animals,
p ima ily p ima es, on hei exchanges, demons a ing o e -
lap a oidance and sequence o ganisa ion (e.g., F öhlich e
al. 2016; Ma in e al. 2017; Rossano 2013; Snowdon and
Cle eland 1984; Takahashi e al. 2013). Hence, some s ud-
ies ha e in es iga ed he in luence o demog aphic and
social ac o s on u n- aking. Fo example, a s udy on cap-
i e Campbell monkeys (Ce copihecus campbelli camp-
belli) sugges ed ha some elemen s o u n- aking migh be
linked o de elopmen and lea ning, wi h only adul indi id-
uals adhe ing o o e lap a oidance (Lemasson e al. 2011).
A s udy on wild whi e- aced capuchins (Cebus Capucinus)
e ealed ha i espec i e o sex, dominan indi iduals we e
mo e likely o elici a ocal esponse om hei conspeci -
ics (Digweed e al. 2007). Conce ning social bonds, a s udy
on zoo-housed bonobos (Pan paniscus) showed ha indi-
iduals esponded p e e en ially o he call o a conspeci ic
wi h whom hey sha ed a s ong social bond (Le e o e al.
2019). Howe e , hese s udies p ima ily ocused on pu ely
ocal exchanges.
The e is cu en ly a lack o unde s anding abou he ole o
demog aphic and social ac o s in o he u n- aking modali-
ies, such as ges u es and ac ions ( e e ing o any socially
di ec ed beha iou s ha lead o he in ended goal h ough
he di ec manipula ion o ano he ’s body o he mo emen
o one’s own body; F öhlich, Wi ig, e al., 2016). Howe e ,
his is essen ial o assesing he complexi y in ol ed, he
mul imodal beha iou al epe oi e o a social in e ac ion
(Mondada 2016), and hei simila i ies and di e ences o
human con e sa ional u n- aking.
Hence, he p esen s udy aimed o add ess his gap by
in es iga ing he impac o demog aphic and social ac o s
on he in as uc u e o non ocal u n- aking. Fo his pu -
pose, we ocused on he g ooming con ex o one o ou
closes li ing ela i es, chimpanzees (Pan oglody es).
G ooming, sugges ed o se e a social bonding unc ion
ha may ha e pa ed he way o language (Dunba 1996,
2004), p o ides a coope a i e con ex and includes a i -
o - a o ganisa ion (i.e., ole e e sal: Machanda e al.
2014; Mi ani 2006), nego ia ions, and a complex a ay o
ges u es (Pika 2014). I pe mea es nea ly e e y aspec o
chimpanzee socie ies (Goodall 1986), wi h g ooming in e -
ac ions, especially o male chimpanzees, being go e ned by
demog aphic and social ac o s such as age (e.g., younge
di ec ed o olde : Rosa i e al. 2020; Sandel e al. 2020),
dominance ank (e.g., lowe di ec ed o highe especially
in s eep-dominan socie ies: New on-Fishe and Lee 2011),
ela edness (e.g., di ec ed be ween ma e nal b o he s: San-
del e al. 2020), and social bonds (e.g., di ec ed o bonded
pa ne s: Mi ani 2009).
The e o e, his p esen s udy ocused on male eas -
e n chimpanzees (P. . schwein u hii) and hei g ooming
in e ac ions om he Ngogo popula ion in Kibale Na ional
Pa k, Uganda. Based on he e idence o he in luence o
demog aphic and social ac o s on ocal u n- aking o non-
human species and gene al engagemen in g ooming beha -
iou , we p edic ed ha hese ac o s may also in luence he
u n- aking in as uc u e o g ooming in e ac ions in chim-
panzees. In pa icula , we an icipa ed an inc eased likelihood
o u n ansi ions (i.e., ole al e na ion wi hin a g ooming
in e ac ion) when ini ia ed by highe - anking indi iduals,
and adul s (p ime indi iduals wi h highe social s and-
ing), as hey migh be mo e inclined o ecei e esponses,
especially om younge indi iduals o hose wi h a lowe
ank. In addi ion, we p edic ed an inc eased likelihood o
u n ansi ions be ween s ongly bonded dyads and ela ed
dyads. Fu he mo e, al hough dominance ank, social
bonds, and ela edness may a ec empo al ela ionship,
we speci ically p edic ed ha adul ecipien s would show a
p e e ence o a oiding empo al o e lap wi h u ns because
hey a e socially and in e ac i ely mo e expe ienced han
younge chimpanzees. These p edic ions a e embedded in
he Communica ion Accommoda ion Theo y and highligh
how communica ion is dynamic and esponsi e, shaped by
con ex ual ac o s and ela ionships among he chimpanzees
in ol ed.
To p o ide con ex o ou esea ch ques ions, we i s
examined u n ansi ions in ol ing ges u es and ac ions o
gain insigh in o hei equency, ypes, and empo al ela-
ionships o u n ansi ions wi hin chimpanzee g ooming
1 3
26 Page 2 o 14
Animal Cogni ion (2025) 28:26
in e ac ions. The u n ansi ions we e classi ied in o ou
ypes: ac ion – ac ion (e.g., A g ooms B, B g ooms A),
ac ion – signal (e.g., A g ooms B, B p oduces a ges u e), sig-
nal – ac ion (e.g., A p oduces a eques ges u e, B g ooms),
and signal – signal (e.g., A p oduces a ges u e, B p oduces a
ges u e, a possible o m o social nego ia ion). Thus, he ol-
lowing ques ions we e add essed in his s udy: Wha is he
ole o age, dominance ank, social bonds, and ela edness
in he (i) likelihood o u n ansi ions, (ii) equency o u n
ansi ions, (iii) ypes o u n ansi ions, and (i ) empo al
ela ionships o u n ansi ions?
Me hods
S udy si e and subjec s
Da a we e collec ed om male eas e n chimpanzees om
wo sub-communi ies (cen al and wes e n) o he Ngogo
popula ion li ing in Kibale Na ional Pa k, Uganda. The a ea
anges om d y-g ound and ea ly- o mid-s age colonis-
ing o es s o swamp o es s and an h opogenic g asslands
(Lwanga 2003). The Ngogo chimpanzee popula ion is a
p o oundly la ge communi y o ~ 200 indi iduals a he ime
o he s udy, consis ing o ~ 116 indi iduals in he cen al
communi y and ~ 84 indi iduals in he wes e n communi y
a he ime o his s udy.
We obse ed g ooming in e ac ions in 42 males (Ncen al
= 27, Nwes e n =15) anging in age om 13 o 56 yea s. Based
on he physical and social a ibu es o he de elopmen o
male chimpanzees (Goodall 1986; Kawanaka 1989), he
age o 13 was selec ed o include adolescen s in he s udy
(Goodall 1986; Pusey 1990; Reddy and Mi ani 2020).
This was o inco po a e age as a ac o in his s udy, which
included indi iduals who we e independen o hei mo h-
e s. Genealogies ha e p e iously been cons uc ed om
gene ic da a a Ngogo (Lange g abe e al. 2007), whe eby
indi iduals we e classi ied as ‘un ela ed’ i hey we e no
ma e nal siblings.
Da a collec ion
Da a we e collec ed om Sep embe 2021 o June 2022,
esul ing in 160 obse a ion days. We used a ocal sampling
app oach (Al mann 1974) using he Cybe acke so wa e
(Ve sion 3.51), while main aining a eco d o he o al du a-
ion o which a ocal was obse ed. In cases whe e mul-
iple ocal indi iduals we e p esen , p io i y was gi en o
hose who had been sampled less equen ly (i.e., lowe
obse a ion ime). Du ing he ocal samples, we also col-
lec ed scan da a e e y 15 min, no ing which indi iduals
we e wi hin he a m leng h (app oxima ely one me e ) o
he ocal. In addi ion, we used ad libi um sampling (Al -
mann 1974) o collec da a on all occu ences o g ooming
and agg essi e in e ac ions, excluding he ocal animal. The
obse a ional hou s pe ocal male anged om en o 41 h
(mean = 25.05 ± 10.95 h), esul ing in a o al o 1064 h (cen-
al communi y = 698.38 h, wes e n communi y = 365.81 h).
G ooming in e ac ions du ing he ocal ollow we e
eco ded using a digi al high-de ini ion came a (Sony
FDR-AX53 4 K) equipped wi h an ex e nal di ec ional
mic ophone (Sony ECM-GZ1M). Ad libi um eco dings o
g ooming in e ac ions, excluding he ocal, we e addi ion-
ally ob ained when he ocal was es ing. We selec ed a o al
o 311 g ooming in e ac ions (31.33 h), 190 in e ac ions o
he cen al communi y (20.01 h, mean/dyad = 1.76 ± 1.13),
and 121 in e ac ions o he wes e n communi y (11.31 h,
mean/dyad = 2.47 ± 1.88), based on he good isibili y o he
in e ac ions while accoun ing o a non-skewed dis ibu ion
o di e en dyads o add ess he esea ch ques ions. G oom-
ing in e ac ions we e di ided in o h ee phases: s a ing,
g ooming, and ending. The s a ing phase e e s o he ini-
ia ion o a g ooming in e ac ion by one o bo h o he in e -
ac an s h ough he p oduc ion o a signal (e.g., p esen 1) o
an ac ion (e.g., app oach), wi h ilming s a ing 30 s be o e
he g ooming session. Howe e , no all s a ing/ini ia ion/
solici a ion phases could be cap u ed, because no e e y
g ooming in e ac ion could be an icipa ed, and he in e ac-
ion had al eady s a ed when he obse e a i ed a he
loca ion. A g ooming phase (also e e ed o as a g ooming
‘session’) was de ined as a con inuous pe iod du ing which
an indi idual b ushes and manually picks h ough he hai
o ano he in e ac an wi h hei inge s o mou h (Goodall
1986), wi hou a change in beha iou (e.g., mo ing, eeding,
o es ing wi hou g ooming; New on-Fishe and Lee 2011).
A g ooming phase could en ail he swi ching o g ooming
oles (also e e ed o as “bou s”), whe e a g ooming ole is
de ined as an episode o pe iod o unidi ec ional g ooming
owa ds a ecipien who is no engaged in g ooming, ha is,
he engagemen o one in e ac an gi en owa ds a ecipien /
o he in e ac an wi hin a g ooming in e ac ion (Fos e e al.
2009). The g ooming phase may also be mu ual (o e lap-
ping g ooming oles, e.g., A g ooms B and B g ooms A).
The ending phase o an in e ac ion was de ined as he ime
poin when ei he he in e ac an ceased g ooming, engaged
in al e na i e ac i i ies, including es ing, o mo e han 30 s
(New on-Fishe and Lee 2011), o pe o med lea e- aking
ac ions.
1 Ges u es a e depic ed in small capi als.
1 3
Page 3 o 14 26
Animal Cogni ion (2025) 28:26
Da a analysis
Dominance ank
To de e mine he dominance hie a chy, we used only
dyadic in e ac ions wi h a clea ‘winne ’ and ‘lose ’ om
an agg essi e in e ac ion (To al Ncen al=180, Nwes e n=69),
along wi h wo gi en submissi e beha iou s (To al
Ncen al=254, Nwes e n=81), pan -g un s (Bygo 1979; Noe
e al. 1980), and pan ba k ocalisa ions (Rosa i e al.
2020). We hen c ea ed an in e ac ion ma ix om his
da ase and applied Ma Man analysis (de V ies e al. 1993)
in R o de e mine he o dinal anking. Males we e a ed
o dinally, wi h he highes - anked male ( he alpha male)
ecei ing a alue o 1 and lowe - anked males ecei ing
nume ically g ea e numbe s. The Ma Man analysis com-
pu es he imp o ed index o linea i y (h0) and accoun s
o ied and unknown ela ionships (de V ies 1995),
whe eby an index sco e o 0.90 o highe is an indica-
ion o a clea linea hie a chy (de V ies 1998). Linea
hie a chies di e among di e en chimpanzee communi-
ies, whe e he le el o despo ism is no uni o m ac oss
communi ies (Kabu u and New on-Fishe 2015). In some
communi ies, he gamma posi ions, in compa ison o he
alpha and be a posi ions, a e no as clea (Nakamu a e
al. 2015), which may be because adolescen males do no
o m a clea dominance hie a chy (Bygo 1979; Sandel
e al. 2017). The e o e, we applied a s eepness measu e
o assess he dis ibu ion o dominance, whe he i was
equally sha ed (egali a ian) o concen a ed among a ew
indi iduals (despo ic). The s eepness measu e conside s
he di e ences in agg essi e and submissi e in e ac ions
ac oss indi iduals and is based on no malised Da id’s
sco e. I was implemen ed using he s eepness package
and s eep es unc ion in R, esul ing in a sco e be ween
ze o and one, wi h ze o indica ing an egali a ian domi-
nance hie a chy and one ep esen ing a despo ic domi-
nance hie a chy.
Social bonds
To de e mine he s eng h o social bonds ac oss dyads,
we applied he Dyadic Composi e Sociali y Index (DSI)
o de e mine he s eng h o social ela ionships (Silk e
al. 2013). The DSI sco e was quan i ied h ough g ooming
equencies (mean pe dyad: Ncen al= 3.03 ± 3.26, Nwes e n=
5.19 ± 5.80), ex ac ed om he ocal da a, and spa ial
p oximi y equencies ex ac ed om he scan da a (mean
pe dyad: Ncen al= 4.46 ± 5.01, Nwes e n=7.62 ± 7.72) o he
da ase . Spa ial p oximi y pa e ns e lec he s eng h
o ela ionships, whe eas g ooming pa e ns e lec he
Coding
F om hese in e ac ions, he iden i ies o he in e ac an s,
dyad, and in e ac ion we e no ed, and we coded he ou
ollowing pa ame e s. Fi s , we coded ges u es (cha ac e -
ised as he signals) and ac ions (see Supplemen a y Ma e-
ial and Table S1 o de ini ions and de ails o each ges u e
and ac ion). Second, we coded u n ansi ions. Following
Hilb ink e al. (2015), u n ansi ions we e ope a ionalised
as a uni (signal/ges u e o ac ion, u n one) p oduced by
one indi idual and ecei ing a uni (signal/ges u e o ac ion)
om ano he indi idual, in ol ing ole e e sal (signale /
ac o - ecipien oles). Tu n ansi ions could ange om
none (i.e., unidi ec ional g ooming in e ac ion) o mul iple
wi hin a single g ooming in e ac ion, so we also coun ed
he equency o u n ansi ions pe in e ac ion. To no only
p edic possible ou comes bu also assess u n ansi ion da a
o po en ial biases, we examined hei equency o ensu e
a comp ehensi e and obus app oach. Thi d, we coded u n
ansi ion ypes, which could be ac ion – ac ion, ac ion –
signal, signal – ac ion, o signal – signal. Fou h, we coded
empo al ela ionships ( imings) be ween he uni s in u n
ansi ions, speci ically om he o se (end imes amp) o a
uni o he onse (beginning imes amp) o he ollowing uni
wi hin 30 s. O e lap a oidance ep esen s esponse la en-
cies ha ha e nega i e iming alues, whe eas posi i e im-
ing alues ep esen esponse la encies ha in ol e o e lap.
We addi ionally calcula ed he empo al ela ionships o
each u n ansi ion ype: ac ion – ac ion, ac ion – signal,
signal – ac ion, and signal – signal o be conside ed as asso-
cia ed (New on-Fishe and Lee 2011; Robe s e al. 2012).
In e - eliabili y
15% o all coded ideo clips we e eco ded by wo addi-
ional code s who we e blind o he esea ch objec i es,
and in e -obse e ag eemen was e alua ed o each coded
Tie using he EasyDIAg so wa e package (Holle and Rein
2015). This p og amme acili a es he assessmen o a e
ag eemen in ELAN, which is de e mined h ough a com-
posi e measu e o bo h he anno a ion alue and du a ion
o a coded segmen be ween a e s. To es ablish ag eemen ,
he du a ion o he coded segmen s was equi ed o o e -
lap by a leas 60%, and he anno a ion alues needed o be
iden ical (Holle and Rein 2015). The esul s o he es s indi-
ca ed a ‘good’ le el o ag eemen be ween he a e s ( a e 1,
Cohen’s κ = 0.81; a e 2, Cohen’s κ = 0.88) and he espec-
i e code . In ins ances whe e ag eemen was no achie ed,
hese cases we e e-e alua ed by he espec i e code and
discussed be ween he code and he a e . Once ag eemen
was a ained, hese ins ances we e esol ed.
1 3
26 Page 4 o 14
Animal Cogni ion (2025) 28:26
we e omi ed om he inal model. To accoun o possi-
ble communi y di e ences, we added communi y iden i y
(cen al o wes e n) as a con ol e ec . To a oid pseudo-
eplica ion, he andom e ec s we e he ID o he ini ia o ,
ecipien , dyad and in e ac ion.
To in e p e he s eng h and unce ain y o he es ima ed
e ec s, he median es ima e and he median absolu e de ia-
ion (MAD) we e epo ed, along wi h he 89% Bayesian
c edible in e al (89% C I) and he p obabili y o di ec-
ion (pd). The p obabili y o di ec ion (posi i e o nega i e)
anges om 50% (high unce ain y abou he di ec ion) o
100% (s ong ce ain y abou he di ec ion), and indica es
he ce ain y wi h which an e ec goes in a pa icula di ec-
ion (Makowski e al. 2019).
Model diagnos ics was pe o med by examining he
model summa ies, which indica ed ha he pos e io dis i-
bu ions accu a ely ep esen ed he o iginal esponse alues.
The R-ha s a is ics we e deemed accep able, wi h alues
o < 1.05. Fu he mo e, we con i med adequa e e ec i e
sample sizes (Bulk_ESS and Tail_ESS) o > 1000, and no
di e gen ansi ions in he MCMC chains. We also assessed
model con e gence by isually inspec ing ace plo s, his-
og ams o pos e io dis ibu ions, and checking o au o-
co ela ion be ween i e a ions (see Supplemen a y Figu e
S2-S10; Depaoli and an de Schoo 2017).
Model 1: in luence o demog aphic and social ac o s on u n
ansi ions To es whe he demog aphic and social ac o s
a ec ed he likelihood o u n ansi ions, e e y occu ence
o uni (signal/ac ion) p oduc ion (N = 6334) ac oss all in e -
ac ions was included as a da a poin . The esponse a iable
included whe he he e was an occu ence (binomial: yes/
no) o a u n ansi ion ( i ing a Be noulli dis ibu ion), and
whe he he p oduced uni en ailed a consequen uni p o-
duced by he ecipien . The in e ac ion ID was added as a
andom e ec .
Model 2: in luence o demog aphic and social ac o s on he
equency o u n ansi ions To es whe he demog aphic
and social ac o s in luenced he equency o u n ansi-
ions in a gi en in e ac ion, each in e ac ion (N = 311) was
conside ed a da a poin , whe e he equencies o u n an-
si ions we e allied pe in e ac ion. The esponse a iable
was he numbe o u n ansi ions ( i ing a Poisson dis-
ibu ion). The log o he o al dyadic in e ac ion ime was
in eg a ed as an o se e m ( o accoun o he numbe o
quali y o he ela ionships (Mi ani 2009). The DSI was
calcula ed using he ollowing equa ion:
DSI =
(
GA+B
GA+B+GA+GB
)+(
PA+B
PA+B+PA+PB
)
GA+B is he combined g ooming equency o indi idual
A owa ds indi idual B and ice e sa, B o A. GA ep e-
sen s he o al g ooming equency obse ed by indi idual
A in he absence o indi idual B, and GB ep esen s he o al
g ooming equency by indi idual B in he absence o indi-
idual A. Simila ly, PA+B is he numbe o scans whe e indi-
iduals A and B we e obse ed o be in p oximi y o each
o he (wi hin one me e ). PA and PB ep esen he o al num-
be o scans o indi iduals A and B, espec i ely, exclud-
ing he o he indi idual (A o B). Highe alues ep esen
s ongly bonded dyads ha had mo e equen in e ac ions
han he a e age dyad in hei communi y. Low alues ep-
esen weakly bonded dyads wi h in equen in e ac ions
compa ed wi h he a e age dyad o he communi y.
Implemen ed models
In con as o equen is s a is ics, which calcula e he p ob-
abili y o obse ing da a unde he null hypo hesis, Bayesian
s a is ics o e insigh s in o he c edibili y o he pa ame e s
used gi en he obse ed da a (K uschke 2014; McEl ea h
2018). The e o e, a se ies o Bayesian gene alised mixed
models was implemen ed in R ( e sion 4.4.2; R Co e Team
2022) using he unc ion “b m” om he b ms package
(Bü kne 2017) o es he p edic ions. Each model included
ou Ma ko chain Mon e Ca lo (MCMC) chains wi h 4000
i e a ions pe chain and speci ied 1000 wa m-up i e a ions,
leading o 16000 pos e io samples. Fo all models, weakly
in o ma i e p io s we e used o p e en o e i ing ( an de
Schoo e al. 2021) and educe in e en ial e o s. Fo in e -
ence, we calcula ed 89% c edi in e als (uppe and lowe )
om he pos e io dis ibu ions and checked whe he ze o
was included in his in e al (K uschke 2014; McEl ea h
2018). He ein, we conside ed ha , i pos e io dis ibu-
ions shi ed subs an ially away om ze o in one di ec ion,
as opposed o cen e ing on ze o (i.e., which would indica e
he null expec a ion o pos e io dis ibu ions), he e was
e idence o an e ec in a speci ic di ec ion (posi i e o
nega i e).
Fo each model (see Supplemen a y Ma e ial o he
gene al model o mula, Figu e S1), he ixed e ec s we e
ini ia o ’s age, ecipien ’s age, ini ia o ’s dominance ank,
ecipien ’s dominance ank, social bond s eng h (DSI), and
ela edness (0 “no ela ed”/1 “ ela ed”). All ixed e ec s,
excep ela edness, we e z- ans o med o imp o e he in e -
p e a ion (Schielze h 2010). Based on ou p edic ions, we
included in e ac ions be ween age and dominance ank (six
pai wise in e ac ions). In e ac ions ha showed no e ec
1 3
Page 5 o 14 26

Animal Cogni ion (2025) 28:26
Dominance ank and social bonds
The dominance hie a chy o he wo communi ies e ealed
a signi ican dominance s uc u e. Howe e , anks wi hin
his hie a chy we e no s ic ly de ined (linea i y index: cen-
al: h’ = 0.417, p = 0.001; wes e n: h’ = 0.355, p = 0.047),
wi h he h’ index being less han 0.90 o bo h communi ies.
The slope s eepness (cen al: 0.17, p < 0.001; wes e n: 0.19,
p < 0.001), in line wi h he dominance hie a chy, sugges ed
ha he ank di e ences be ween consecu i e posi ions in
bo h communi ies we e ela i ely small (e.g., no s ong
dominance be ween posi ions 4 and 5). These esul s indi-
ca e modes powe di e ences be ween adjacen ly anked
indi iduals, and hus, a mo e egali a ian s yle in his popu-
la ion a he ime o he s udy. The DSI alues ac oss 157
dyads om bo h communi ies anged om 0 o 0.2069.
Demog aphic and social ac o s on he likelihood
and equency o u n ansi ions
Model 1 showed ha ini ia o ’s age, ecipien ’s age, and
ecipien ’s ank in luenced he likelihood o u n ansi ions
(Fig. 1; see Supplemen a y Ma e ial Table S3). Tu n ansi-
ions we e mo e likely when hey we e ini ia ed by an olde
indi idual (es ima e [MAD] = 0.30 [0.08], 89% CI [0.20,
0.41], pd = 100%), when hey in ol ed younge ecipien s
(es ima e [MAD] = − 0.34 [0.09], 89% CI -0.46, -0.23],
pd = 100%), and when ecipien s had a lowe ank (es ima e
[MAD] = 0.20 [0.09], 89% CI 0.09, 0.32], pd = 98.01%).
The in e ac ion be ween ini ia o ’s age and ecipien ’s ank
showed s ong e idence o an e ec (es ima e [MAD] = 0.10
[0.04], 89% CI 0.04, 0.15], pd = 98.61%). Howe e , he
social bond s eng h and ela edness did no in luence he
likelihood o u n ansi ions. These indings a e pa ially
suppo he p edic ion, whe e ou indings con i med ha
u n ansi ions a e mo e likely when ini ia ed by olde indi-
iduals and when ecipien s a e lowe - anking and younge .
Howe e , ou indings did no suppo he p edic ion ha
u n ansi ions a e mo e p obable when ini ia ed by highe -
anking indi iduals, and be ween s ongly bonded dyads.
Model 2 showed ha ecipien ’s age had an impac on
he numbe o u n ansi ions (Fig. 2, see Supplemen-
a y Ma e ial Table S4). Speci ically, u n ansi ions we e
mo e equen when ecipien s we e younge (es ima e
[MAD] = − 0.11 [0.05], 89% CI -0.16, -0.05, pd = 66.61%).
Howe e , he pd was close o 50%, sugges ing ela i ely
weak e idence o he di ec ion o he e ec . The o he
e ec s had no in luence on he equency o u n ansi ions
(see Supplemen a y Ma e ial, Table S3). In ela ion o ou
p edic ion, despi e he weak e idence o he e ec and ind-
ing no in luence o he ini ia o , we ind pa ial suppo o
he p edic ion ha u n ansi ions a e mo e equen when
u n ansi ions ela i e o he in e ac ion ime). The in e ac-
ion ID was added as a andom e ec .
Model 3: in luence o demog aphic and social ac o s on u n
ansi ion ypes To es whe he demog aphic and social
ac o s a ec ed he likelihood o di e en u n ansi ion
ypes, a mul i a ia e esponse model was de eloped using
each u n ansi ion (N = 2275) as an indi idual da a poin . In
his model, each u n ansi ion ype was ans o med in o a
binomial a iable by con e ing each ype in o a nume ical
a iable wi hin he da a ame and subsequen ly agg ega ing
hem o o m he esponse a iable. This app oach allowed
each u n ansi ion ype o be ea ed as a dis inc bina y
a iable in he same model. The in e ac ion ID was added
as a andom e ec .
Model 4: in luence o demog aphic and social ac o s on
empo al ela ionships To es whe he demog aphic and
social ac o s in luenced empo al ela ionships, each u n
ansi ion was conside ed a da a poin (N = 2275). The im-
ings we e log- ans o med o no malise he da a and educe
he e oscedas ici y, and we e inpu as he esponse a iable
( i ing a Gaussian dis ibu ion). Sum- o-ze o con as was
applied p io o he u n ansi ion ypes o acili a e in e -
p e a ion and ob ain he g and mean as he e e ence le el
ins ead o one o he ou le els (ac ion – ac ion, ac ion –
signal, signal – ac ion, and signal – signal). The u n an-
si ion ypes we e addi ionally added as a ixed e ec in
in e ac ion wi h all o he ixed e ec s, o accoun o he
in luence o demog aphic and social ac o s on he empo al
ela ionships ac oss each ype o u n ansi ion.
Resul s
Ou analysis o he 311 g ooming in e ac ions ac oss 157
dyads yielded 2275 u n ansi ions ou o 6334 occu ences
o signals o ac ions. Among hese in e ac ions, 275 in ol ed
a leas one u n ansi ion (N = 19 in e ac ions; 0.07% wi h
only one u n ansi ion). O he 2275 u n ansi ions, 38%
we e classi ied as ac ion – ac ion (N = 872), 32% as signal
– ac ion (N = 732), 16% as ac ion – signal (N = 353), and
14% as signal – signal (N = 318). Fu he mo e, in 47% o
he 2275 u n ansi ions he e was o e lap a oidance. In
pa icula , ac ion – ac ion u n ansi ions (median = -0.14 s)
we e p edominan ly cha ac e ised by o e lap a oidance,
whe eas o e lap was mainly obse ed in ac ion – signal
(median = 0.10 s), signal – ac ion (median = 0.06 s), and sig-
nal – signal (median = 0.74 s) u n ansi ions.
1 3
26 Page 6 o 14
Animal Cogni ion (2025) 28:26
we e mo e likely when he ecipien was younge (es ima e
[MAD] = -0.12 [0.07], 89% CI [-0.20, -0.03], pd = 94.60%;
Fig. 3a). Ac ion – signal u n ansi ions we e mo e p obable
when he ini ia o was younge (es ima e [MAD] = -0.11
[0.09], 89% CI [-0.22, -0.01], pd = 90.65%; Fig. 3b). Sig-
nal – ac ion u n ansi ions we e mo e p obable when hey
in ol ed lowe - anking ecipien s (es ima e [MAD] = 0.13
[0.09], 89% CI [0.02, 0.24], pd = 92.61%; Fig. 3c) and dyads
wi h weake bonds (es ima e [MAD] = -0.11 [0.08], 89% CI
in ol ing younge ecipien s, bu no suppo ha ecipien s
a e o lowe - ank.
Demog aphic and social ac o s impac on u n
ansi ion ypes
Model 3 showed ha he e was an e ec o age, domi-
nance ank, and social bond s eng h ac oss he u n an-
si ion ypes. Speci ically, ac ion – ac ion u n ansi ions
Fig. 2 The plo displays he
s ong e ec s om Model 2,
showing he equency o u n
ansi ions in ela ion o ecipi-
en ’s age. The e o ba s ep esen
he s anda d e o , and he ci cles
indica e he means, wi h hei
size e lec ing he numbe o
obse a ions. The lines ep esen
a linea eg ession model i , and
he shaded a ea indica es he
con idence in e al
Fig. 1 The plo s display he s ong e ec s om Model 1, showing
he p edic ed p obabili y o u n ansi ions in ela ion o (a) ini ia-
o ’s age, (b) ecipien ’s age, and (c) ecipien ’s ank. The e o ba s
ep esen he s anda d e o , and he ci cles indica e he means, wi h
hei size e lec ing he numbe o obse a ions. The lines ep esen
a linea eg ession model i , and he shaded a ea indica es he con i-
dence in e al
1 3
Page 7 o 14 26
Animal Cogni ion (2025) 28:26
ha lowe - anking ecipien s we e mo e likely o be in ol ed
in signal – ac ion u n ansi ions, and younge ecipien s in
ac ion–ac ion u n ansi ions suppo s he p edic ion ha
younge indi iduals and hose wi h lowe ank a e mo e
likely o gi e esponses. Howe e , ou esul s ha signal –
signal u n ansi ions we e mo e likely when hey in ol ed
ha olde ecipien s and ac ion – signal ansi ions we e
mo e likely when hey in ol ed younge ini ia o ’s, showed
he opposi e o ou p edic ion ha u n ansi ions a e
mo e p obable wi h adul ini ia o s and younge ecipien s.
Addi ionally, we p edic ed ha dyads ha a e ela ed and
hose ha ha e s onge social bonds would ha e a highe
[-0.20, -0.01], pd = 91.61%; Fig. 3d). Signal – signal u n
ansi ions we e mo e p obable when hey in ol ed olde
ecipien s (es ima e [MAD] = 0.26 [0.09], 89% CI [0.15,
0.37], pd = 99.86%; Fig. 3e). The o he ixed e ec s did no
in luence he likelihood o u n ansi ion ypes (see Supple-
men a y Ma e ial, Table S5). In addi ion, ac ion – ac ion
u n ansi ions we e less p obable in he wes e n commu-
ni y (es ima e [MAD] = -0.32 [0.20], 89% CI [-0.57, -0.07],
pd = 94.29%), whe eas signal – ac ion u n ansi ions
we e mo e p obable in he wes e n communi y (es ima e
[MAD] = 0.30 [0.21], 89% CI [0.05, 0.56], pd = 92.21%).
Ou indings pa ially suppo ou p edic ions. The inding
Fig. 3 Plo s display he s ong
e ec s om Model 3. Each plo
showing he p edic ed p obabili y
o (a) ac ion – ac ion u n ansi-
ions in ela ion o ecipien ’s age,
(b) ac ion – signal u n ansi-
ions in ela ion o ini ia o ’s age,
signal – signal u n ansi ions in
ela ion o (c) ecipien ’s ank and
(d) social bond s eng h (DSI),
and signal – signal u n ansi-
ions in ela ion o ecipien ’s
age. The e o ba s ep esen he
s anda d e o , and he ci cles
indica e he means, wi h hei
size e lec ing he numbe o
obse a ions. The lines ep esen
a linea eg ession model i , and
he shaded a ea indica es he
con idence in e al
1 3
26 Page 8 o 14
Animal Cogni ion (2025) 28:26
males (e.g., Hayaki 1990; Sandel e al. 2017), and senio
males a e subo dina e o adul males (e.g., Bake 2000; Hay-
aki e al. 1989). Ou indings u he suppo his obse a-
ion, wi h u n ansi ions being mo e p obable when olde
indi iduals ini ia ed a u n in ol ing lowe - anking males,
and when u n ansi ions in ol ed younge ecipien s (i.e.,
adolescen s). Thus, younge and lowe - anking indi iduals
(i he ini ia o was olde ) we e mo e likely o espond. I
is possible ha lowe - anking males exchange beha iou s
such as g ooming o acqui e social knowledge (e.g., iden-
i ying po en ial allies o social in eg a ion), in line wi h
he Biological Ma ke Theo y (Noë and Hamme s ein 1994,
1995). This heo y posi s ha in e ac ions be ween indi idu-
als can be concep ualised as a ma ke , whe e in e ac ions
possess in insic alue and indi iduals a e ega ded as ad-
e s whose selec ion o social pa ne s is con ingen upon
he bene i s hese pa ne s may p o ide. In his line, a s udy
on eas e n chimpanzees showed ha olde males engaged
in agg essi e in e ac ions wi h conspeci ics less equen ly
han o he age classes (Rosa i e al. 2020), sugges ing ha
esponding o hese indi iduals is less likely o esul in a
possible agg essi e in e ac ion o lowe - anking males.
Addi ionally, as he dominance o he wo communi ies o
he Ngogo popula ion a he ime o he s udy was no s eep,
agonis ic suppo may no ha e been an impo an commod-
i y (Kabu u and New on-Fishe 2015). Thus, ini ia o domi-
nance ank may no be ele an , explaining he possible
lack o an e ec o his ac o . The highe likelihood and
equency o u n ansi ions wi h younge ecipien s can
be a ibu ed o adolescen s’ endency o in e ac mo e wi h
olde indi iduals, as obse ed in a p e ious s udy on he
Ngogo popula ion (Sandel e al. 2020). These young indi-
iduals may be a emp ing o build connec ions wi h olde
males a he han wi h hei pee s, as a means o g adually
in eg a ing hemsel es in o he adul male dominance hie -
a chy. This is consis en wi h p e ious s udies on g ooming
in e ac ions in eas e n chimpanzees, p o iding e idence ha
adolescen s (younge indi iduals) in e ac selec i ely wi h
adul males a he han wi h indi iduals o a simila age o
ank (Hayaki 1988; Kawanaka 1993; Pusey 1990). In ela-
ion o u n- aking, ou indings mi o he esul s o s udies
on call exchanges in o he p ima e species. Fo example,
calls om olde indi iduals a e mo e likely o elici ocal
esponses bo h in cap i e Campbell monkeys (Lemasson
e al. 2010) and cap i e common ma mose s (Calli h ix
jacchus; Chen e al. 2009). Conce ning dominance ank,
s udies ha e shown ha dominan indi iduals ecei ed ela-
i ely mo e esponses han subo dina e ones as o example
in cap i e and wild whi e- aced capuchins (Digweed e al.
2007) and cap i e wes e n lowland go illas (Go illa go illa
go illa; Lemasson e al. 2018).
likelihood o u n ansi ions, bu his was no suppo ed, as
weakly bonded dyads we e mo e likely o engage in signal
– ac ion ansi ions and he e was no e ec o ela edness.
Demog aphic and social ac o s impac on empo al
ela ionships
Model 4 showed no e ec o demog aphic and social ac o s
on he o se – onse imings (see Supplemen a y Ma e ial
Table S6). Ou p edic ion ha social and demog aphic ac-
o s migh in luence he empo al ela ionship, speci ically
ha o e lap a oidance would be mo e likely when u n an-
si ions in ol ed adul ecipien s, was no suppo ed by ou
indings.
Discussion
This s udy aimed o assess he impac o di e en demo-
g aphic and social ac o s on non ocal u n- aking in
non-human p ima es. We ocused on male chimpanzee
in e ac ions in he coope a i e con ex o g ooming, o
in es iga e he impac o age, dominance ank, ela edness,
and social bonds on he (i) likelihood o u n ansi ions, (ii)
equency o u n ansi ions, (iii) likelihood o u n an-
si ion ypes, and (i ) empo al ela ionships. We p edic ed
an inc ease in he likelihood o u n ansi ions when hese
we e ini ia ed by highe - anking indi iduals, and adul s as
hese indi iduals can ha e o ha e highe social s anding
and migh be mo e inclined o ecei e esponses, especially
by ecipien s o lowe ank and younge age. Fu he mo e,
ha he u n ansi ion likelihood would be mo e p obable
be ween s ongly bonded dyads and ela ed dyads. Hence,
we p edic ed adul ecipien s would mo e likely a oid em-
po al o e laps. O e all, ou indings showed ha u n an-
si ions and ansi ion ypes in g ooming in e ac ions we e
in luenced by demog aphic and social ac o s, bu no by
ela edness. None o hese ac o s had an e ec on empo-
al ela ionships. These indings pa ially suppo he p e-
dic ions o his s udy. In he ollowing sec ions, we discuss
hese esul s in de ail.
The ole o demog aphic and social ac o s on u n
ansi ions
Ou esul s showed ha ini ia o ’s age a ec ed he likeli-
hood o u n ansi ions con ingen on he dominance ank
o he ecipien and ice e sa, and he age o he ecipien
a ec ed he likelihood o u n ansi ions. This inding may
be due o he age and dominance ank o chimpanzee males
being linked o each o he (Rod igues e al. 2022), whe e
adolescen males a e unequi ocally subo dina e o adul
1 3
Page 9 o 14 26