B ain & Language 229 (2022) 105106
A ailable online 4 Ap il 2022
0093-934X/© 2022 The Au ho s. Published by Else ie Inc. This is an open access a icle unde he CC BY license (h p://c ea i ecommons.o g/licenses/by/4.0/).
The one a las o pe cep ual disc iminabili y and pe cep ual dis ance: Fou
one languages and i e language g oups
Liquan Liu
a
,
b
,
c
,
d
,
*
, Regine Lai
e
, Lehe Singh
, Ma ina Kalashniko a
b
,
g
,
h
, Pa ick C.M. Wong
e
,
i
,
Benjawan Kasisopa
b
, Ao Chen
j
, Chu amanee Onsuwan
k
, Denis Bu nham
b
,
*
a
School o Psychology, Wes e n Sydney Uni e si y, Aus alia
b
The MARCS Ins i u e o B ain, Beha iou and De elopmen , Wes e n Sydney Uni e si y, Aus alia
c
Cen e o Mul ilingualism in Socie y Ac oss he Li espan, Uni e si y o Oslo, No way
d
Cen e o Excellence o he Dynamics o Language, Aus alian Resea ch Council, Aus alia
e
Depa men o Linguis ics and Mode n Languages, The Chinese Uni e si y o Hong Kong, Hong Kong
Depa men o Psychology, Na ional Uni e si y o Singapo e, Singapo e
g
Basque Cen e on Cogni ion, B ain and Language, Spain
h
Ike basque, Basque Founda ion o Science, Spain
i
B ain and Mind Ins i u e, The Chinese Uni e si y o Hong Kong, Hong Kong
j
School o Communica ion Sciences, Beijing Language and Cul u e Uni e si y, China
k
Depa men o Linguis ics, Facul y o Libe al A s and Cen e o Excellence in In elligen In o ma ics, Speech and Language Technology, and Se ice Inno a ion (CILS),
Thammasa Uni e si y, Thailand
ARTICLE INFO
Keywo ds:
Tone
C oss-linguis ic pe cep ion
Tone sys em
Con as ype
Pe cep ual asymme y
Cue-weigh ing
Mul i-dimensional analysis
ABSTRACT
Some p io in es iga ions sugges ha one pe cep ion is lexible, easonably independen o na i e phonology,
whe eas o he s sugges i is cons ained by na i e phonology. We add ess his issue in a sys ema ic and
comp ehensi e in es iga ion o adul one pe cep ion. Sampling om di e se one and non- one speaking
communi ies, we es ed disc imina ion o he h ee majo one sys ems (Can onese, Thai, Manda in) ha
domina e he one pe cep ion li e a u e, in ela ion o na i e language and language expe ience as well as
s imulus a ia ion ( one p ope ies, p esen a ion o de , pi ch cues) using linea mixed e ec modelling and
mul idimensional scaling. The e was an o e all disc imina ion ad an age o one language speake s and o
na i e ones. Howe e , language- and one-speci ic e ec s, and p esen a ion o de e ec s also eme ged. Thus,
o e and abo e na i e phonology, s imulus a ia ion exe s a powe ul in luence on one disc imina ion. This
s udy p o ides a one a las, a e e ence guide o in o m empi ical s udies o one sensi i i y, bo h e ospec i ely
and p ospec i ely.
1. In oduc ion
A ound 60–70% o he wo ld’s languages a e one (o pi ch-accen )
languages – hey use pi ch o di e en ia e wo d meanings (Yip, 2002).
O e hal o he wo ld’s popula ion speaks a one language (F omkin,
1978). Despi e his ubiqui y, he de e minan s o one pe cep ion a e no
well unde s ood. Some sugges one pe cep ion is lexible, easonably
independen o na i e phonology, whe eas o he s sugges i is con-
s ained by na i e phonology o pe cep ual asymme ies. This s udy
aims o in es iga e he ela i e oles o he pe cei e , he s imulus and
he con ex in lexical one pe cep ion. To his end, he s udy in ol es a
comp ehensi e in es iga ion o how di e en ones in di e en one
languages a e pe cei ed by lis ene s om di e en language back-
g ounds in di e en ask con ex s.
In e ms o he classi ica ion o lexical ones, unlike o he phone ic
uni s (consonan s, owels), which a e o en de ined by a icula o y
ges u es, ones a e de ined mos ly by hei pi ch cha ac e is ics (Bu n-
ham, A ina, Xu, & Kasisopa, 2011). Fundamen al equency (
0
), he
acous ic basis o pi ch pe cep ion, is a widely ecognized index o one
di e ences (Ba y & Blamey, 2004), e en hough ones a e also di e -
en ia ed, o a lesse ex en , on o he dimensions, such as du a ion,
ampli ude, and oice egis e (Feng, Wu, & Nissenbaum, 2020). Gi en
he p edominance o pi ch in he pe cep ion o one, Chao (1930)
in oduced a sys em o desc ibe indi idual ones, whe eby each one is
* Co esponding au ho s.
E-mail add esses: [email p o ec ed] (L. Liu), [email p o ec ed] (D. Bu nham).
Con en s lis s a ailable a ScienceDi ec
B ain and Language
jou nal homepage: www.else ie .com/loca e/b&l
h ps://doi.o g/10.1016/j.bandl.2022.105106
Recei ed 15 Ma ch 2021; Recei ed in e ised o m 2 Ma ch 2022; Accep ed 8 Ma ch 2022
B ain and Language 229 (2022) 105106
2
desc ibed on a no malized pi ch scale om 1 (low) o 5 (high), one alue
a he s a , ano he a he end, and a hi d medial alue when he e is an
in lexion poin wi hin he one. As his index is no malized, i is inde-
penden o he language and speake and is used he e o desc ibe and
compa e ones wi hin and be ween languages.
The co e issue in s udying pe cep ual disc iminabili y and pe cep ual
dis ance lies in he unde s anding o he de e minan s o one disc im-
ina ion. Lis ene s’ one pe cep ion has been shown o be in luenced by
mul iple ac o s. He e, we discuss he e ec s o he pe cei e ( one
language expe iences); he s imulus (including one sys ems, con as
ype) and he s imulus i sel ( one p ope ies and cues); and he ask
con ex (o de o s imulus p esen a ion) on one pe cep ion.
Like he pe cep ion o o he phone ic uni s, one pe cep ion is shaped
by linguis ic expe ience om he beginning o li e (Fikke , Liu, & O a,
2020; We ke , 2018). Unlike consonan s and owels o which a clea
de elopmen al ajec o y has been epo ed in ol ing pe cep ual na -
owing, i.e., om lexible (c oss-language, phone ic) o cons ained
(language-speci ic, phonological) pe cep ion, he pic u e is mixed o
ones. Some s udies ha e demons a ed pe cep ual na owing, indexed
by dec eased onal sensi i i y o e age in non-na i e one lea ne s
(Ma ock & Bu nham, 2006; Ma ock, Molna , Polka, & Bu nham, 2008;
Quam & Swingley, 2010). O he s ha e epo ed acili a ion, an age-
based inc eased sensi i i y by bo h one and non- one language
lea ning in an s (Chen & Kage , 2016; Chen, S e ens, & Kage , 2017;
Ramache s, B ouwe , & Fikke , 2018; Singh e al. 2018; Tsao, 2017).
Ye o he s udies epo esul s ha a es o he lexibili y o one
pe cep ion – a enewed sensi i i y o one by non- one language lea ning
in an s in hei second yea (G¨
o z, Yeung, K aso kina, Schwa ze , &
H¨
ohle, 2018; Liu & Kage , 2014, 2017). Bu nham and Singh (2018)
poin ou ha he mixed esul s a e he p oduc o s udies using di e en
one pai s and di e en one languages and could well be due o he
psychophysics o one salience. Acco dingly, mo e acous ically salien
ones may be mo e esilien o lis ene s ac oss language backg ounds in
he cou se o pe cep ual na owing. Howe e , he measu e o inhe en
onal salience has been di icul and is an issue o con o e sy, exace -
ba ed by he absence o one pe cep ion me ics solidly g ounded on
a icula o y ges u es.
The e ec o language backg ound is e iden in s udies o adul lis-
ene s’ one pe cep ion, in which one language speaking adul s show
dis inc and enhanced one disc imina ion compa ed wi h hei non-
one language-speaking pee s (Bu nham & F ancis, 1997; Kaan,
Wayland, Bao, & Ba kley, 2007; Malins & Joanisse, 2012; also see
Maggu, Zong, Law, & Wong, 2018 and Tong, Lee, Lee, & Bu nham, 2015
o he e ec o monolingual e sus bilingual one and non- one lan-
guages). In a s udy measu ing Thai one pe cep ion by speake s o Thai,
Can onese, Manda in ( one languages), Swedish (pi ch-accen ) and En-
glish, be e pe cep ion was e iden in na i e speake s o a one/pi ch-
accen language o e na i e non- one (English) speake s (Bu nham
e al., 2015).
Di ec compa isons o one language speake s’ na i e e sus non-
na i e one pe cep ion a e a e, bu indi ec e idence sugges s di e -
ences in he pe cep ion o na i e and non-na i e ones. As ea ly as 4
mon hs a e bi h, Can onese- and Manda in-lea ning in an s display
language-speci ic pe cep ion o Can onese ones (Yeung, Chen, &
We ke , 2013). Such an e ec also appea s o ca y o e in o adul hood:
While Thai speake s ha e no di icul y iden i ying Thai ones, Manda in
and Can onese speake s’ Thai one pe cep ion appea s o be modula ed
by hei own onal in en o y (Reid e al., 2015). I has been a gued ha
lis ene s pe o m be e when non-na i e ones can be assimila ed o
hei na i e one ca ego ies (Chen, Bes , & An oniou, 2020), hus
adhe ing o assimila ion accoun s o na i e e sus non-na i e speech
pe cep ion (e.g., Pe cep ual Assimila ion Model, Bes , 1995). These
combined esul s s ongly sugges , bu do no di ec ly es he hypo h-
esis, ha one language speake s’ pe cep ion o na i e ones should be
supe io o hei pe cep ion e sus non-na i e ones In his pape , we
will di ec ly es he hypo hesis ha one language speake s’ pe cep ion
o na i e ones will be be e han hei pe cep ion o non-na i e ones.
Tone language speake s’ p ocessing o ones in ol es no only he
bila e al on opa ie al b ain egions used o acous ic analysis and
abs ac ion (Gandou , Dzemidzic, e al., 2003; Gandou , Wong, e al.,
2003; Gandou e al., 2004; Li e al., 2010), bu also on o empo al
egions used o p ocessing phonological and seman ic in o ma ion
(Kwok, Dan, Yakpo, Ma hews, & Tan, 2016; Kwok e al., 2017; Liang &
Du, 2018; Wong e al., 2004). In addi ion, one p ocessing in lis ene s
om one language backg ounds becomes mo e le -la e alised when
language comp ehension asks e oke mo e seman ic p ocessing (Gan-
dou , Dzemidzic, e al., 2003; Gandou , Wong, e al., 2003; Gandou ,
Wong, & Hu chins, 1998; Gandou e al., 2004). Mo eo e , Chen, Pe e ,
e al. (2018) ound ha while one (Manda in Chinese) and non- one
(Du ch) language adul s show simila igh -la e alised misma ch nega-
i i y (MMN) o 3-no e musical melodies based on Manda in ones, o
he lexical ones on which he
0
o hese 3-no e melodies we e based,
Chinese adul s showed a la e MMN peak o lexical one oddballs han
did non- one Du ch adul s. These esul s imply ha , consis en wi h
indings ha one language speake s p ocess ones ca ego ically (Feng,
Gan, Wang, Wong, & Chand aseka an, 2018; Gandou & K ishnan,
2016; Peng e al., 2010; Xi, Zhang, Shu, Zhang, & Li, 2010), a la ge
0
di e ence is necessa y o Chinese han o Du ch adul s o de ec he
lexical one change. Mo eo e , Chen and colleagues (Chen, e al., 2018;
Chen, Liu, & Kage , 2016; Liu, Chen, & Kage , 2020) ound signi ican
co ela ions be ween lexical one and music pi ch o non- one, bu no
one language speake s. I is known ha adul -like igh -hemisphe ic
la e alisa ion o piano ones de elops ea ly a 2 mon hs a e bi h (He,
Ho son, & T aino , 2007), and he e o e, gi en his and he abo e
indings, i appea s ha a he neu ophysiological le el non- one lan-
guage adul s pe cei e lexical ones in a simila ashion as hey do non-
lexical (musical) pi ch a ia ions, whe eas adul one language
speake s shi he manne o p ocessing o lexical (bu no musical) ones
some ime a e ea ly in ancy, and his acili a es he ca ego isa ion o
0
modula ions ha signal phonological dis inc ions.
A he beha iou al le el, bo h he absence and p esence o one-
ele an expe ience a ec s non- one language speake s’ lexical one
pe cep ion. On he one hand, he absence o one- ele an expe ience
limi s one disc imina ion (Bu nham & Ma ock, 2007). Bu nham e al.
(2015) showed ha lexical ones p esen ed in a speech con ex we e less
well disc imina ed by non- one (English) language lis ene s han by one
o pi ch-accen lis ene s, bu once he e y same
0
pi ch con ou s we e
p esen ed as in a non-linguis ic con ex , as hums o iolin glides, di -
e ences in he pe cep ual pe o mance be ween one- and non- one
language lis ene s we e diminished. On he o he hand, he p esence o
one- ele an expe ience shows ha one disc imina ion is lexible; adul
non- one language lis ene s’ one pe cep ion is imp o ed by bo h one
aining (An oniou & Wong, 2015; 2016; Chand aseka an, Sampa h, &
Wong, 2010; F ancis, Ciocca, Ma, & Fenn, 2008; Ing alson, Ba , &
Wong, 2013) and musical aining (Bu nham, B ooke , & Reid, 2015;
Coope & Wang, 2012; Wong & Pe achione, 2007). Thus, while he e is
de elopmen al di e gence in lexical one pe cep ion as a unc ion o
language en i onmen , expe ience wi h a non- one e sus a one lan-
guage does no esul in comple e one insensi i i y in in an s. In adul s,
non- one language lis ene s’ one disc imina ion is a unc ion o he
acous ic con inuum o
0
, and can be imp o ed as a unc ion o one-
ele an aining and expe ience.
Wi h espec o he one language sys em, i is unclea whe he
ce ain sys ems a e easie o pe cei e han o he s. F om a phonological
densi y pe spec i e, i is easonable o p edic ha one sys ems wi h less
densi y (i.e., ewe ones) would be easie o pe cei e. The e is an
addi ional ac o ha is ela ed o he one language sys em: he ype and
mix o one ypes in a one language. Lexical ones ha e been classed as
s a ic o dynamic (Ab amson, 1978). S a ic ones (S) a e hose wi h a
ela i ely le el con ou (e.g., Can onese 55) o a mild slope (e.g., Thai
45), di e ing om one ano he mainly in pi ch heigh ; dynamic ones
(D) in ol e dis inc pi ch con ou s (changes in pi ch heigh o e ime),
L. Liu e al.
B ain and Language 229 (2022) 105106
3
such as dynamic ising, (D , e.g., Can onese 25), dynamic alling (D , e.
g., Manda in 51), o mo e complex con ou s (e.g., ise- all, Thai 241). In
e ms o con as ype, he e is e idence om s udies examining mul iple
one con as s ha lis ene s’ one disc imina ion is ela ed o he ype o
one con as hey hea . Fo ins ance, in Huang and Johnson (2010),
disc imina ion accu acy o he Manda in ising- alling 35–51 one
con as was he bes , sugges ing ha dynamic ising e sus dynamic
alling ones may be he easies o disc imina e. Simila ly, in a com-
pa ison ac oss one con as ypes, Bu nham and colleagues ound ha
D -D one pai s a e mo e easily disc imina ed han S-S o S-D one pai s
(Bu nham, Ki kwood, Luksaneeyanawin, & Panso ee, 1992; Bu nham,
Kasisopa, Reid, Luksaneeyanawin, Lace da, A ina, Xu Ra anasone,
Schwa z, & Webs e , 2015). Mo eo e , i appea s ha he s a ic e sus
dynamic dis inc ion has psychological eali y, o i has been epo ed
ha lis ene s’ na i e s a ic one expe ience acili a es hei pe cep ion o
s a ic ones in ano he one language (Chiao, Kabak, & B aun, 2011; Qin
& Mok, 2011). Mo e comp ehensi e empi ical s udies in es iga ing
mo e con as ypes ac oss languages a e needed.
Ano he ac o ha can in luence pe cep ion is he o de in which
s imuli a e p esen ed. “Pe cep ual asymme y” occu s when he o de o
p esen a ion leads o asymme ical pe cep ual pa e ns. This phenome-
non has been epo ed o segmen al ea u es, especially owels (Polka,
Ruan, & Masapollo, 2018; Zhao, Masapollo, Polka, M´
ena d, & Kuhl,
2019). Wi h espec o lexical ones, Chen, Liu and Kage (2015) ha e
shown ha Manda in one sandhi (e.g., a icula ion o dipping-dipping
214–214 one sequences as ising-dipping, 35–214) may es upon a
pe cep ual e ec – di ec ionally-speci ic acous ic masking. In line wi h
his, i has been ound ha Manda in ising- o-dipping, 35–214 and
ising- o- alling, 35–51 one pai s elici la ge MMN esponses han he
same ones in he opposi e o de (214–35 and 51–35), among na i e bu
no non-na i e Manda in speake s, sugges ing a pe cep ual cons ain
ela ed o s imulus o de on na i e lis ene s’ phonological ep esen a-
ions (Poli ze -Ahles, Schlu e , Wu, & Almeida, 2016). Simila ly,
Wayland and Chen (2018) ound changes om Manda in la (55) o
o he ones qui e challenging, and om Manda in alling (51) o o he
ones easie han he e e se o de (o he ones o 51) o bo h one
(Manda in) and non- one (English) speake s. Howe e , a ollow-up
s udy epo ed he opposi e pa e n o Manda in la ones, and a
change om Manda in dipping o o he ones is mo e di icul han he
e e se di ec ion (Wayland, Chen, Zhou, & Hong, 2019). Mo eo e ,
p ocessing load (Liu, Ong, Tunine i, & Escude o, 2018) and acous ic
cues such as spec al dynamici y (Masapollo, Zhao, F anklin, & Mo gan,
2019) o b ea hiness (Yang & Sunda a, 2019) may all play a ole in
whe he pe cep ual asymme y occu s in one disc imina ion. Theo e -
ical models ha e been p oposed o accoun o he cons ain s imposed
on one pe cep ion by he asymme y o he o de e ec s, including he
unde -speci ica ion hypo hesis (e.g., Poli ze -Ahles e al., 2016), p o o-
ypicali y heo y (Bes , 1994), na u al e e en owel amewo k (Polka
& Bohn, 2003) and psychophysics o one salience (Bu nham & Singh,
2018). Fu he s udy is equi ed o in es iga e he ela i e me i o hese
heo ies and po en ial cons ain s h ough empi ical esea ch.
Finally, wi h espec o one p ope ies, pi ch heigh and pi ch di-
ec ion ha e consis en ly been a gued o be he wo dominan cues in
one pe cep ion (Gandou , 1983; Gandou & Ha shman, 1978; Lin,
1987; Vance, 1977) and ecogni ion (Howie, 1976; Xu, 1997). While his
is he case o bo h one and non- one language lis ene s, unc ional
magne ic esonance imaging ( MRI) (Feng e al., 2018) and mul idi-
mensional scaling (MDS) s udies (Chand aseka an e al., 2010; F ancis
e al., 2008) ha e shown ha one language speake s a end mo e o
di ec ional cues and non- one language speake s mo e o heigh cues,
indica ing a na i e onological/phonological cons ain (Chan-
d aseka an e al., 2010; F ancis e al., 2008).
He e, in o de o ex ac gene al p inciples and speci ic d i e s o
one pe cep ion and o in es iga e lexibili y and cons ain in one
p ocessing, lis ene s om ou di e en one language backg ounds
(Chinese Manda in, Singapo ean Manda in, Hong Kong Can onese, and
Bangkok Thai) and one non- one language (Aus alian English) we e
es ed on one disc imina ion in ou di e en one sys ems (Chinese
Manda in, Singapo ean Manda in, Can onese, and Thai), all o which
ha e bo h s a ic and dynamic ones bu di e in he numbe o ones. We
expec ed supe io disc imina ion o lexical ones by one- o e non- one
language speake s and, among one language speake s, o na i e o e
non-na i e ones. We also expec ed acous ic p ope ies o one would
modula e lis ene s’ pe cep ual abili y o ones ac oss language sys ems
and one ypes (Choi, Tong, & Singh, 2017). As mixed indings ha e
been epo ed ega ding pe cep ual asymme y, we keep he p edic ion
open o whe he and in wha di ec ion onal asymme ies may occu .
Finally, we ocused on he wo p ima y dimensions o one pe cep ion,
pi ch heigh and pi ch di ec ion, explo ing he ole o 23 acous ic
measu emen s/cues in hese wo dimensions o one pe cep ion (see
Appendix III) and p edic ed ha lis ene s would use he cues and/o
pi ch- ela ed knowledge ha a e ele an in hei na i e language o
non-na i e one pe cep ion. To inc ease he alidi y and gene alisabili y
o he s udy, wo disc imina ion asks, AX and AXB, we e used in wo
sepa a e g oups o pa icipan s. The esul s o he AX ask a e p esen ed
in he body o he pape , wi h he me hods o he AXB ask epo ed in
Supplemen a y ma e ial along wi h any AXB esul s ha showed
meaning ul di e gen pa e ns om he AX ask esul s.
2. Me hods
2.1. Pa icipan s
One hund ed and wen y adul s om i e language backg ounds
pa icipa ed in an AX disc imina ion ask (N =24 pe language back-
g ound): Can onese speake s (M =21.00 yea s, SD =1.64) es ed a he
Chinese Uni e si y o Hong Kong, Thai speake s (M =27.33 yea s, SD =
5.41) es ed a he Thammasa Uni e si y, Manda in Chinese speake s
(M =21.25 yea s, SD =2.44) es ed a he Beijing Language and Cul u e
Uni e si y, Singapo ean Manda in speake s (M =21.70 yea s, SD =
2.18) es ed a he Na ional Uni e si y o Singapo e, and Aus alian
English speake s (M =28.38 yea s, SD =10.43) es ed a he Wes e n
Sydney Uni e si y (see Appendix I o de ails including a sepa a e g oup
o 120 pa icipan s o AXB). Tone language speake g oups included
hose who knew o he languages including one/pi ch-accen languages,
bu all used hei na i e one language in hei daily li es. Aus alian
English speake s had no p io one language expe ience.
2.2. S imuli
All ones om he ou one sys ems, Can onese (6 ones, 15 pai -
wise con as s), Thai (5 ones, 10 con as s), Chinese Manda in (4
ones, 6 con as s) and Singapo ean Manda in (4 ones, 6 con as s),
we e used as s imuli, esul ing in 37 wi hin-language one con as s
(Appendix II). Tones we e ca ego ized in o h ee ypes: s a ic (S) ones,
dynamic ising (D ) ones, and dynamic alling (D ) ones, esul ing in
i e con as ypes: S-S, S-D , S-D , D -D , D -D . To inc ease gene al-
isabili y, each one was p esen ed on six di e en CV syllables ( wo
ini ials (/p, pʰ/) × h ee owels (/i, a, u/)). All s imuli (N =90) wo ds o
phonologically legal non-wo ds in hei espec i e languages, we e
spoken by a na i e speake o each language (all young emales:
Can onese, 24 yea s; Chinese Manda in, 30 yea s; Singapo ean Man-
da in, 28 yea s; Thai, 27 yea s). The speake p oduced six o eigh okens
o each one wo d. Two mos ep esen a i e okens (no he i s o he
las o a oid ‘lis ’ in ona ion, o p oduc ions wi hou hal s, oice
b eaking o o he impe ec ions) we e chosen o he s udy, wi h he
ide ha o e all
0
was oughly equa ed be ween languages. The chosen
s imuli we e manipula ed o equa e ampli ude (a 65 dB) wi hin and
be ween language in PRAAT (Boe sma & Weenink, 2013), bu no
du a ion o keep he s imuli sound na u al. Among he ou language
sys ems, only Thai ones sha e simila du a ions ac oss ones (due o
owel leng h being phonological); du a ions di e ac oss ones in each
L. Liu e al.
B ain and Language 229 (2022) 105106
4
o he o he h ee languages.
2.3. P ocedu e
The ask was un using he DMDX expe imen al pla o m (Fo s e &
Fo s e , 2003). AX ials consis ed o wo syllables om he same lan-
guage di e ing only in one. The A-X in e -s imulus in e al was
1000msec, and he e we e ou ypes o one pai s (Same: AA, BB;
Di e en : AB, BA). To encou age phonemic a he han acous ic p o-
cessing, in he same ials he wo ins ances o he same one we e
di e en exempla s o ha one (i.e., (A
1
A
2
, B
1
B
2
); and in he di e en
ials (AB, BA) he one exempla s in A and B posi ions di e ed be ween
any pa icula se o ials. The o al numbe o AX ials was 888, gi en
by: [37 di e en one pai ials] * [2 iden i y (same/di e en ) * 2 o de
(AA s BB o AB s BA o de s)] * [2C (/p, p
h
/) ×3 V (/a, i, u/) =6
syllable con ex s]. To a oid a igue e ec s, hal he pa icipan s in each
language g oup (12 ou o 24) we e es ed on he 444 ials in which /p/
was he ini ial consonan and he o he hal on he 444 ials on which
/p
h
/ was he ini ial consonan . Pa icipan s we e equi ed o p ess he
le / igh shi key on a compu e keyboa d i hey pe cei ed he wo
sounds (AX) o be he same/di e en as quickly and accu a ely as
possible.
2.4. Da a ea men and analyses
CV con ex s (/p, p
h
/ x /a, i, u/) we e included simply o enhance
gene alizabili y and, as hey a e o no speci ic in e es , da a we e
collapsed ac oss CV con ex s o p o ide g ea e powe . To con ol o
esponse biases, he dependen a iable was d’, de ined as: d’ =Z(hi
a e, whe e hi =‘di e en ’ esponse on an AB o BA ial) – Z ( alse ala m
a e, whe e alse ala m =‘di e en ’ esponse on an AA o BB ial).
3. Resul s
The da a we e submi ed o linea mixed-e ec s modelling (LME)
analyses using he lme4 (Ba es e al., 2015) package’s lme unc ion in R
(R Co e Team, 2018; R S udio Team, 2015). The e we e h ee key ac-
o s: lis ene s’ language backg ound (5 le els), one sys em (4 le els)
and con as ype (5 le els) and he andom ac o was pa icipan .
Resul s indica ed ha all h ee key ac o s played a ole in lis ene s’
o e all pe o mance. Pos -hoc e ec s we e ob ained h ough pai wise
compa isons using emmeans. Tone pai o de ac oss he a ious con as
ypes was cha ac e ized as ‘high- o-low’ e sus ‘low- o-high’ as ollows:
Fo S-S ials, ‘low- o-high’ was de ined as lowe pi ch heigh i s and
highe pi ch heigh second, and ice e sa o ‘high- o-low’; o S-D /S-
D ials ‘low- o-high’ was de ined as S as he i s sound in he pai and D
he second and ice e sa o ‘high- o-low’ ials; o D -D ials ‘low- o-
high’ was de ined as D (dec easing pi ch) as he i s sound and D
(inc easing pi ch) as he second and ice e sa o ‘high- o-low’ ials;
and o D -D ials ‘low- o-high’ was de ined as he ising one wi h he
lesse dynamic con ou as he i s sound and ha wi h he g ea e dy-
namic con ou as he second and ice e sa o ‘high- o-low’ ials. See
Supplemen a y Ma e ials o ables epo ing he ull model ou pu .
3.1. Disc imina ion Pe o mance: C oss-language one pe cep ion
Fig. 1 shows he d’ esul s o each o he ou one sys ems o each
o he i e pa icipan g oups. Linea mixed-e ec s modelling analyses
e ealed ha one language speake s pe o med signi ican ly be e han
he non- one language speake s (es ima e =0.518, s anda d e o (SE) =
0.0946, z =5.473, p <.0001). This e ec was e iden o con as s in
Can onese, Thai and Chinese Manda in one sys ems (es ima es >0.351,
SEs <0.142, zs >2.908, ps <0.0037), bu no o he Singapo ean
Manda in one sys em (es ima e =0.241, SE =0.141, z =1.713, p
=.0868) in which sco es ac oss pa icipan s we e high, a ‘ceiling’ e ec .
Tone language lis ene s gene ally disc imina ed na i e one con-
as s signi ican ly be e han non-na i e con as s (es ima e =0.374,
SE =0.0521, z =7.178, p <.0001) (Fig. 2). This was e iden o con-
as s in he Can onese, Thai and Chinese Manda in (es ima es >0.2991,
SEs <0.1024, zs >2.922, ps <0.0036), bu no he Singapo ean Man-
da in one sys em (es ima e =0.0258, SE =0.1255, z =0.205, p =.8374)
whe e pa icipan s ac oss one language backg ound exhibi ed a ceiling
e ec .
O e and abo e modula ions due o lis ene s’ language backg ound,
disc iminabili y o he ou one sys ems (Fig. 1A-1D) di e ed signi i-
can ly (es ima es >0.257, SEs <0.0648, zs >4.457, ps <0.0001).
Can onese ones (1A) we e he mos di icul o disc imina e (emmean =
2.61, SE =0.0477), wi h bes disc imina ion by he na i e (do ed-line
ba ) Can onese backg ound lis ene s, (ps <0.0066), and he wo s by
he non- one language Aus alian English lis ene s (le mos ba ) (ps <
0.0180). This is ollowed by he Thai ones (1B) (emmean =3.04, SE =
0.0519) wi h he Aus alian English lis ene s pe o ming signi ican ly
wo se han he na i e Thai (p =.0025) and he non-na i e Can onese (p
=.0134) lis ene s, al hough he e was no di e ence be ween Thai and
Fig. 1. Tone Language Sys em and Language Backg ound. Fo he ou one sys ems (a. Can onese, b. Thai, c. Chinese Manda in, d. Singapo ean Manda in), mean d’
sco es (y-axis) ep esen language backg ound lis ene s’ pe o mances – ac oss he x-axes, om le - o- igh : he non- one Aus alian English lis ene s, he mean
pe o mance ac oss one language speake s, hen na i e speake s o ha pa icula one sys em (ba enclosed by a discon inuous line), hen each o he one language
speake s lis ening o non-na i e ones. E o ba s = ±1 SE.
L. Liu e al.
B ain and Language 229 (2022) 105106
5
Can onese lis ene s (p =.9896). Disc imina ion o Chinese Manda in
ones (1C) was easie han ei he he Can onese (p <.0001) o he Thai
(<0.0001) ones (emmean =3.30, SE =0.0595), wi h he bes pe o -
mance by he na i e Chinese Manda in lis ene s (ps <0.0467) and he
wo s by he non- one Aus alian English lis ene s (ps <0.0211). Sin-
gapo ean Manda in ones (1D) we e he easies o disc imina e (emmean
=4.11, SE =0.0595) wi h no signi ican di e ences be ween lis ene
g oups (ps >0.1805).
Ac oss lis ene g oups, disc imina ion pe o mance di e ed be ween
one con as ypes (Fig. 2) (es ima es >0.2761, SEs <0.0836, zs >
4.739, ps <0.0001). D -D (2D) con as s we e he easies o disc imi-
na e (emmean =3.62, SE =0.0583), hen S-D (2B) (emmean =3.11, SE
=0.0561) and S-D (2C) (emmean =3.07, SE =0.0498) (which did no
di e om each o he (es ima e =0.0426, SE =0.0531, z =0.802, p
=.9301), hen S-S (2A) (emmean =2.80, SE =0.0611) wi h he D -D
(2E) con as s, being he mos di icul (emmean =2.37, SE =0.0778).
In e ms o he speci ic na i e language in luence, Can onese lis ene s
ma ginally ou pe o med Thai (p =.0685), Chinese (p =.0540), and
signi ican ly ou pe o med Singapo ean (p =.0350) and Aus alian lis-
ene s (p =.0001) in S-S disc imina ion, and Aus alian English lis ene s
unde pe o med o he lis ene s (ps <0.0227) in S-D and D -D
disc imina ion.
3.2. Disc imina ion Pe o mance: Pe cep ual asymme y among con as
ypes
Fig. 2 shows he d’ sco es o each o he ou Tone Con as Types
and he wo Tone Pai O de s o each o he i e Language Backg ound
g oups. Tes s o o ien a ional asymme ies o con as ype e ealed
ha he e we e no o e all c oss-language backg ound di e ences in he
e ec o o de o he D -D o S-S one con as ypes (es ima es <
0.0509, SEs >0.0746, zs <0.681, ps >0.4967). The e we e, howe e ,
signi ican o e all c oss-language backg ound di e ences o : (i) D -D
(mo e- o-less ising (ligh e ba s) be e han less- o-mo e ising
(da ke ba s), (ii) S-D (D → S pai s (ligh e ba s) be e han S → D
pai s (da ke ba s); and (iii) S-D (D → S pai s (ligh e ba s) be e han
S → D pai s (da ke ba s) (es ima es >0.2521, SEs <0.1142, zs >2.209,
ps <0.0272). Wi hin hese h ee con as ypes showing signi ican
di ec ional di e ences: (i) o D -D , he e we e no signi ican in-
e ac ions wi h language backg ound, i.e., he mo e- o-less ising
(ligh e ba s) >less- o-mo e ising (da ke ba s) e ec was consis en
ac oss language backg ound g oups, (ii) o S-D , he D → S pai s
(ligh e ba s) >S → D pai s (da ke ba s) e ec was g ea e o
Can onese and Thai han he o o he pa icipan s (es ima es >0.4382,
SEs <0.157, zs >2.804, ps <0.0051), and (iii) o S-D , he D → S pai s
(ligh e ba s) >S → D pai s (da ke ba s) e ec was g ea e o
Aus alian English and Thai han o he pa icipan s (es ima es >0.5520,
SEs <0.122, zs >4.503, ps <0.0001).
Taken oge he , in con as ypes whe e di e ences occu , he e is a
gene al end ha disc imina ion o one pai s is be e when he i s
one is mo e dynamic (highe pi ch, g ea e pi ch con ou , o g ea e
inc ease in pi ch con ou ) han he second one compa ed o when he
i s one is less dynamic han he second one.
3.3. Modelling
Fo MDS analyses, he d’ sco es we e u he i ed by he Indi idual
Di e ences Scaling INDSCAL model (Ca oll & Chang, 1970), a 3-way
ex ension o he classical MDS using he smaco IndDi unc ion in he
smaco R package (Bo g, G oenen, & Mai , 2013; De Leeuw & Mai ,
2009). A wo-dimensional solu ion (pi ch heigh , pi ch di ec ion) was
op ed o based on p e ious one pe cep ion li e a u e, isual inspec ion
o he sc een plo s, and he in e p e abili y o he dimensions. Bo h
o e all and language-speci ic cue weigh ing by lis ene s is epo ed
using K uskal’s s ess (S ess-1) o e alua e he model i (K uskal &
Wish, 1978). Resul s wi h andoms ess() baseline compa isons indica ed
Fig. 2. Tone Con as Types and Tone Con as O ien a ion. Fo he i e con as ypes (a. S a ic-S a ic, b. S a ic-Falling, c. S a ic-Rising, d. Falling Rising, e. Rising-
Rising), mean d’ sco es (y-axis) ep esen lis ene s’ pe o mances – ac oss he x-axes, om le - o- igh : he non- one Aus alian English lis ene s, hen mean pe -
o mance ac oss one language speake s, hen one language speake s. Ba colou s ep esen whe he one pai s we e p esen ed in he o de om low- o-high (da ke )
o high- o-low (ligh e ) dynamic o ien a ion. E o ba = ±1 SE.
L. Liu e al.
B ain and Language 229 (2022) 105106
6
a sa is ac o y model i o ou models co esponding o each language
sys em (Mai , de Leeuw, & G oenen, 2015). The dimensions o each one
language we e in e p e ed by ma ching each one’s
0
and i s de i a i es
(slope, accele a ion e c., see Appendix III o de ailed acous ic mea-
su emen s) o he wo dimensions o each one sys em as closely as
possible as ollows: The acous ic measu emen s ha we e he mos
simila o he dimension coo dina es we e i s selec ed (simila i y
measu ed by he Pea son’s co ela ion coe icien ), and hen con i med
by a se o eg ession analyses o ensu e hey we e signi ican ac o s in
p edic ing each dimension (heigh , di ec ion). These heigh and di ec-
ion weigh s we e in e p e ed o index he ela i e impo ance a
pa icipan placed on he wo cues in one pe cep ion in a one sys em
and we e compa ed wi hin each INDSCAL model/ one sys em.
Depending on he no ming o he common space (o a one sys em), hey
we e conside ed as anked da a (Bo g e al., 2013) and analysed by an
o dinal eg ession model: Cumula i e Link Mixed Model in R. Model
assump ions we e checked wi h he o dinal packages nominal_ es () and
scale_ es () unc ions (Ch is ensen, 2015). I any independen a iable
ailed hese es s (i.e., a signi ican p- alue was e u ned), ha a iable
was handled di e en ly in he model using he nominal and scale op-
ions in he clm() unc ion.
3.3.1. Modelling: Dimensionali y
The INDSCAL model (Ca oll & Chang, 1970) e ealed pi ch heigh
and di ec ion cues o he ou one sys ems (Fig. 3). The mos salien
heigh and di ec ion cues o lis ene s we e (i) o Thai ones (s ess-1 =
0.134): a e age pi ch ( =0.99, p =.001) and mean slope a e he i s
in lec ion ( =0.90, p =.038); (ii) o Can onese ones (s ess-1 =0.168):
s a pi ch ( =-0.97, p =.001) and end slope ( =0.96, p =.002); (iii) o
Chinese Manda in ones (s ess-1 =0.090): end pi ch ( =0.97, p =.030)
and mean magni ude o slope change a e o e he las wo ime poin s
( =-0.99, p =.010); and (i ) o Singapo ean Manda in ones (s ess-1 =
0.077): min pi ch ( =0.99, p =.010) and end slope ( =-0.94, p =.060).
Fig. 3. Dimensionali y modelling (INDSCAL schema ics) o he ou one sys ems: (a. Can onese, b. Thai, c. Chinese Manda in, d. Singapo ean Manda in). The
connec ed poin s ep esen acous ic space ac oss lis ene s (black line) and in e p e ed dimensional solu ions ( ed line) along he bes i heigh (x-axis) and di ec ion
(y-axis) cues mos co ela ed o he wo pe cep ual dimensions in each o he languages. The g ea e he o e lap o he black and ed spaces, he highe he likelihood
ha he in e p e ed pe cep ual dimensions a e co ec . (Fo in e p e a ion o he e e ences o colou in his igu e legend, he eade is e e ed o he web e sion o
his a icle.)
L. Liu e al.
B ain and Language 229 (2022) 105106
7
3.3.2. Modelling: Cue weigh ing
The main e ec s o language backg ound and cue dimensions in each
o he ou one sys ems we e examined (Fig. 4 and Table 1, see Sup-
plemen a y ma e ial o de ailed weigh ing magni udes ac oss one
sys ems and language backg ounds). O e all, when one backg ound
pa icipan s lis en o ones in hei na i e languages (do ed-line ba )
Can onese lis ene s a ou ed di ec ion (End Slope) o e heigh (S a
Pi ch) (D >H); Chinese Manda in lis ene s a ou ed heigh (End Pi ch)
o e di ec ion (End Acc (abs) (H >D); Singapo ean Manda in lis ene s
a ou ed di ec ion (End Slope) o e heigh (Min Pi ch), and o Thai
lis ene s, A ePi ch and SlopeA e In we e he mos salien , bu he e
was no bias o one o he o he (D =H). These na i e language biases
did no ans e di ec ly o non-na i e one languages, indeed he e we e
many ins ances in which lis ene s adap ed o he a ge language, e.g.
when Can onese speake s wi h hei na i e bias o D >H lis ened o
Chinese Manda in, hei bias changed o ha o he na i e Chinese
Manda in speake s, H >D. The e is, in ac , mo e adap a ion o he L2
han he e is L1 → L2 ca yo e bias. Mos no able is ha he non- one
language speake s did no adap a all, hey main ained an H >D bias,
sugges ing ha expe ience wi h a pa icula one language acili a es
u he acili y wi h o he one languages.
4. Discussion
This s udy o he pe cep ion o ou one sys ems (Can onese, Thai,
Chinese Manda in, Singapo ean Manda in) by lis ene s om i e lan-
guage backg ounds (Aus alian English, Can onese, Thai, Chinese, Sin-
gapo ean) showed ha one pe cep ion is cons ained by (i) language
backg ound, (ii) one sys em, (iii) one con as ype, (i ) o de o p e-
sen a ion, and ( ) salience weigh ing o heigh and di ec ion cues. De-
ails o hese indings a e se ou below.
4.1. Language backg ound
Language backg ound is a d i ing ac o in one pe cep ion.
Consis en wi h p e ious li e a u e (Bu nham & F ancis, 1997; Bu n-
ham, e al., 2015; Kaan, Wayland, Bao, & Ba kley, 2007; Maggu, Zong,
Law, & Wong, 2018), one language adul s show enhanced disc imina-
ion o one compa ed wi h non- one language adul s. This acili a ion,
also e iden in in ancy esea ch (Ma ock & Bu nham, 2006), was
consis en ac oss ou o he i e one sys ems bu was diminished in
Singapo ean Manda in, which was ound o be highly disc iminable (see
one sys ems below).
While p e ious s udies ha e indi ec ly shown suppo o he hy-
po hesis ha one language lis ene s disc imina e hei na i e ones
be e han he non-na i e ones by compa ing he pe cep ual ou comes
o he same ones/ onal con as s om di e en one language speake s
(e.g., Reid e al., 2015), ou esul s show di ec ly ha one language
speake s’ pe cep ion o na i e ones is be e han hei pe cep ion o
non-na i e ones and do so qui e con incingly – o h ee o he ou one
language speake g oups he e – Can onese, Chinese Manda in, and Thai
(wi h he lack o e ec o Singapo ean Manda in speake s being
possibly due o all one language speake s doing well on Singapo ean
ones). These obse a ions likely poin o he e ec o speci ic linguis ic
expe iences, such as pe cep ual assimila ion be i on a phone ic (So &
Bes , 2014) o phonological (Bohn, A esani, Bes , & Vay a, 2011) le el.
These indings also do e ail wi h exis ing neu ophysiological e idence
o one pe cep ion. In a posi on emission omog aphy s udy compa ing
Thai and non-speech ones, Thai lis ene s showed signi ican ac i a ion
in he le on al ope culum, sugges ing phonological p ocessing,
whe eas Chinese and English lis ene s exhibi ed sensi i i y in he an e-
io insula egion, indica ing phone ic o non-linguis ic p ocessing.
Hence, one p ocessing is cons ained by language ( one e sus non-
one; na i e e sus non-na i e) expe iences (Gandou e al., 2000).
4.2. Tone sys em
Disc iminabili y o he ou one sys ems di e ed, wi h an o dinal
anking in disc imina ion ease, o Singapo ean Manda in (4 ones) >
Chinese Manda in (4 ones) >Thai (5 ones) >Can onese (6 ones). A
i s glance, such anking may be ela ed o he complexi y/densi y o
one space. Howe e , ha explana ion canno accoun o he di e ence
be ween he wo Manda in one sys ems which should sha e simila
complexi y. A ecen s udy has p oposed ha Singapo ean Manda in
ones may ha e a unique a icula o y o igin. Using ul asound o ack
la yngeal mo emen , Yun and Moisik (2019) ound ha compa ed o
Chinese Manda in speake s, Singapo ean Manda in speake s’ a icula-
ion o ones in ol es he addi ional use o la yngeal medializa ion and
app oxima ed hy oid lamina. I needs o be in es iga ed whe he hese
mechanisms also acili a e one pe cep ion. The ad an age o Singa-
po ean Manda in o e Chinese Manda in ones was e iden in Aus a-
lian English, Can onese and Thai speake s, bu he ela i e supe io i y
was no so g ea o he Can onese and Thai speake s. This may e lec
simila s a ic and dynamic onal dis ibu ions/spaces in Chinese
Fig. 4. Cue weigh ing sco es o he ou one sys ems (a. Can onese, b. Thai, c. Chinese Manda in, d. Singapo ean Manda in). Mean weigh sco es (y-axis) ep esen
lis ene s’ pe o mances (x-axis, le - o- igh : Aus alian English lis ene s, hen mean pe o mance ac oss one language speake s, hen na i e speake s o ha
pa icula one sys em, hen one language speake s lis ening o non-na i e ones. Ba colou s ep esen di ec ion (da ke ) e sus heigh (ligh e ) cues. E o ba =
±1 SE.
L. Liu e al.
B ain and Language 229 (2022) 105106
8
Manda in and Singapo ean Manda in one sys ems, and is also in line
wi h indings showing he numbe o s a ic ones in lis ene s’ L1 can
p edic hei disc imina ion abili y a leas o s a ic ones (Chiao e al.,
2011; Qin & Mok, 2011). Thus, he e is e idence o a gene al cons ain
ha he densi y/di e si y o one space in he na i e language a ec s
one disc iminabili y i espec i e o he lis ene s’ language backg ound.
While his migh be a key cons ain a ec ing one disc iminabili y, i is
no he only ac o .
4.3. Tone con as ype
In o de om easie o mo e di icul , he disc imina ion ease o he
i e con as ypes was: Dynamic-Dynamic (D -D ) >S a ic-S a ic (S-S)
>S a ic-Dynamic (S-D =S-D ) >Dynamic-Dynamic (D -D ), which is
in line wi h p e ious s udies (Bu nham e al., 1992; Bu nham & F ancis,
1997; Huang & Johnson, 2010; Liu & Kage , 2018). Tu ning o he e ec
o language backg ound on con as ype, o D -D , Aus alian English
lis ene s pe o med mo e poo ly han all one language lis ene s, whose
pe o mance did no di e . Fo he mos di icul D -D ( ising e sus.
ising) con as ype, he e was no in luence o language backg ound on
pe o mance, which could be in e p e ed as being due o he acous ic/
phone ic simila i y o he wo ising ones. Howe e , i is equally
plausible ha o non- one language speake s he di icul y is a he
acous ic/phone ic le el (So & Bes , 2014), bu o he one language
speake s, he di icul y may be mo e phonological (Reid e al., 2015). In
suppo o such a conclusion, (i) he non- one (Aus alian English) lan-
guage lis ene s pe o m much be e on he high- o-low o de o p e-
sen a ion o D -D con as s, whe eas o he one language backg ound
speake s his is no so appa en (g ea e acous ic e ec o he non- one
language speake s), and (ii) Thai is he only one language sys em which
does no ha e a D -D con as (see Supplemen a y ma e ial I) and Thai
language backg ound speake s pe o m mo e poo ly on D -D con as s
han do he o he h ee one language backg ound g oups.
4.4. Tone o de
The o de in which ones a e p esen ed a ec ed disc iminabili y.
Be e disc imina ion o one pai s was obse ed when he i s one is
mo e dynamic (highe pi ch, g ea e pi ch con ou , o g ea e inc ease in
pi ch con ou ) han he second one compa ed o he o he way a ound.
Fi e explana ions ha ha e been p oposed o accoun o such pe cep-
ual asymme ies can be conside ed. Fi s , he unde -speci ica ion hy-
po hesis (Co nell, Lahi i, & Euli z, 2013; Lahi i & Ree z, 2010) has been
used o explain asymme ies in consonan (Gaskell, 2003; Hes ik &
Du asula, 2016), owel (De Jonge & Boe sma, 2015; Scha inge e al.,
2012), and one pe cep ion (Poli ze -Ahles e al., 2016). This hypo hesis
a gues ha unde speci ied (and he e o e mo e inclusi e, lexible) ea-
u es a e be e disc imina ed han speci ied ea u es (Sha e , Schwa z,
& Ku zbe g, 2004; Schlu e e al., 2016; Schlu e , Poli ze -Ahles, Al
Kaabi, & Almeida, 2017), esul ing in he p edic ion ha a p esen a ion
o de wi h high- o-low dynamici y would be mo e unde speci ied.
Second, in p o o ypicali y heo ies, i is conside ed ha non-na i e
phonemes ha a e mo e de ian om p o o ypical membe s o a
phonological ca ego y should be mo e salien and mo e easily disc im-
ina ed om he p o o ype, han hose close o he p o o ypical membe s
(Bes , 1994; Bes & Tyle , 2007; Ikeda e al., 2002; K iengwa ana &
Escude o, 2017). This would imply ha he mo e dynamic a one he
mo e i is p o o ypical. Thi d, in he Na u al Re e en Vowel amewo k
Table 1
Summa y o he weigh ing o ela i e heigh (H) and di ec ion (D) cues o he 4 one sys ems and he 5 language expe ience g oups plus means o one and non-
na i e one g oups, along wi h he speci ic signi ican heigh and di ec ion cues wi h es ima es (es ), s anda d e o s (SE) and z sco es (* p <.015; ** p <.001; *** p
<.0005) and gloss, i.e., whe he D >H, H >D o D =H.
Tone
Sys em
Language Backg ound
Can onese
Chinese Manda in
Singapo ean
Manda in
Thai
Aus alian
(Non-Tone)
All Tone
Languages
All Non-Na
Tone Langs
Can onese
D>H
End Slope > S a
Pi ch
es = -2.59
SE = 0.300,
z = -8.652**
D=H
NSDs
D>H
End Slope >
S a Pi ch
es = -2.59
SE = 0.300,
z = -8.652**
D>H
End Slope >
S a Pi ch
es = -2.59
SE = 0.300
z = -8.652**
H>D
S a Pi ch > End
Slope
es = 1.7810 SE =
0.509
z = 3.497***
D>HD>H
Ch-Mand
H>D
H>D
H>D
H>D
H>D
--- H>D
Ch-Mand s es
es = -1.787
SE = 0.493
z = -3.624***
All 5: End Pi ch > End Acc (abs), es > -1.786, SEs<0.579 zs>-3.624**
L. Liu e al.
B ain and Language 229 (2022) 105106
9
(Polka & Bohn, 2003, 2011), owels ha a e mo e pe iphe al in F1-F2
space a e disc imina ed be e om mo e cen al owels han ice
e sa, leading o he p edic ion ha ones wi h highe dynamici y a e
conside ed mo e cen al in his model. The ou h explana ion a ge s
he psychophysics o one salience (Bu nham & Singh, 2018; Chan-
d aseka an, K ishnan, & Gandou , 2007, 2009; K ishnan, Gandou , &
Bidelman, 2010) and is g ounded in empi ical indings demons a ing
ela ionships be ween acous ic–phone ic p ope ies and phonological
de elopmen (Polka & Bohn, 2011; Tsuji, Mazuka, C is ia, & Fikke ,
2015). This would indica e an ini ial highe salience would be mo e
dis inguishable han i he ini ial one we e less salien . Finally, as pi ch
is a salien linguis ic ea u e humans a e exposed o since and be o e
bi h, we hypo hesize ha a en ional ac o s may also play a ole, such
ha o de s leading o highe a en ion and a ousal will be be e
disc imina ed, in ou case he o de o high- o-low dynamici y. Table 2
summa izes hese explana ions in he cu en con ex . These da a ha
we epo a e he i s sys ema ic demons a ion o one asymme y;
con as s p esen ed om a high- o-low dynamic di ec ion a e be e
disc imina ed han hose p esen ed om a low- o-high dynamic di ec-
ion. We u he hypo hesize his as a po en ial gene al cons ain in one
disc imina ion and lea ning.
4.5. Cue dimensionali y and weigh ing
The esul s ega ding pi ch heigh and di ec ion cues e lec bo h
lexibili y and cons ain , illus a ing he igh bond be ween lis ene s’
one pe cep ion and he ela i e weigh o heigh and di ec ion cues in
hei na i e one language (F ancis e al., 2008). O g ea impo ance a e
(i) di e ences in cue weigh ing be ween languages and (ii) how hose
di e ences may in luence pe cep ual cue weigh ing in L2 one
pe cep ion.
Fi s , he e we e sys ema ic di e ences in how lis ene s om
di e en language backg ounds weigh pe cep ual cues (Table 1). In he
pe cep ion o hei na i e one sys ems, Chinese Manda in lis ene s
weigh ed heigh o e di ec ion cues whe eas Singapo ean Manda in
lis ene s weigh ed di ec ion o e heigh . Nei he o hese Manda in
a ie ies has wo ones ha di e only in
0
heigh , lea ing he unex-
plained inding o Chinese Manda in lis ene s’ g ea e eliance on heigh
cues a subjec o u he explo a ion. Unlike Manda in, in addi ion o
con ou ones, Can onese and Thai each has h ee le el ones con as ing
in
0
heigh wi h li le
0
modula ion ac oss ime (Ab amson, 1962;
Mo ´
en & Zsiga, 2006), so he equal weigh ing o heigh and di ec ion by
na i e Thai speake s o Thai, and o di ec ion o e heigh by na i e
Can onese speake s o Can onese a e ela i ely explicable (al hough
gi en he numbe o s a ic: dynamic ones in each, 3:3 o Can onese and
3:2 o Thai, he opposi e could ha e been expec ed).
Second, ega ding cues o L2 one pe cep ion, p e ious s udies
indica e ha Can onese lis ene s show mo e eliance on di ec ion han
heigh cues (Gandou 1983; Li & Shuai, 2011), and ha one and non-
one language speake s a end o bo h heigh and di ec ion cues when
lis ening o Thai (Chand aseka an e al., 2010). Ou esul s depic a
somewha di e en pic u e (Table 1). Chinese Manda in lis ene s, when
lis ening o Singapo ean Manda in, shi ed hei na i e one H >D
eliance o align wi h ha a ou ed by na i e Singapo ean lis ene s, D >
H; bu while hei eliance also changed o Thai ( o D >H) and
Can onese (D =H), nei he aligned wi h he na i e speake s’ eliance.
Singapo ean Manda in lis ene s exhibi ed pe ec adap abili y; gi en hei
D >H o na i e ones, hey emained wi h D >H o Can onese, shi ed
o H >D o Chinese Manda in, and o H =D o Thai, weigh ings ha
migh be conside ed op imal as hey a e he weigh ings ha na i e lis-
ene s use o hose languages (shaded cells in Table 1). Can onese lis-
ene s we e mo e simila o Chinese Manda in lis ene s; hey changed
hei na i e D >H weigh ings o all h ee o he o he languages, bu
his only acco ded wi h he na i e weighing in he case o Chinese
Manda in (H >D). Finally, Thai lis ene s showed some deg ee o
adap a ion: om hei na i e D =H, hey shi ed o na i e language-
app op ia e weigh ings o Can onese (D >H) and Chinese Manda in
(H >D), bu no o Singapo ean Manda in (D >H a he han D =H).
The non- one Aus alian English lis ene s showed s ikingly di e en
esul s. They we e insensi i e o he op imal cue weigh ing in di e en
one languages and consis en ly elied on H >D ac oss he one sys ems
(excep o Singapo ean Manda in – H =D). This is in line wi h p e ious
s udies on English lis ene s’ eliance on heigh cues in hei pe cep ion
o Can onese (F ancis e al., 2008; Gandou , 1983; Gandou & Ha sh-
men, 1978) and Thai (Bu nham & F ancis, 1997).
O e all, hese esul s depic a sligh ly di e en pic u e om p e ious
cue weigh ing s udies which sugges ha lis ene s’ c oss-language one
pe cep ion is de e mined by he cue weigh ing in hei na i e language
(e.g., F ancis e al., 2008). Ra he , i appea s ha one language speake s
a e ela i ely lexible, shi ing cues o a s a egy ha esembles he cue
weigh ing o na i e speake s. Singapo ean Manda in and Thai lis ene s
show he g ea es adap abili y, and hei Can onese and Chinese Man-
da in coun e pa s do so wi h a ying success. In con as , he non- one
language lis ene s appea ed insensi i e o he di e ences be ween one
languages; hey main ained an H >D s a egy (H =D o Singapo ean
Manda in) ac oss one languages.
The e o e, a he han a ans e o cue weigh ing p io i y om one
language speake s’ L1 o L2, ou esul s sugges ha one language
speake s a e sensi i e o di e ences in one sys ems, and ha i is he
na u e o he one sys em a he han lis ene s’ na i e one language back-
g ound ha go e ns cue weigh ing in one pe cep ion. Tha said, ha ing a
one language backg ound is c ucial o such adap abili y. Aus alian
English (non- one) lis ene s we e esis an and nonadap i e; hey
consis en ly weigh ed heigh o e di ec ion cues. This can be iewed in
wo ways, as (i) insensi i i y o he nuances o one languages esul ing
in non-op imal pe o mance, o (ii) main aining a p e e ed app oach
ha , as a gene al s a egy, wo ks easonably well ac oss all one lan-
guages, and speci ically, selec i e a en ion o he gene al heigh cues
ha a e p esen in mos one con as s and selec i ely es ic ed a en-
ion o cues ha a e no uni e sally use ul ac oss one con as s.
In sum, one language lis ene s do no necessa ily use he same cues
in na i e and non-na i e one pe cep ion. In ac , hey end o adap
lexibly o he acous ics o he a ge language a he han ixedly
ans e ing hei L1 p io i ies. Non- one language lis ene s, on he o he
hand, do no show such adap a ion, o hem i appea s ha all one
languages a e equal, as in equally o eign.
5. Conclusion
Some p e ious esea ch suppo s he no ion ha one pe cep ion is
lexible, independen o na i e phonology, whe eas o he esea ch
sugges s i is cons ained by na i e phonology o pe cep ual asymme-
ies. Gi en ha such mixed esul s could be due o s imulus and/o ask
Table 2
A summa y o pe cep ual asymme y explana ions agains he cu en inding.
Unde hese explana ions, he p esen a ion o de s om a high o low dynamici y
would be conside ed as being mo e speci ied, mo e ypical and cen al ep e-
sen a ion, wi h highe salience and a ousal han he o he way a ound.
Resul s/
Hypo heses
Tone ype A Tone ype B Disc imina ion
pe o mance
Empi ical
Resul s (he e)
High
dynamici y
Low
dynamici y
Hi → Lo >Lo → Hi
(dynamici y)
Unde -
speci ica ion
Speci ied Unde speci ied Spec → Unde >Unde
→ Spec
P o o ypes Typical A ypical Typical → A ypical >
A ypical → Typical
NRV Cen al Pe iphe al Cen al → Pe iph >
Pe iph → Cen al
Salience High
Salience
Low Salience Hi → Lo >Lo → Hi
(salience)
A en ion High
A en ion
Low A en ion Hi → Lo >Lo → Hi
(a en ion)
L. Liu e al.
