This is a sel -a chi ed e sion o an o iginal a icle. This e sion
may di e om he o iginal in pagina ion and ypog aphic de ails.
Au ho (s):
Ti le:
Yea :
Ve sion:
Copy igh :
Righ s:
Righ s u l:
Please ci e he o iginal e sion:
CC BY-NC 4.0
h ps://c ea i ecommons.o g/licenses/by-nc/4.0/
Human B ain Ages Wi h Hie a chy-Selec i e A enua ion o P edic ion E o s
© The Au ho (s) 2020. Published by Ox o d Uni e si y P ess. All igh s ese ed.
Published e sion
Hsu, Yi-Fang; Waszak, Flo ian; S ömme , Juho; Hämäläinen, Ja mo A.
Hsu, Y.-F., Waszak, F., S ömme , J., & Hämäläinen, J. A. (2021). Human B ain Ages Wi h
Hie a chy-Selec i e A enua ion o P edic ion E o s. Ce eb al Co ex, 31(4), 2156-2168.
h ps://doi.o g/10.1093/ce co /bhaa352
2021
© The Au ho (s) 2020. Published by Ox o d Uni e si y P ess. All igh s ese ed. Fo pe missions, please e-mail: [email p o ec ed]
This is an Open Access a icle dis ibu ed unde he e ms o he C ea i e Commons A ibu ion Non-Comme cial License (h p://c ea i ecommons.o g/
licenses/by-nc/4.0/), which pe mi s non-comme cial e-use, dis ibu ion, and ep oduc ion in any medium, p o ided he o iginal wo k is p ope ly ci ed.
Fo comme cial e-use, please con ac [email p o ec ed]
Ce eb al Co ex, 2020;00: 1–13
doi: 10.1093/ce co /bhaa352
O iginal A icle
ORIGINAL ARTICLE
Human B ain Ages Wi h Hie a chy-Selec i e
A enua ion o P edic ion E o s
Yi-Fang Hsu 1,2, Flo ian Waszak3,4, Juho S ömme 5and
Ja mo A. Hämäläinen5
1Depa men o Educa ional Psychology and Counselling, Na ional Taiwan No mal Uni e si y, 106308 Taipei,
Taiwan, 2Ins i u e o Resea ch Excellence in Lea ning Sciences, Na ional Taiwan No mal Uni e si y, 106308
Taipei, Taiwan, 3Cen e Na ional de la Reche che Scien i ique (CNRS), In eg a i e Neu oscience and Cogni ion
Cen e (INCC), Uni é Mix e de Reche che 8002, 75006 Pa is, F ance, 4Uni e si é de Pa is, 75006 Pa is, F ance and
5Jy äskylä Cen e o In e disciplina y B ain Resea ch, Depa men o Psychology, Uni e si y o Jy äskylä,
40014 Jy äskylä, Finland
Add ess co espondence o Yi-Fang Hsu, No.162, Sec. 1, Heping E. Rd, Da-an Dis ic , 106308 Taipei, Taiwan. Email: yi- ang.hsu@can ab.ne
Abs ac
F om he pe spec i e o p edic i e coding, ou b ain embodies a hie a chical gene a i e model o ealize pe cep ion, which
p oac i ely p edic s he s a is ical s uc u e o senso y inpu s. How a e hese p edic i e p ocesses modi ied as we age?
Recen esea ch sugges ed ha aging leads o dec eased weigh ing o senso y inpu s and inc eased eliance on p edic ions.
He e we in es iga ed whe he his age- ela ed shi om senso ium o p edic ions occu s a all le els o hie a chical
message passing. We eco ded he elec oencephalog aphy esponses wi h an audi o y local–global pa adigm in a coho o
108 heal hy pa icipan s om 3 g oups: senio s, adul s, and adolescen s. The de ec ion o local de iancy seems la gely
p ese ed in olde indi iduals a ea lie la ency (including he misma ch nega i i y ollowed by he P3a bu no he
eo ien ing nega i i y). In con as , he de ec ion o global de iancy is clea ly comp omised in olde indi iduals, as hey
showed wo se ask pe o mance and a enua ed P3b. Ou indings demons a e ha olde b ains show li le decline in
senso y (i.e., i s -o de ) p edic ion e o s bu signi ican diminu ion in con ex ual (i.e., second-o de ) p edic ion e o s.
Age- ela ed de icien main enance o audi o y in o ma ion in wo king memo y migh a ec whe he and how lowe -le el
p edic ion e o s p opaga e o he highe le el.
Key wo ds: aging, audi o y pe cep ion, elec oencephalog aphy (EEG), p edic ion e o s, p edic i e coding
In oduc ion
P edic i e p ocessing is undamen al o audi ion.This is because
he audi o y sys em essen ially deals wi h sequen ial inpu s,
which con ains ele an in o ma ion in he imminen u u e.
The p edic i e coding model o pe cep ion u he delinea es
ha he pe cep ual sys em uses a hie a chical gene a i e model
o p oac i ely ep esen he s a is ical s uc u e o senso y
inpu s (F is on 2005,2009;Egne e al. 2010;Feldman and F is on
2010; o e iews see Cla k 2013;Heilb on and Chai 2018).
Speci ically, pe cep ion in ol es he ma ching o a cascade
o op-down p edic ions wi h senso y inpu s, which aims
o minimize p edic ion e o s (o he inconsis ency be ween
an icipa ed and incoming signals) in he sys em.
How a e hese p edic i e p ocesses modi ied as we age?
No mal aging ypically deg ades p ecision o pe iphe al and
cen al p ocessing, which leads o dec eased weigh ing o sen-
so y inpu s and inc eased eliance on p edic ions (Wolpe e al.
2016;Chan e al. 2017). Recen neu ocompu a ional esea ch
Downloaded om h ps://academic.oup.com/ce co /ad ance-a icle/doi/10.1093/ce co /bhaa352/6014977 by Jy askyla Uni e si y use on 08 Decembe 2020
2Ce eb al Co ex, 2020, Vol. 00, No. 00
quan i ying he synap ic coupling unde lying he misma ch
nega i i y (MMN) also ound an age- ela ed a enua ion o
lea ning-dependen inc ease in o wa d connec i i y om
p ima y audi o y co ex speaking o a educed sensi i i y o
ascending a e en s o p edic ion e o s (Mo an e al. 2014)
as well as an age- ela ed inc ease o inhibi o y e ec a
in e io on al gy us indica ing inc eased i ing a e o he
inhibi o y neu ons (Coo ay e al. 2014) o e li espan. I seems
ha olde b ains a e less p edisposed o upda ing he p io
p obabili y es ima e, leading o a pe cep ion o he en i onmen
inc easingly domina ed by op-down in o ma ion. In o he
wo ds, age u ns ou b ain in o a s ubbo n p edic ion machine
whe e he senso ium is weigh ed less o op imize p edic ions.
This is consis en wi h epo s on age- ela ed shi s in neu onal
ec ui men om senso y o on al egions (Da is e al. 2008).
Howe e ,i is cu en ly unclea whe he his age- ela ed shi
om senso ium o p edic ions occu s a all le els o hie a chical
message passing, which includes senso y (i.e., i s -o de )
p edic ions and con ex ual (i.e., second-o de ) p edic ions. A
well-es ablished audi o y local–global pa adigm (Bekinsch ein
e al. 2009) p obing audi o y p edic ion a 2 hie a chical le els
o de iancy ound ha he de ec ion o local/global de iancy
in ol es dis inc neu onal sys ems (Wacongne e al. 2011;
Chennu e al. 2013;El Ka oui e al. 2015). Local de iancy elici ed
he MMN ( ega dless o pa icipan s’ ask), which e lec s a
p ea en i e, au oma ic, and nonconscious esponse o a i s -
o de iola ion o pa icipan s’ expec a ion cons uc ed on he
basis o ecen senso y his o y. I o igina ed om a es ic ed
ne wo k in ol ing bila e al Heschl’s gy us, supe io empo al
gy us, and on al egion a a smalle ex en (Chennu e al.
2016; o e iews see Alho 1995) wi h a sligh delay (Rinne
e al. 2000). Global de iancy, on he o he hand, elici ed he
P3b (only when pa icipan s we e awa e o he iola ions o
global egula i ies), which e lec s a second-o de iola ion
o pa icipan s’ expec a ion o a gi en con ex ual ule. I is
hough o index conscious access and upda ing in wo king
memo y (Donchin and Coles 1988;Polich 2004,2007). I s pu a i e
gene a o s in ol e a b ain-scale co ical ne wo k including
p e on al, pa ie al, empo al, and cingula e egions (Ta kka
e al. 1995;Polich 2004,2007;Pa el and Azzam 2005). Such
dissocia ion o he consciousness-independen /-dependen
capaci y o hie a chically p edic he local/global s uc u e o
he audi o y en i onmen is sha ed by human (Fauge as e al.
2011,2012;King e al. 2013;S auss e al. 2015) and nonhuman
p ima es (Uh ig e al. 2014,2016).
An indisc imina e age- ela ed shi om senso ium o p e-
dic ions should be associa ed wi h educed MMN and P3b, sig-
naling educed p edic ion e o s due o dec eased weigh ing o
senso y inpu s a bo h senso y and con ex ual le els. Ne e he-
less, conclusions a e di icul o d aw om he li e a u e. P e-
ious esea ch examining he age- ela ed changes in he MMN
yielded equi ocal esul s. While some s udies ound no e idence
o age-dependen a enua ion o he MMN (Pekkonen e al. 1993,
1996;Sch oede e al. 1995;Gun e e al. 1996;Kazme ski e al.
1997;Amenedo and Díaz 1998), o he s udies ound senio s o
exhibi a smalle MMN (Czigle e al. 1992;Gae a e al. 1998,2001;
Alain and Woods 1999;Coope e al. 2006;Schi e al. 2008). The
lack o consensus was likely due o a ia ions in how pa ici-
pan s’ a en ion was moni o ed, he leng h o he s imulus onse
asynch ony (SOA), e c., making i di icul o judge whe he
age a enua es he i s -o de p edic ion e o s. On he o he
hand, while P3b was commonly epo ed o show age- ela ed
changes in i s ampli ude, la ency, and opog aphy (Schi e al.
2008; an Din e en e al. 2018; o e iewsseeF iedman e al.
1997;F iedman 2003), i s measu emen was usually based on
he classi ica ion o he audi o y de ian /s anda d s imuli ha
i s elici ed he MMN. The e o e, i emains unde e mined o
wha ex en he P3b uncon oundedly e lec s he second-o de
p edic ion e o s.
To es whe he he age- ela ed shi om senso ium
o p edic ions is hie a chy selec i e, he e we eco ded he
elec oencephalog aphy (EEG) esponses wi h an audi o y local–
global pa adigm, which o hogonally manipula ed local (i.e.,
i s -o de ) and global (i.e., second-o de ) egula i ies (Fig. 1).
In a coho o 108 heal hy pa icipan s om 3 g oups: senio s
(≥55 yea s old), adul s (19–54 yea s old), and adolescen s
(≤18 yea s old), we ound ha olde b ains show li le decline
in he MMN signaling senso y (i.e., i s -o de ) p edic ion e o s
bu signi ican diminu ion in he P3b signaling con ex ual (i.e.,
second-o de ) p edic ion e o s, which is ela ed o de icien
main enance o audi o y in o ma ion in wo king memo y.
Ma e ials and Me hods
Pa icipan s
A o al o 108 heal hy olun ee s including 36 senio s (age
be ween 55 and 82 yea s, mean = 65.31, s anda d de ia ion
[SD] = 7.03), 36 adul s (age be ween 19 and 27 yea s, mean= 21.17,
SD = 2.01), and 36 adolescen s (age be ween 15 and 18 yea s,
mean = 16.97, SD = 0.81) pa icipa ed in he expe imen wi h
no his o y o neu ological, neu opsychia ic, o isual/hea ing
impai men s as indica ed by sel - epo . Pa icipan s p o ided
us wi h in o ma ion abou hei gende , handedness, le el
o educa ion, and ma i al s a us. Pa icipan s ga e w i en
in o med consen and we e paid o pa icipa ion. The s udy
was conduc ed in acco dance wi h he Decla a ion o Helsinki
and app o ed by he Resea ch E hics Commi ee a Na ional
Taiwan No mal Uni e si y. A powe analysis was conduc ed
in G∗Powe 3.1.9.7 (Faul e al. 2007,2009) using a model o 1-
way analysis o a iance (ANOVA) wi h 3 g oups, each wi h a
sample size o 36. Wi h alpha se a 0.05, he powe o de ec a
medium-sized e ec ( = 0.25) was 0.63 and he powe o de ec
a la ge-sized e ec ( = 0.40) was 0.96.
Pa icipan s unde wen a neu opsychological e alua ion
o cogni i e unc ions. Thei wo king memo y was measu ed
wi h 3 sub es s in Wechsle Adul In elligence Scale-Fou h
Edi ion, including Digi Span, Le e -Numbe Sequencing, and
A i hme ic. Thei dep essi e symp oms we e measu ed wi h
he Cen e o Epidemiological S udies-Dep ession (CES-D) scale
(Radlo 1977;Chien and Cheng 1985). I is a 20-i em sel - epo
scale ha asks one o a e how o en o e he pas week he/she
expe ienced symp oms associa ed wi h dep ession. Response
op ions ange om1 o4 o eachi em(1 a elyo noneo
he ime [<1 day], 2 some o li le o he ime [1–2 days], 3
occasionally o a mode a e amoun o ime [3–4 days], 4 mos
o all o he ime [5–7 days]). Thei le el o pe cei ed s ess was
measu ed wi h he Pe cei ed S ess Scale (PSS) (Cohen e al.
1983;Chu 2005). I is a 14-i em sel - epo scale ha asks abou
one’s eelings and hough s du ing he las mon h. Response
op ions ange om 0 o 4 o each i em (0 ne e , 1 almos
ne e , 2 some imes, 3 ai ly o en, 4 e y o en). All es s we e
adminis e ed and sco ed by a licensed counseling psychologis
acco ding o s anda d p ocedu es. Demog aphic cha ac e is ics
and cogni i e unc ions o he pa icipan s a e summa ized in
Table 1.
Downloaded om h ps://academic.oup.com/ce co /ad ance-a icle/doi/10.1093/ce co /bhaa352/6014977 by Jy askyla Uni e si y use on 08 Decembe 2020
Age-Rela ed A enua ion o P edic ion E o s Hsu e al. 3
Figu e 1. Schema ic illus a ion o he audi o y local–global pa adigm. (A). On each ial, 5 sinusoidal ones o 50-ms du a ion we e p esen ed wi h a 200 ms SOA. Each
ial was ollowed by a ji e ed ITI o 700–1000 ms. (B). In each block, 2 di e en ypes o one quin e s we e p esen ed, he i s using he 5- ime epe i ion o he same
one (i.e., XXXXX) and he second wi h a change in he las one (i.e., XXXXY). Hal o he blocks con ained equen epe i ion and in equen change in he las one
(uppe : 80% XXXXX and 20% XXXXY) and hal con ained in equen epe i ion and equen change in he las one (lowe : 20% XXXXX and 80% XXXXY). The las
one in each one quin e (i.e., p obe) is ma ked in g ay as i is he one o in e es . Each block s a ed wi h a leas 20 equen one quin e s o es ablish he global
egula i y be o e he i s in equen one quin e appea ed. Pa icipan s we e ins uc ed o 1) iden i y in equen ials and 2) coun he numbe o in equen ials
in each block.
S imuli
A o al o 14 sinusoidal ones we e gene a ed using Sound Fo ge
P o 10.0 (Sony C ea i e So wa e Inc.). The du a ion o each one
was 50 ms (including 5 ms ise/ all imes). The equency o each
one was wi hin he ange o 261.626–987.767 Hz, ma ching he
absolu e equency o a se ies o 14 na u al keys on a mode n
piano (i.e., C4 D4 E4 F4 G4 A4 B4 C5 D5 E5 F5 G5 A5 B5) (Table 2).
Unlike p e ious esea ch using 2 ones in he audi o y local–
global pa adigm (Bekinsch ein e al. 2009;Wacongne e al. 2011;
Chennu e al. 2013;El Ka oui e al. 2015) (i.e., blocks o equen
epe i ion and in equen change: 80% AAAAA +20% AAAAB o
80% BBBBB +20% BBBBA; blocks o in equen epe i ion and
equen change: 20% AAAAA +80% AAAAB o 20% BBBBB +80%
BBBBA), he e we in oduced a ia ion in pi ch using 14 ones.
Speci ically, om he pool o 14 ones, 2 di e en ypes o one
quin e s we e c ea ed, he i s using he 5- ime epe i ion o he
same one (i.e., he XXXXX ype, which could be C4-C4-C4-C4-
C4, D4-D4-D4-D4-D4, E4-E4-E4-E4-E4, e c.) and he second wi h
a change in he las one (i.e., he XXXXY ype, which could be
C4-C4-C4-C4-A5, D4-D4-D4-D4-B4, E4-E4-E4-E4-C5, e c.). The e-
a e , o blocks o equen epe i ion and in equen change
(Fig. 1Buppe ) 80% o he ials we e andomly sampled om
he XXXXX ype and 20% om he XXXXY ype, whe eas o
blocks o in equen epe i ion and equen change (i.e., Fig. 1B
lowe ) 20% o he ials we e andomly sampled om he XXXXX
ype and 80% om he XXXXY ype. Such a ia ion in pi ch no
only dec eased s imulus-speci ic e ec s (ensu ing ha neu o-
physiological esponses indeed e lec p edic ion e o s pe se)
bu also inc eased ask di icul y (which helps e eal indi idual
di e ences in ask pe o mance).
P ocedu es
A o al o 1000 one quin e s we e p esen ed in 8 blocks in
andomized o de . Hal o he blocks con ained equen epe i-
ion and in equen change in he las one (i.e., Fig. 1Buppe :
80% XXXXX and 20% XXXXY) and hal con ained in equen
epe i ion and equen change in he las one (i.e.,Fig. 1Blowe :
20% XXXXX and 80% XXXXY). Bo h block ypes p esen ed a local
egula i y whe e he i h one could be iden ical o di e en
om p eceding ones (hence he local s anda d and he local
de ian ) and a global egula i y whe e one one quin e could be
mo e o less common han he o he (hence he global s anda d
and he global de ian ) (Fig. 1B). In each block, he numbe o
one quin e s a ied be ween 110 and 140, wi h he numbe
o equen one quin e s a ied be ween 88 and 112 and he
numbe o in equen one quin e s a ied be ween 22 and 28
o main ain he 80:20 equency a io. Each block s a ed wi h a
leas 20 equen one quin e s o es ablish he global egula i y
be o e he i s in equen one quin e appea ed.
A g ay ixa ion c oss agains black backg ound emained
on he sc een o he du a ion o each block ( iewed om a
dis ance o 120 cm). In each ial, a one quin e was p esen ed
wi h a 200 ms SOA and an in ensi y o maximum 82 dB (56–
82 dBA; 65–82 dBC). Each ial was ollowed by a ji e ed in e ial
in e al (ITI) o 700–1000 ms. Pa icipan s we e ins uc ed o 1)
iden i y in equen ials and 2) coun he numbe o in equen
ials in each block. The whole expe imen ook a ound 29 min
(i.e., 1000 ials ×1700 ms). E-p ime e sion 2.0 (Psychology So -
wa e Tools) was used o s imulus p esen a ion. S imula ion
was andomized indi idually o each pa icipan and deli e ed
binau ally ia headphones (Sennheise PX200-II).
Downloaded om h ps://academic.oup.com/ce co /ad ance-a icle/doi/10.1093/ce co /bhaa352/6014977 by Jy askyla Uni e si y use on 08 Decembe 2020
4Ce eb al Co ex, 2020, Vol. 00, No. 00
Table 1 Demog aphic cha ac e is ics and cogni i e unc ions o he pa icipan s
Senio s Adul s Adolescen s
Gende Male 10 13 20
Female 26 23 16
Handedness Righ 33 33 34
Le 1 2 2
Ambidex ous 2 1 0
Le el o educa ion P ima y 2 0 0
Seconda y192636
Te ia y 15 10 0
Ma i al s a us Single 3 35 36
Ma ied 27 1 0
Di o ced200
Widowed 4 0 0
Wo king memo y (Digi Span) Min 17 29 24
Max434647
Mean 30.39 36.56 36.64
SD 5.93 4.69 5.54
Wo king memo y (Le e -Numbe Sequencing) Min 13 16 17
Max263029
Mean 19.36 23.86 24.00
SD 3.02 3.36 2.81
Wo king memo y (A i hme ic) Min 9 11 12
Max232223
Mean 15.69 18.42 19.00
SD 4.03 2.64 2.56
Dep essi e symp oms (CES-D) Min 20 23 23
Max406858
Mean 27.92 35.58 34.14
SD 4.87 9.84 8.23
Le el o pe cei ed s ess (PSS) Min 0 17 14
Max354131
Mean 25.72 25.58 25.33
SD 6.21 5.18 4.60
Table 2 F equency (Hz) o each one
C4 D4 E4 F4 G4 A4 B4
F equency 261.63 293.67 329.63 349.23 392.00 440.00 493.88
C5 D5 E5 F5 G5 A5 B5
F equency 523.25 587.33 659.26 698.46 783.99 880.00 987.77
Da a Reco ding and Analysis
EEG Reco ding and P ep ocessing
EEG was eco ded om 62 sin e ed Ag/AgCl elec odes on a
Neu oscan quik-cap acco ding o he ex ended 10–20 sys em.
The g ound elec ode was placed a AFz and he e e ence
elec ode was placed be ween Cz and CPz. Eye mo emen s we e
moni o ed by addi ional 4 elec odes placed abo e and below he
le eye and a he ou e can hi o bo h eyes, which we e bipo-
la ized online o yield e ical and ho izon al elec ooculog am,
espec i ely.All signals we e ampli ied and online il e ed a 0.1–
100 Hz wi h he Neu oscan Synamps 2 ampli ie (Compumedics
Neu oscan) and sampled a 500 Hz.
Epochs ex ended om −900 o 498 ms ela i e o p obe onse ,
using a −100 o 0 ms p es imulus baseline. Ocula a i ac co -
ec ion was conduc ed wi h independen componen analysis
in EEGlab (Delo me and Makeig 2004). The da a we e ecom-
pu ed o a e age e e ence, high-pass il e ed a 0.5 Hz and
low-pass il e ed a 20 Hz. Epochs con aining ol age de ia ions
exceeding ±100 μV ela i e o baseline a any o he elec odes
we e ejec ed. The ial numbe s a e a i ac ejec ion in each
condi ion a e lis ed in Table 3.
S a is ical Analysis o E en -Rela ed Po en ial (ERP) Ampli ude
Local e ec was de ined as he di e ence be ween all local
de ian s and local s anda ds (a e aged ac oss global manip-
ula ion). Global e ec was de ined as he di e ence be ween
all global de ian s and global s anda ds (a e aged ac oss local
manipula ion). G oup di e ence in local and global e ec s
was de ined as local as well as global di e ence ERPs (i.e.,
de ian −s anda d) be ween senio s, adul s, and adolescen s.
Clus e -based pe mu a ion s a is ics we e pe o med in BESA
S a is ics 2.0 (B ain Elec ic Sou ce Analysis, Ge many) on he
ERPs om 0 o 498 ms ela i e o p obe onse . Clus e -based
pe mu a ion s a is ics ha e 2 essen ial s eps (Ma is and Oos en-
eld 2007). Fi s , he desi ed es (local and global e ec s: pai ed-
sample - es be ween de ian and s anda d; g oup e ec : 1-way
Downloaded om h ps://academic.oup.com/ce co /ad ance-a icle/doi/10.1093/ce co /bhaa352/6014977 by Jy askyla Uni e si y use on 08 Decembe 2020

Age-Rela ed A enua ion o P edic ion E o s Hsu e al. 5
Table 3 Range, mean, and SD o ial numbe s a e a i ac ejec ion in each condi ion
Senio s Adul s Adolescen s
Local s anda d
Global s anda d
Min 190 299 259
Max 420 414 424
Mean 361.56 367.36 375.64
SD 50.95 29.28 30.81
Local de ian
Global de ian
Min 42 71 63
Max 105 103 106
Mean 91.33 93.00 93.94
SD 12.82 7.98 8.05
Local s anda d
Global de ian
Min 59 70 64
Max 102 104 104
Mean 92.17 93.72 93.67
SD 9.92 7.55 8.46
Local de ian
Global s anda d
Min 211 298 235
Max 406 407 412
Mean 363.17 370.39 374.36
SD 42.61 30.14 33.37
ANOVA be ween senio s, adul s, and adolescen s) is un o e
all ime poin s and elec odes. Values below a ce ain h eshold
(clus e alpha 0.05) a e clus e ed based on empo al adjacency
(consecu i e ime poin s) and spa ial adjacency (elec ode dis-
ance 4 cm). All - alues/F- alues o he clus e s a e summed
o o m he clus e es s a is ic. Second, a pe mu a ion dis i-
bu ion o es ima e he clus e es s a is ic is gene a ed wi h
he boo s apping me hod,which andomly eassigns he condi-
ion/g oup labels and pe o ms he a o emen ioned es ing and
clus e ing p ocedu e again (numbe o pe mu a ions 3000).I he
obse ed clus e es s a is ic in con as o pe mu a ion dis i-
bu ion is <0.05/6 (local and global e ec s in 3 g oups)= 0.008, he
obse ed clus e is conside ed o be s a is ically signi ican .
S a is ical Analysis o Beha io al Measu es
One-way ANOVA (pe o med in SPSS 23) be ween senio s, adul s,
and adolescen s was conduc ed o examine g oup di e ence
on cogni i e unc ions (including wo king memo y, dep essi e
symp oms, and le el o pe cei ed s ess) as well as ask pe o -
mance (including iden i ica ion accu acy and coun ing e o s).
To u he explo e he ela ion be ween cogni i e unc ions
and EEG, we conduc ed a pa ial co ela ion be ween wo king
memo y and he size o local and global e ec s, whe e pa ic-
ipan s’ sco es on CES-D and PSS we e con olled o o emo e
he po en ial in luences o psychological ac o s (i.e., dep essi e
symp oms and le el o pe cei ed s ess) on cogni i e unc ions.
We also conduc ed a s epwise mul iple linea eg ession o
e alua e whe he all sco es on wo king memo y sub es s we e
necessa y o p edic he size o local and global e ec s.
Resul s
Cogni i e Func ions
Signi ican g oup di e ence was ound on all bu he las
measu e (Digi Span: F2,105 = 15.81, P<0.001; Le e -Numbe
Sequencing: F2,105 = 26.63, P<0.001; A i hme ic: F2,105 = 11.29,
P<0.001; CES-D: F2,105 = 9.52, P<0.001; PSS: F2,105 = 0.05, P= 0.95).
Pos hoc compa ison o he g oup di e ence using Tukey’s
Hones Signi ican Di e ence (HSD) es indica ed ha , in
wo king memo y sub es s and CES-D (bu no PSS), senio s
sco ed signi ican ly lowe han he 2 younge g oups, whe eas
he 2 younge g oups did no signi ican ly di e om each o he
(Table 4 and Fig. 2).
Task Pe o mance
Pa icipan s We e Asked o Iden i y In equen T ials
Pa icipan s could iden i y which one quin e s we e global
de ian s wi h an a e age accu acy o 0.91 ±0.02 s anda d
e o (SE) (senio s: 0.78±0.04 SE; adul s: 0.97 ±0.01 SE; ado-
lescen s: 0.98 ±0.01 SE). The e was a signi ican g oup di -
e ence (F2,105 = 22.93, P<0.001) whe e senio s showed lowe
iden i ica ion accu acy han he 2 younge g oups, whe eas
he 2 younge g oups did no signi ican ly di e om each
o he (Tukey’s HSD es : senio s–adul s: −0.18 ±0.03, P<0.001;
senio s–adolescen s: −0.20 ±0.03, P<0.001; adul s–adolescen s:
−0.02 ±0.03, P= 0.85).
Pa icipan s We e Asked o Coun he Numbe o In equen T ials
Pa icipan s’ coun ing e o s we e quan i ied as he absolu e
di e ence be ween he ac ual and epo ed numbe o global
de ian s. The a e age disc epancy was 6.47 ±0.56 SE (senio s:
8.99 ±1.00 SE; adul s: 5.49 ±0.97 SE; adolescen s: 4.93 ±0.79 SE).
The e was a signi ican g oup di e ence (F2,105 = 5.64, P<0.01)
whe e senio s showed la ge coun ing e o s han he 2 younge
g oups, whe eas he 2 younge g oups did no signi ican ly
di e om each o he (Tukey’s HSD es : senio s–adul s:
3.51 ±1.31, P<0.05; senio s–adolescen s: 4.06 ±1.31, P<0.01;
adul s–adolescen s: 0.55 ±1.31, P= 0.91).
EEG
Figu e 3 shows he opog aphical dis ibu ions o local and
global e ec s (i.e., de ian −s anda d) in each g oup. O e all, he
pa e n was simila ac oss g oups, wi h senio s di e ging mo e
om he 2 younge g oups. Local e ec mainly mani es ed as
a on ocen al nega i i y a a ound 100–150 ms (i.e., MMN),
a on ocen al posi i i y a a ound 200–300 ms (i.e., P3a),
and a on ocen al nega i i y a a ound 350–450 ms (i.e.,
eo ien ing nega i i y [RON]). Global e ec mainly mani es ed
as a on ocen al nega i i y a a ound 200 ms and a pa ie al
posi i i y a a ound 250–450 ms (i.e., P3b).
Downloaded om h ps://academic.oup.com/ce co /ad ance-a icle/doi/10.1093/ce co /bhaa352/6014977 by Jy askyla Uni e si y use on 08 Decembe 2020
6Ce eb al Co ex, 2020, Vol. 00, No. 00
Table 4 Pos hoc compa ison o he g oup e ec using Tukey’s HSD es (mean di e ence ±SE)
Senio s−adul s Senio s−adolescen s Adul s−adolescen s
Digi Span −6.17 ±1.28, P<0.001 −6.25 ±1.28, P<0.001 −0.08 ±1.28, P=1.00
Le e -Numbe Sequencing −4.50 ±0.72, P<0.001 −4.64 ±0.72, P<0.001 −0.14 ±0.72, P=0.98
A i hme ic −2.72 ±0.74, P≤0.001 −3.31 ±0.74, P<0.001 −0.58 ±0.74, P=0.71
CES-D −7.67 ±1.87, P<0.001 −6.22 ±1.87, P<0.010 1.44 ±1.87, P=0.72
Figu e 2. Raincloud plo s (Allen e al. 2019) o cogni i e unc ions measu ed wi h wo king memo y sub es s, CES-D, and PSS. Signi ican g oup e ec s we e ound in
all bu he las measu e (∗∗∗P<0.001).
Figu e 4Ashows he g and a e aged ERPs o local and global
e ec s on ep esen a i e elec odes (i.e., FCz and Pz) in each
g oup. Figu e 4Bshows he 2-D ( ime-by-pa icipan ) iews o
he ERPs o illus a e he pu a i e g oup di e ence in local and
global e ec s. Fo local e ec (uppe ), he e was a lack o on o-
cen al nega i i y a a ound 350–450 ms (i.e., RON) in senio s. Fo
global e ec (lowe ), he e was a lack o on ocen al nega i i y
a a ound 200 ms and a diminished pa ie al posi i i y a a ound
250–450 ms (i.e., P3b) in senio s.
Local and Global E ec s
Clus e -based pe mu a ion s a is ics sugges ed ha signi ican
local and global e ec s appea ed in all 3 g oups. Viola ion o he
local egula i y (Fig. 5Auppe ) modula ed 3 clus e s o ERPs in
senio s (clus e 1: 56–498 ms [P<0.0001]; clus e 2: 96–380 ms
[P<0.0001]; clus e 3: 374–498 ms [P<0.008]), adul s (clus e
1: 94–498 ms [P<0.0001]; clus e 2: 172–498 ms [P<0.0001];
clus e 3: 68–168 ms [P≤0.001]), and adolescen s (clus e 1: 98–
498 ms [P<0.0001]; clus e 2: 172–498 ms [P<0.0001]; clus e
3: 60–164 ms [P<0.001]). Viola ion o he global egula i y
(Fig. 5Alowe ) modula ed 2 clus e s o ERPs in senio s (clus e
1: 150–498 ms [P<0.0001]; clus e 2: 134–498 ms [P<0.0001]),
adul s (clus e 1: 102–498 ms [P<0.0001]; clus e 2: 124–498 ms
[P<0.0001]), and adolescen s (clus e 1: 96–498 ms [P<0.0001];
clus e 2: 124–498 ms [P<0.0001]).
G oup Di e ence in Local and Global E ec s
Clus e -based pe mu a ion s a is ics e ealed signi ican g oup
di e ence in bo h local and global e ec s. Local e ec s modu-
la ed 2 clus e s o ERPs (clus e 1: 14–498 ms [P<0.0001]; clus e
2: 262–422 ms [P<0.008]) (Fig. 5Buppe ). Pos hoc compa isons
showed signi ican di e ence be ween senio s and adul s (clus-
e 1: 274–498 ms [P<0.001]; clus e 2: 274–412 ms [P<0.008]) as
well as be ween senio s and adolescen s (clus e 1: 266–498 ms
[P<0.0001]; clus e 2: 262–422 ms [P≤0.001]). The e was no di -
e ence be ween he 2 younge g oups. Global e ec s modula ed
1 clus e o ERPs (130–498 ms [P<0.0001]) (Fig. 5Blowe ). Pos
hoc compa isons showed signi ican di e ence be ween senio s
and adul s (clus e 1: 166–498 ms [P<0.0001]; clus e 2: 176–
498 ms [P<0.001]) as well as be ween senio s and adolescen s
(132–498 ms [P<0.0001]). The e was no di e ence be ween he 2
younge g oups.
Co ela ion Be ween Cogni i e Func ions and EEG
To explo e he ela ion be ween cogni i e unc ions and EEG,
we conduc ed a pa ial co ela ion be ween wo king memo y
and he size o local and global e ec s, while con olling o
pa icipan s’ sco es on CES-D and PSS. Wo king memo y was
quan i ied as he sco es on wo king memo y sub es s. The size
o local and global e ec s was quan i ied as di e ence ERPs
(i.e., de ian −s anda d) on elec odes and ime poin s showing
Downloaded om h ps://academic.oup.com/ce co /ad ance-a icle/doi/10.1093/ce co /bhaa352/6014977 by Jy askyla Uni e si y use on 08 Decembe 2020
Age-Rela ed A enua ion o P edic ion E o s Hsu e al. 7
Figu e 3. Topog aphical dis ibu ions o local and global e ec s (i.e., de ian −s anda d) in each g oup. Scalp maps a e plo ed om 50 o 450 ms a e he onse o he
p obe (i.e., one o in e es ).
Table 5 Pa ial co ela ion able (∗P<0.05 a e Benjamini–Hochbe g p ocedu e)
Local e ec F2 o e 358 ±10 ms Local e ec P8 o e 360 ±10 ms Global e ec CP1 o e 460 ±10 ms
Digi Span (104) = −0.30∗ (104) = 0.28∗ (104) = 0.40∗
Le e -Numbe Sequencing (104) = −0.26∗ (104) = 0.31∗ (104) = 0.37∗
A i hme ic (104) = −0.20∗ (104) = 0.17 (104) = 0.35∗
he mos signi ican g oup di e ence in clus e -based pe mu-
a ion s a is ics (local e ec s: F2 o e 358 ±10 ms and P8 o e
360 ±10 ms; global e ec : CP1 o e 460 ±10 ms). The esul s
showed ha wo king memo y was signi ican ly ela ed o he
size o local and global e ec s (Table 5 and Fig. 6). Wo se wo k-
ing memo y was associa ed wi h smalle RON in local e ec
(mani es ing as smalle on ocen al nega i i y and smalle
pa ie al posi i i y) and smalle P3b in global e ec . No ably, no
co ela ion emained signi ican when we also con olled o
pa icipan s’ age in addi ion o hei sco es on CES-D and PSS.
A s epwise mul iple linea eg ession was conduc ed o e al-
ua e whe he all sco es on wo king memo y sub es s we e nec-
essa y o p edic he size o local and global e ec s. Fo local
e ec on F2 (i.e., on al nega i i y o he RON), Digi Span sco e
en e ed in o he eg ession model a s ep 1 and a signi ican
eg ession equa ion was ound (F1,106 = 12.81, P≤0.001, adjus ed
R2= 0.10). The eg ession equa ion was: p edic ed size o local
e ec = −0.07 ×Digi Span sco e +1.81. Fo local e ec on P8 (i.e.,
pa ie al posi i i y o he RON), Le e -Numbe Sequencing sco e
en e ed in o he eg ession model a s ep 1 and a signi ican
eg ession equa ion was ound (F1,106 = 12.03, P≤0.001, adjus ed
R2= 0.09). The eg ession equa ion was: p edic ed size o local
e ec = 0.09 ×Le e -Numbe Sequencing sco e−1.48. Fo global
e ec on CP1 (i.e., P3b), Digi Span sco e en e ed in o he eg es-
sion model a s ep 1 and a signi ican eg ession equa ion was
ound (F1,106 = 21.43, P<0.001, adjus ed R2= 0.16). The eg ession
equa ion was: p edic ed size o global e ec = 0.09 ×Digi Span
sco e−1.60.
Discussion
He e we used an audi o y local–global pa adigm o in es i-
ga e whe he he age- ela ed shi om senso ium o p edic-
ions occu s a all le els o hie a chical message passing. An
indisc imina e age- ela ed change should be associa ed wi h
Downloaded om h ps://academic.oup.com/ce co /ad ance-a icle/doi/10.1093/ce co /bhaa352/6014977 by Jy askyla Uni e si y use on 08 Decembe 2020
8Ce eb al Co ex, 2020, Vol. 00, No. 00
Figu e 4. (A). G and a e aged ERPs o local and global e ec s on ep esen a i e elec odes (i.e., FCz and Pz) in each g oup. Shaded a ea ep esen s he s anda d e o o
he mean. Time 0 ms ma ks he onse o he p obe (i.e., one o in e es ). Do ed lines ma k he onse o he 4 p eceding ones. (B). The 2-D ( ime-by-pa icipan ) iews
o he ERPs in which local and global e ec s (i.e., de ian −s anda d) a e so ed along pa icipan s’ age, smoo hed ac oss neighbo ing pa icipan s using a ec angula
(boxca ) mo ing a e age (smoo hing wid h = 10), and colo coded. The ace below each ERP image shows he g and-a e aged ERPs ac oss all pa icipan s. The ed do s
on scalp maps ma k he posi ion o ep esen a i e elec odes o ease o e e ence.
educed MMN and P3b, signaling educed p edic ion e o s due
o dec eased weigh ing o senso y inpu s a bo h senso y and
con ex ual le els. Measu ing hallma ks o hie a chical p edic-
ion e o s, we ound ha olde b ains show li le educ ion in
he MMN signaling senso y (i.e., i s -o de ) p edic ion e o s
bu signi ican a enua ion in he P3b signaling con ex ual (i.e.,
second-o de ) p edic ion e o s. The esul s sugges ha , in no -
mal aging, ou b ain as a p edic i e machine does no unde go
an indisc imina e decline bu a speci ic a enua ion o Bayesian
upda ing.
Age- ela ed changes in no el y de ec ion ha e been he sub-
jec o nume ous s udies. Ou indings ex end p e ious empi -
ical wo k by demons a ing he hie a chy-selec i e changes in
p edic i e p ocessing as we age. The de ec ion o local de iancy
seems la gely p ese ed in olde indi iduals a ea lie la ency
(including he MMN ollowed by he P3a bu no he RON).
The lack o age-dependen di e ence a ea lie la ency is in
line wi h p e ious esea ch whe e pa icipan s’ a en ion was
ocused on he audi o y de ian /s anda d s imuli o less con-
s ained on he co e ask (MMN: Pekkonen e al. 1993,1996;
Sch oede e al. 1995;Gun e e al. 1996;Kazme ski e al. 1997;
Amenedo and Díaz 1998; P3a: Mage e al. 2005;Be i e al. 2013).
These indings sugges ed ha audi o y disc imina ion abili y
emains e ec i e h oughou he li espan (a leas wi h he help
o a en ion). In e es ingly, he de ec ion o local de iancy did
di e ge by age a la e la ency, wi h olde indi iduals missing
he RON (Mage e al. 2005;Ge zmann e al. 2013). Mo eo e ,
pa icipan s’ wo king memo y was a signi ican p edic o o he
size o local e ec s a he RON, whe e wo se wo king memo y
was associa ed wi h smalle RON. Toge he , he MMN/P3a/RON
complex was hough o p o ide a neu ophysiological index
o he cascade o 3 main p ocesses in ol ed in in olun a y
a en ion con ols o de ian s (i.e., au oma ic change de ec ion,
o ien ing o a en ion, and eco e y om dis ac ion o eo ien -
ing o a en ion as a wo king memo y p ocess) (Sch öge and
Wol 1998;Sch öge e al. 2000;Be i and Sch öge 2003;Be i
e al. 2004;Be i 2008;Ho á h e al. 2008). The no mal MMN
and P3a bu he missing RON in olde indi iduals likely e lec s
hei in ac capaci y o de ec and o ien o senso y (i.e., i s -
o de ) p edic ion e o s; hei challenge lies in he mo e gene al
a emp o eco e om such pe u bing dis ac ion (i.e., local
de iancy) o e ocus on hei ask (i.e., o 1) iden i y in equen
ials and 2) coun he numbe o in equen ials in each block)
due o limi a ions in wo king memo y.
In con as , he de ec ion o global de iancy is clea ly com-
p omised in olde indi iduals, as hey showed lowe accu acy
in he iden i ica ion o global de ian s, mo e e o s in coun -
ing global de ian s, and a enua ed P3b. The a enua ed P3b
Downloaded om h ps://academic.oup.com/ce co /ad ance-a icle/doi/10.1093/ce co /bhaa352/6014977 by Jy askyla Uni e si y use on 08 Decembe 2020