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 4.0
h ps://c ea i ecommons.o g/licenses/by/4.0/
Cohe ence be ween b ain ac i a ion and speech en elope a wo d and sen ence le els
showed age- ela ed di e ences in low equency bands
© 2021 Massachuse s Ins i u e o Technology
Published e sion
Kolozs á i, O solya B; Xu, Weiyong; Ge ike, Geo gia; Pa iainen, Tiina; Nieminen,
Lea; Noi ay, Aude; Hämäläinen, Ja mo A
Kolozs á i, O. B., Xu, W., Ge ike, G., Pa iainen, T., Nieminen, L., Noi ay, A., & Hämäläinen, J. A.
(2021). Cohe ence be ween b ain ac i a ion and speech en elope a wo d and sen ence le els
showed age- ela ed di e ences in low equency bands. Neu obiology o Language, 2(2), 226-
253. h ps://doi.o g/10.1162/nol_a_00033
2021
RESEARCH ARTICLE
Cohe ence Be ween B ain Ac i a ion and Speech
En elope a Wo d and Sen ence Le els
Showed Age-Rela ed Di e ences
in Low F equency Bands
O solya B. Kolozs á i
1,2
, Weiyong Xu
1,2
, Geo gia Ge ike
1,2,3
, Tiina Pa iainen
1,2
,
Lea Nieminen
4
, Aude Noi ay
5
, and Ja mo A. Hämäläinen
1,2
1
Depa men o Psychology, Uni e si y o Jy äskylä, Finland
2
Cen e o In e disciplina y B ain Resea ch (CIBR), Uni e si y o Jy äskylä, Finland
3
Niilo Mäki Ins i u e, Jy äskylä, Finland
4
Cen e o Applied Language S udies, Uni e si y o Jy äskylä, Finland
5
Labo a o y o O al Language Acquisi ion (LOLA), Uni e si y o Po sdam, Ge many
Keywo ds: speech pe cep ion, de elopmen , magne oencephalog aphy, speech acking,
cohe ence, audi o y esponses
ABSTRACT
Speech pe cep ion is dynamic and shows changes ac oss de elopmen . In pa allel, unc ional
di e ences in b ain de elopmen o e ime ha e been well documen ed and hese di e ences
may in e ac wi h changes in speech pe cep ion du ing in ancy and childhood. Fu he , he e
is e idence ha he wo hemisphe es con ibu e unequally o speech segmen a ion a he
sen ence and phonemic le els. To disen angle hose con ibu ions, we s udied he co ical
acking o a ious sized uni s o speech ha a e c ucial o spoken language p ocessing in
child en (4.7–9.3 yea s old, N= 34) and adul s (N= 19). We measu ed pa icipan s’
magne oencephalog am (MEG) esponses o syllables, wo ds, and sen ences, calcula ed he
cohe ence be ween he speech signal and MEG esponses a he le el o wo ds and sen ences,
and u he examined audi o y e oked esponses o syllables. Age- ela ed di e ences we e
ound o cohe ence alues a he del a and he a equency bands. Bo h equency bands
showed an e ec o s imulus ype, al hough his was a ibu ed o he leng h o he s imulus and
no he linguis ic uni size. The e was no di e ence be ween hemisphe es a he sou ce le el
ei he in cohe ence alues o wo d o sen ence p ocessing o in e oked esponse o syllables.
Resul s highligh he impo ance o he lowe equencies o speech acking in he b ain
ac oss di e en lexical uni s. Fu he , s imulus leng h a ec s he speech–b ain associa ions
sugges ing me hodological app oaches should be selec ed ca e ully when s udying speech
en elope p ocessing a he neu al le el. Speech acking in he b ain seems decoupled om
mo e gene al ma u a ion o he audi o y co ex.
INTRODUCTION
B ain s uc u e and unc ion con inue o de elop in o ea ly adul hood, wi h some e idence o
di e en ajec o ies o he le and igh hemisphe es (Gog ay e al., 2004;Pang & Taylo ,
2000;Pa iainen e al., 2019). In adul s, impo an unc ional di e ences be ween he le
an open access jou nal
Ci a ion: Kolozs á i, O. B., Xu, W.,
Ge ike, G., Pa iainen, T., Nieminen, L.,
Noi ay, A., & Hämäläinen, J. A. (2021).
Cohe ence be ween b ain ac i a ion
and speech en elope a wo d and
sen ence le els showed age- ela ed
di e ences in low equency bands.
Neu obiology o Language,2(2),
226–253. h ps://doi.o g/10.1162
/nol_a_00033
DOI:
h ps://doi.o g/10.1162/nol_a_00033
Suppo ing In o ma ion:
h ps://doi.o g/10.1162/nol_a_00033
Recei ed: 17 July 2020
Accep ed: 17 Feb ua y 2021
Compe ing In e es s: The au ho s ha e
decla ed ha no compe ing in e es s
exis .
Co esponding Au ho :
O solya Bea ix Kolozs á i
[email p o ec ed]
Handling Edi o :
Da id Poeppel
Copy igh : © 2021
Massachuse s Ins i u e o Technology.
Published unde a C ea i e Commons
A ibu ion 4.0 In e na ional
(CC BY 4.0) license.
The MIT P ess
Downloaded om h p://di ec .mi .edu/nol/a icle-pd /2/2/226/1915943/nol_a_00033.pd by JYVASKYLAN YLIOPISTO KIRJASTO use on 12 May 2021
and igh hemisphe es ha e been demons a ed when p ocessing syllable and phonemic in-
o ma ion (e.g., Poeppel, 2014). Howe e , li le is known abou he de elopmen o his unc-
ional specializa ion in child en. Func ional nea -in a ed spec oscopy ( NIRS) and magne ic
esonance imaging (MRI) ha e p o ided e idence o a signi ican le wa d asymme y o
speech p ocessing ha is al eady p esen om bi h (Dehaene-Lambe z e al., 2002;Pena
e al., 2003). D awing on hese indings, we used magne oencephalog aphy (MEG) o examine
how hemisphe ic specializa ion is e lec ed in b ain esponses o a ious speech uni s (sen-
ence, wo ds, syllables) and o unco e whe he his specializa ion di e s be ween child en
and adul s. To achie e hose goals, we combined wo expe imen al app oaches: examining
gene al indices o audi o y ma u a ion as e lec ed in he age- ela ed changes o onse -
esponses (e en - ela ed ields [ERF]) o simple speech sounds alongside examina ion o wo d
and sen ence acking in di e en equency bands, as measu ed by cohe ence.
P e iously, long las ing ma u a ional e ec s ha e o en been s udied using he e en - ela ed
po en ials (ERPs) and hei magne ic equi alen ERFs o sho sounds wi h EEG and MEG. The
audi o y ERPs in in ancy and in he p eschool age show p ominen P1 and N2 esponses,
which as child en en e childhood s a o become ea lie in la ency and dec ease in ampli-
ude. Addi ionally, P1 and N2 esponses a e sepa a ed by eme ging N1 and P2 esponses
a ound he age o 8 o 9 yea s (Alb ech e al., 2000;Pon on e al., 2000).
Di e ences in hemisphe ic ma u a ion a es ha e also been obse ed using ERFs. The N1m
pa e ns measu ed wi h MEG we e mo e adul -like in 7- o 8-yea -olds in he igh hemisphe e
han in he le (Pa iainen e al., 2019). This sugges s ine-g ained de elopmen al ajec o ies
o he di e en audi o y egions wi h clea ly imma u e pa e ns o ac i a ion in he audi o y
co ex a ound ea ly school age (8 o 9 yea s old).
While s udying he e en - ela ed po en ials and ields in esponse o indi idual phonemes
and syllables is a use ul me hod o in es iga e he well-known ma u a ional e ec s o audi o y
p ocessing, audi o y in o ma ion in speech spans ac oss mul iple imescales encompassing
phonemes, syllables, wo ds, and ph ases. Mul i- ime esolu ion models o speech p ocessing
(Ghi za, 2011;Ghi za & G eenbe g, 2009;Poeppel, 2003;Poeppel & Assaneo, 2020) p opose
ha speech in o ma ion is p ocessed and in eg a ed in a hie a chical and in e dependen man-
ne by phase alignmen o neu al en ainmen o he in ol ed oscilla o y ne wo ks in he au-
di o y co ices wi h di e en specializa ion o he le and igh audi o y a eas.
Cohe ence analysis can be used o s udy speech pe cep ion in hese longe speech seg-
men s. Cohe ence is he compu a ion o synch ony be ween wo signals in he equency do-
main. The cohe ence alue e lec s he consis ency o phase di e ence be ween wo signals
(he e be ween he speech en elope and b ain ac i i y) a any gi en equency. This echnique
can be used o in es iga e acking o he speech signal in he b ain, which has been a gued o
e lec ele an linguis ic ope a ions such as pa sing and chunking o hie a chical linguis ic
s uc u es o speech (Bou guignon e al., 2013;Ding e al., 2016;G oss e al., 2013;Molina o
& Liza azu, 2018;Peelle & Da is, 2012).
Neu onal oscilla ions in equency bands p esen in speech (del a, 1–3 Hz, he a, 4–8 Hz,
be a, 15–30 Hz, and low gamma, 30–50 Hz; Poeppel, 2014) ha e been heo ised o p o ide a
basis o pa sing he con inuous speech signal in o di e en linguis ic uni s (e.g., del a: syllable
s ess pa e ns; he a: syllables; be a: onse - ime uni s; low gamma: phone ic in o ma ion;
Ghi za e al., 2013;Leong & Goswami, 2014;Poeppel, 2014;Poeppel & Assaneo, 2020).
In his amewo k, he linguis ic in o ma ion associa ed wi h he di e en imescales would
be hen in eg a ed o gi e he inal speech pe cep . Low equency co ical ac i i y appea s
o synch onise o he hy hms o mul iple linguis ic uni s (Ding e al., 2016,2017), while highe
Func ional nea -in a ed
spec oscopy ( NIRS):
Func ional neu oimaging echnique
using nea -in a ed spec oscopy
whe e ce eb al hemodynamic
esponses a e measu ed.
Cohe ence:
A alue ha e lec s how simila he
oscilla o y ac i i y p esen in wo
signals is.
E en - ela ed po en ial (ERP):
The b ain esponse o a p esen ed
s imulus, measu ed using
elec oencephalog aphy (EEG).
N1m:
Audi o y e oked esponse ela ed
o N1 esponse, measu ed using
magne oencephalog aphy
Neu obiology o Language 227
Age- ela ed di e ences in cohe ence
Downloaded om h p://di ec .mi .edu/nol/a icle-pd /2/2/226/1915943/nol_a_00033.pd by JYVASKYLAN YLIOPISTO KIRJASTO use on 12 May 2021
equencies (such as be a and gamma) may be mo e sensi i e o syn ac ic and seman ic in o -
ma ion (Ding e al., 2016). Toge he , hese esul s sugges ha du ing lis ening o connec ed
speech, he b ain synch onizes co ical hy hms o ack he hy hm o he di e en linguis ic
uni s (Ding e al., 2017).
Speech p ocessing in ol es bo h le and igh audi o y co ices (Poeppel, 2003;Poeppel &
Assaneo, 2020). In i s ea ly s age he ep esen a ion o he inpu speech signal has a bila e al
symme y, which hen b anches ou in subsequen p ocessing s eps. Le audi o y a eas ha e
been sugges ed o sample in o ma ion om sho (20–40 ms) in eg a ion windows (Gi aud
e al., 2007;Poeppel, 2003;Poeppel & Assaneo, 2020), and igh a eas o sample in o ma-
ion om longe (150–200 ms) in eg a ion windows (Gi aud e al., 2007;Luo & Poeppel,
2007;Poeppel, 2003;Poeppel & Assaneo, 2020). These di e ences a e e lec ed in oscilla o y
neu onal ac i i y in di e en bands (mos ly in gamma and he a bands, espec i ely).
Howe e , changes in b ain ac i i y ha e been epo ed du ing childhood wi h espec o
gene al audi o y sound p ocessing as well as mo e speci ic speech p ocessing (e.g., Ríos-
López e al., 2020;Uhlhaas e al., 2010). De elopmen al changes in neu al synch ony ha e
been demons a ed ( o a e iew, see Uhlhaase al.,2010) using audi o y s imula ion (Mülle
e al., 2009), whe eby young child en showed educed synch onisa ion in he del a and he a
(Mülle e al., 2009) equencies compa ed o adolescen s and adul s.
The e is con e ging e idence ha hemisphe ic specialisa ion o di e en windows o in e-
g a ion o audi o y in o ma ion and speech is p esen om he i s yea o li e; howe e , esul s
di e as o which hemisphe e shows he s onges esponse o long speech-like chunks
(Telkemeye e al., 2009,2011). The de elopmen al pa e n o hemisphe ic dominance o
p ocessing spoken sen ences seems o shi be ween b ain hemisphe es wi h age. G ea e
en ainmen o speech was ound in he le hemisphe e compa ed o igh in he he a band wi h
7-mon h-old in an s (Kalashniko a e al., 2018). Howe e , his specializa ion was no ound in
young child en be ween he ages o 4 and 7 yea s (Ríos-López e al., 2020) in he del a band.
Finally, a highe co ela ion in he igh as compa ed o he le hemisphe e be ween he ampli-
ude en elope o sen ences and hei co esponding b ain esponses was ound in olde 9- o
13-yea -old child en (Ab ams e al., 2008,2009).
Building on hose indings, he cu en s udy in es iga ed (a) age- ela ed di e ences and (b)
hemisphe ic balance in wo d and sen ence acking in low equency bands o sepa a e he
wo d o ph asal le els o p ocessing. Based on p e ious s udies on adul s and olde child en,
we expec ed hemisphe ic di e ences o al eady be p esen in 5- o 9-yea -olds in he del a
(1–4Hz)and he a(4–8 Hz) bands wi h he igh hemisphe e showing highe cohe ence han
he le hemisphe e.
To examine i and how he ma u a ion o he synch ony measu es is ela ed o he es ab-
lished ma u a ion o he onse esponse ( e lec ed in he changes in ERFs o syllables), we
compa ed he cohe ence alues o wo ds and sen ences wi h he age- ela ed changes in
he N1m esponse o syllables. E oked b ain ac i i y o sounds has been shown o change
om p eschool o school age and o adul hood. While he speci ic N1m esponse is absen
in ea ly childhood, i seems o eme ge a a ound 8 o 9 yea s o age and only become ully
ma u e in adul hood (Alb ech e al., 2000;Pon on e al., 2000). I he N1m ampli ude has a
common unde lying ma u a ional mechanism wi h he speech acking index, ou esul s
should show simila de elopmen al e ec s. On he o he hand, synch oniza ion o b ain ac-
i i y o speech could u ilize pa ly sepa a e b ain mechanisms ha ollow a di e en de el-
opmen al ajec o y and a e a ec ed mo e by en i onmen al inpu han by de elopmen al
changes e lec ed by N1m.
Neu obiology o Language 228
Age- ela ed di e ences in cohe ence
Downloaded om h p://di ec .mi .edu/nol/a icle-pd /2/2/226/1915943/nol_a_00033.pd by JYVASKYLAN YLIOPISTO KIRJASTO use on 12 May 2021
We also examined co ela ions be ween he p ocessing o speech en elopes and phono-
logical skills. Speech en elope p ocessing has been ela ed o segmen a ion in o syllable and
phoneme le el elemen s (Poeppel, 2014). As o phonological skills, b oadly de ined hey in-
clude he awa eness o a ious speech uni s (e.g., phonemes, syllables, wo ds), wo king mem-
o y ope a ions o speech sounds, and access o phonological ep esen a ions (e.g., Fowle ,
1991;Goswami & B yan , 2016; o a e iew, see Noi ay, Popescu e al., 2019). These a e
hough o be ep esen ed, o example, by apid naming, phoneme dele ion, and speech ep-
e i ion asks. Based on his, we hypo hesized ha speech en elope p ocessing could be linked
o phonological skill de elopmen (Goswami, 2011).
MATERIALS AND METHODS
Pa icipan s
Two age g oups pa icipa ed in he s udy: ypically de eloping child en and young adul s. The
adul s we e s udying a he Uni e si y o Jy äskylä, Finland. Table 1 shows he numbe o pa -
icipan s, mean age and age- ange, gende , handedness, and a e age hea ing le el o each
g oup. All pa icipan s we e Finnish na i e speake s.
The child en we e ec ui ed ia he Na ional Regis y o Finland and he adul s ia email
lis s o he uni e si y. Exclusion c i e ia a he ime o ec ui men we e head inju ies, ADHD o
lea ning di icul ies, neu ological diseases and medica ion a ec ing he cen al ne ous sys-
em, o any epo ed hea ing de ici s. Child en ec ui ed o he s udy we e ypically de elop-
ing and did no p esen any neu ological, cogni i e, o language- ela ed de iciency. In
addi ion, he hea ing le el o he pa icipan s was es ed using audiome y, wi h mos o hem
pe o ming a o below 25 dBs o 250 Hz, 500 Hz, 1000 Hz, and 2000 Hz sounds in he le
and igh ea s.
A e da a collec ion 13 pa icipan s we e excluded o e all, all o hem om he child
g oup. Fi e we e excluded based on he measu emen because o oo much mo emen and
inabili y o ollow ins uc ions du ing he eco ding, wo because o noisy da a, ou because o
echnical p oblems (ins umen a ion ailu e o so wa e issues), one based on inciden al ind-
ings du ing he measu emen s (based on he neu ologis ’s epo ), and one because o high
ampli ude luc ua ions in he da a.
En olmen in he s udy was olun a y; all adul s and child en pa icipan s as well as hei
pa en /ca egi e p o ided w i en in o med consen p io o hei pa icipa ion in he s udy.
Subsequen o he MEG s udy, all pa icipan s ecei ed ei he a mo ie icke o a gi ca d
Table 1. Desc ip ion o pa icipan s
Child en Adul s
# o pa icipan s included in he analysis (measu ed in MEG) 34 (47) 19 (19)
Mean age (SD) 7.53 (1.34) 24.80 (3.73)
Age ange (Minimum–Maximum; y = yea s, m = mon hs) 4y8m–9y4m 20y3m–35y2m
Gende a io (M:F) 18:16 2:17
Handedness (le :bo h: igh ) 5:1:28 0:1:18
A e age hea ing le el in DBs (le : igh ea ) 21.25:21.37 Sel - epo o no mal
hea ing le el
Neu obiology o Language 229
Age- ela ed di e ences in cohe ence
Downloaded om h p://di ec .mi .edu/nol/a icle-pd /2/2/226/1915943/nol_a_00033.pd by JYVASKYLAN YLIOPISTO KIRJASTO use on 12 May 2021

as compensa ion o hei pa icipa ion. Indi idual s uc u al MR images we e acqui ed om a
p i a e company o e ing MRI se ices (Synlab Jy äskylä). T1-weigh ed 3D-SE images we e
collec ed on a GE 1.5 T (GoldSeal Signa HDx ) MRI scanne using a s anda d head coil and
wi h he ollowing pa ame e s: TR/TE = 540/10 ms, lip angle = 90, ma ix size = 256 × 256,
slice hickness = 1.2 mm, sagi al o ien a ion.
This s udy was ca ied ou in acco dance wi h he Decla a ion o Helsinki and app o ed by
he E hical Commi ee o he Uni e si y o Jy äskylä, Finland.
Beha iou al Tes Ba e y
Fi s , we conduc ed a ba e y o beha iou al es s assessing he child en’s gene al cogni i e
abili ies, wi h an emphasis on language- ela ed skills. Fo a desc ip ion o he beha iou al
es s, see Table 2.
Th ee di e en age-app op ia e es s, WPPSI-III (Wechsle , 2003a), WISC-IV (Wechsle ,
2003b), and WAIS-IV (Wechsle , 2008), we e used o measu e pa icipan s’ isuo-spa ial ea-
soning and ocabula y, and wo es s, WISC-IV and WAIS-IV, we e used o wo king memo y.
The mo o de elopmen o he pa icipan s was es ed using sub es s om he De elopmen al
Neu opsychological Assessmen (NEPSY; Ko kman e al., 1998), he o o-mo o ask, and he
Table 2. Desc ip ion o beha iou al es sco es
Beha iou al measu e Sub es
Mean (SD)
sco e epo ed Child en Adul s
WPPSI-III, WISC-IV,
WAIS-IV
Block design sp 10.00 (3.82) 10.96 (3.15) 11.26 (3.21)
Vocabula y sp 11.36 (3.01) 10.46 (3.07) 13.79 (2.02)
Digi span sp x 9.96 (2.15) 11.26 (2.96)
NEPSY Repe i ion o nonsense wo ds sp 10.09 (2.43) x
O o-mo o ask sp 10.30 (2.81) x
NEPSY-II Visuo-mo o ask
(ca and mo o cycle)
combined sp 9.42 (3.02) x
Phonological p ocessing sp 11.33 (2.33) x
Repe i ion o sen ences # co ec 25.62 (3.08) x
Rapid au oma ized
naming (RAN)
Objec s ime (s) 64.25 (13.66) 34.37 (7.78)
Le e s ime (s) 34.70 (10.64) 18.90 (4.62)
Le e knowledge ask o al 22.67 (8.27) x
Lukilasse Wo d eading pe cen ile 46.38 (40.32) x
Pseudowo d lis eading o al 34.45 (13.98) x
Pseudowo d ex eading luency 99.59 (0.22) x
Lukilasse Dic a ion pe cen ile 58.10 (39.06) x
No e. sp: s anda d poin ; SD: s anda d de ia ion; WPPSI-III: Wechsle P eschool and P ima y Scale o In elligence (Wechsle , 2003a); WISC-IV: Wechsle
In elligence Scale o Child en (Wechsle , 2003b); WAIS-IV: Wechsle Adul In elligence Scale (Wechsle , 2008); NEPSY: Neu opsychological Assessmen es
ba e y I (Ko kman e al., 1998); NEPSY II (Ko kman e al., 2008); RAN: Rapid au oma ized naming (Denckla & Rudel, 1976); Lukilasse (Häy inen e al., 1999).
Neu obiology o Language 230
Age- ela ed di e ences in cohe ence
Downloaded om h p://di ec .mi .edu/nol/a icle-pd /2/2/226/1915943/nol_a_00033.pd by JYVASKYLAN YLIOPISTO KIRJASTO use on 12 May 2021
NEPSY II isuo-mo o ask (Ko kman e al., 2008). Pa icipan s’phonological p ocessing was
es ed using he NEPSY II sub es . To assess speed o lexical e ie al, he Rapid au oma ized
naming (RAN; Denckla & Rudel, 1976) Objec s and Le e s sub es s we e used. To measu e
memo y o sen ences, he NEPSY II Sen ence Repe i ion sub es was used.
Reading skills we e es ed using he wo d eading ask om he Lukilasse es ba e y
(Häy inen e al., 1999), he pseudowo d eading ask adap ed om TOWRE (To gesen e al.,
1999), and he pseudowo d ex eading ask (Eklund e al., 2015).
Fo a de ailed desc ip ion o he beha iou al es s, see Supplemen a y Ma e ial 1 (suppo ing in-
o ma ion can be ound online a h ps://www.mi p essjou nals.o g/doi/suppl/10.1162/nol_a_00033).
S imuli
Th ee ypes o s imuli cha ac e izing a ious empo al and linguis ic s uc u es we e used o
he speech acking ask: syllables, wo ds, and sen ences. Syllables a ied in consonan s’place
o a icula ion (mo ing om on o back: bilabial s op /p/, den al s op / /, and pala al s op /k/),
while he owel emained iden ical (/a/).
Each syllable was p esen ed 18 imes ( o al o 54 syllable p esen a ions), and wo ds s a ing
wi h he same syllables (18 wo ds o each syllable, o al o 54 wo ds), as well as 54 sen ences,
each s a ing wi h one o he wo d ca ego y s imuli. Fo a desc ip ion and exempla s o he
s imuli, see Table 3.
All wo ds we e common, e e yday nouns. The wo ds we e 2 o 3 syllables long. Sen ences
we e composed o 3 o 4 wo ds and always s a ed wi h a noun ollowed by a o m o he e b
“ o be”in he p esen ense. S imuli we e chosen wi h he help o an expe de elopmen al
linguis . The s imuli we e p oduced by a emale na i e Finnish speake . All s imuli we e sep-
a a e, unique okens p oduced sepa a ely. S imuli we e eco ded using a 44 kHz sampling
equency, 32-bi quan isa ion in a p o essional eco ding s udio. The sound iles we e cu in o
indi idual segmen s using P aa (Boe sma & Weenink, 2018).
The same syllables and wo ds we e used o each s imulus ype o ge compa able onse
e oked b ain esponses. To see he lis o s imuli used, see Supplemen a y Ma e ial 2.
P ocedu e
Expe imen al design
Each speech acking ial consis ed o a ixa ion c oss in he middle o he sc een o 500 ms,
hen an exclama ion ma k appea ed in he same space o 1,000 ms signalling ha a sound is
going o come soon, ollowed by he ixa ion c oss o 750 ms. The audi o y s imuli we e hen
Table 3. Desc ip ion o s imuli
S imulus
ype
A e age du a ion
(ms) SD Range (ms) Exempla s: Finnish English ansla ion
Syllable 209.33 25.58 185–236 ka, a, pa
Wo d 574.54 103.22 352–797 kala, paju, alo ish, willow, house
Sen ence 1,438.54 240.49 1,039–2,051 Kala on ak aa iossa.
Paju on aipuisa puu.
Talo on ai an uusi.
The ish is in he aqua ium.
A willow is a lexible ee.
The house is b and new.
Neu obiology o Language 231
Age- ela ed di e ences in cohe ence
Downloaded om h p://di ec .mi .edu/nol/a icle-pd /2/2/226/1915943/nol_a_00033.pd by JYVASKYLAN YLIOPISTO KIRJASTO use on 12 May 2021
p esen ed ia ea phones, wi h he ixa ion c oss on he sc een. The ixa ion c oss emained on
he sc een o 750 ms a e he end o sound. This was ollowed by a s ill image o a pa o
appea ing o 1,250–4,250 ms (p esen a ion du a ion depended on he ype o s imuli hea d)
which p o ided he cue o he pa icipan s o epea he p e iously hea d s imuli aloud (see
Figu e 1).
Pa icipan s we e ins uc ed o i s lis en o a speech sequence (i.e., a syllable, a wo d, o a
sen ence) and o epea i a e seeing he isual cue on he sc een (a pa akee ). The isual
s imuli we e p esen ed on a black backg ound wi h whi e s anda d cha ac e s (a c oss o ix-
a ion and an exclama ion ma k ale ing o he audi o y s imuli) in Times New Roman on and
a on size 64. The bi d s imuli we e 9 × 15 cm in size on he p ojec ion sc een. He e only he
ime-window o he audi o y s imulus p esen a ion was analysed.
Pa icipan s we e i s gi en ins uc ions and 6 p ac ice ials (2 o each ype o s imuli, p e-
sen ed in andom o de ). In he ac ual expe imen 162 s imuli ( he 3 syllables epea ed 18
imes each, 54 wo ds, and 54 sen ences) we e p esen ed in andom o de .
S imuli we e p esen ed in 9 blocks, wi h 2 longe b eaks a e 3 blocks and sho e b eaks
(du a ion de e mined by he pa icipan ) in be ween he blocks. Th ee blocks las ed app oxi-
ma ely 8 min, and i ook app oxima ely 30 min o comple e he ask, ins uc ions and p ac ice
included.
The ask was embedded in a child- iendly na a i e o s imula e child en’s a en ion and
mo i a ion o comple e he ask. Pa icipan s we e old hey a e eaching 3 pa o s how o
“speak.”Thei ask was o wai o he pa o o s a lis ening (when he cue appea ed on
he sc een) and hei ins uc ions included keeping eye-con ac wi h he pa o o make su e
he bi d is paying a en ion ( o minimize mo emen - ela ed a e ac s in he eco ding).
Fu he mo e, hey we e asked o epea wha hey hea d a a no mal speaking loudness
(i.e., o no mumble he syllables, wo ds, o sen ences) since he pa o s will “no be able o
lea n i hey don’ hea he speech p ope ly”( o be able o eco d he p oduc ion as clea ly as
possible). This also ensu ed he child en we e ully engaged in he ask. Co ec p oduc ion
was on a e age o child en 88.41% and o adul s 97.86%. A he end o each hi d block
(i.e., be o e he longe b eaks and he end o he es ), he pa o s “ epea ed”some o he hea d
sounds, which we e new sounds c ea ed by aising he pi ch o he o iginal s imuli, o gi e he
imp ession ha he pa o s we e he ones epea ing hem. The i s and second ime i was he
syllables, while a he end o he MEG eco ding i was one sen ence om each syllable ype.
Pa icipan s sa in a magne ically shielded, sound a enua ed oom unde he MEG helme ,
a a 68 deg ee posi ion. The s imuli we e p esen ed h ough inse ea phones (Ro el RA-1570
Figu e 1. Schema ic ep esen a ion o one ial o he expe imen al pa adigm. Da a analysis was
ocused on he ime-window du ing he s imuli p esen a ion, indica ed by he pic u e o he
loudspeake .
Neu obiology o Language 232
Age- ela ed di e ences in cohe ence
Downloaded om h p://di ec .mi .edu/nol/a icle-pd /2/2/226/1915943/nol_a_00033.pd by JYVASKYLAN YLIOPISTO KIRJASTO use on 12 May 2021
sys em; ea ips we e ER3-14B o child en and ER3-14A o adul s) a a com o able loudness
le el. The pa icipan s sa 1 m om he p ojec ion sc een. Du ing measu emen , a esea ch
assis an was also p esen in he oom when necessa y o he child en. P esen a ion so wa e
( e sion 18.1; Neu obeha io al Sys ems, Inc., Albany, CA, USA) was used o p esen he s im-
uli, unning on a Mic oso Windows compu e (sound ca d: Sound Blas e Audigy RX; ideo
ca d: NVIDIA Quad o K5200).Measu emen s we e ideo moni o ed o make su e pa icipan s
we e paying a en ion and doing he ask.
MEG eco ding
306-channel (102 magne ome e s and 102 plana g adiome e pai s) MEG da a we e eco ded
in a magne ically shielded oom using he Elek a Neu omag® TRIUX™sys em (Elek a AB,
S ockholm, Sweden) a he Cen e o In e disciplina y B ain Resea ch, a he Uni e si y o
Jy äskylä, Finland.
The head posi ion in ela ion o he senso s in he helme was moni o ed con inuously wi h
i e digi ised head posi ion indica o (HPI) coils a ached o he scalp. Th ee HPI coils we e
placed on he o ehead and one behind each ea . The posi ion o he HPI coils was de e mined
by h ee ana omic landma ks (nasion, le and igh p eau icula poin s) using he Polhemus
Iso ak digi al acke sys em (Polhemus, Colches e , VT) a he beginning o he eco ding. An
addi ional se o poin s (>100) andomly dis ibu ed o e he scalp was also digi ised. Elec o-
oculog am was eco ded wi h wo elec odes a ached diagonally close o he le and igh
eyes and one g ound elec ode a ached o he colla bone.
The sampling a e o he eco ding was 1000 Hz and a 0.03–330 Hz online band-pass il e
was used.
Da a Analysis
P e-p ocessing
All da a we e p e-p ocessed using he empo al ex ension o he signal space sepa a ion me hod
wi h bu e s o 30 s (Taulu & Kajola, 2005;Taulu e al., 2005) in Max il e 2.2™(Elek a AB)
o emo e ex e nal in e e ence and co ec o head mo emen s. Bad channels we e iden i ied
and econs uc ed by he Max il e p og am. Head posi ion was es ima ed in 200 ms ime-
windows and 10 ms s eps o mo emen compensa ion. Da a we e sa ed in h ee sepa a e
iles con aining h ee eco ding blocks. Ini ial head posi ion o he i s ile was used o ans-
o ming he head posi ion o he same posi ion ac oss he iles.
Da a we e p e-p ocessed using independen componen analysis (ICA) using as ICA algo-
i hm (Hy ä inen & Oja, 2000) o emo e eye blinks, ho izon al eye mo emen s, and ca diac
a i ac s in MNE Py hon (0.16.2; G am o e al., 2013), and he sepa a e MEG eco dings we e
conca ena ed. The es o he da a analysis was done in he FieldT ip oolbox (Oos en eld e al.,
2011) in MATLAB R2016 (h ps://www.ma hwo ks.com/).
The con inuous MEG eco ding was epoched o 100 ms be o e and 1,000 ms a e he onse
o sound in he syllable s imuli ( o analysis o he e oked ields), and 100 ms be o e he onse o
sound and 100 ms a e he end o he sound in he wo d and sen ence s imuli ( o analysis o
he equency con en s). Epochs we e isually inspec ed and bad ials we e ejec ed, wi h an
a e age o 2.18% o epochs ejec ed o he child en and 0.78% o epochs ejec ed o he
adul s. Da a we e low-pass il e ed a 45 Hz. The epoched da a was baseline co ec ed using
he 100 ms p eceding he onse o he s imuli.
Neu obiology o Language 233
Age- ela ed di e ences in cohe ence
Downloaded om h p://di ec .mi .edu/nol/a icle-pd /2/2/226/1915943/nol_a_00033.pd by JYVASKYLAN YLIOPISTO KIRJASTO use on 12 May 2021
The a e aged squa ed alues we e hen co ela ed wi h he co esponding hemisphe e’s
cohe ence alues in he equency bands. No signi ican co ela ions we e ound. (Fo he a-
ble o co ela ion coe icien s and p alues, see Supplemen a y Ma e ial 7, Table 7.1.)
Topog aphy o he a e ages was isually inspec ed hen o con i m he co ec N1m e-
sponse pa e n in each pa icipan , o le and igh hemisphe es sepa a ely.
The N1m esponse in he le hemisphe e was obse ed in 4 (11.76%) child en, wi h an
a e age la ency o 130 ms, and 17 (89.47%) adul s, wi h an a e age la ency o 104 ms. Six
(17.65%) child en wi h an a e age la ency o 151 ms showed an ac i a ion pa e n wi h an
opposi e cu en di ec ion o he adul -like N1m.
The N1m esponse in he igh hemisphe e was obse ed in 8 (23.53%) o he child en’s
e oked esponses, wi h an a e age la ency o 141 ms, and in 17 (89.47%) o he adul s’e oked
esponses, wi h an a e age la ency o 105 ms. Fi e (14.71%) o he child en wi h an a e age
la ency o 143 ms showed an ac i a ion pa e n wi h an opposi e cu en di ec ion o he
adul -like N1m.
To examine whe he he N1m ampli ude and cohe ence alues in he del a and he a bands
would ollow a simila de elopmen al pa e n, co ela ions we e calcula ed o quan i y he
possible de elopmen al ela ionship be ween he measu es. Cohe ence alues we e plo ed
Table 7. Resul s o epea ed measu es mixed ANOVA o he he a equency band a sou ce le el in he wo egions o in e es s
The a –Tempo al egion Main e ec s and in e ac ions d F alue p alue pa ial η2
Type 1,51 44.799 0.000 0.468
Hemisphe e 1,51 5.850 0.019 0.103
G oup 1,51 6.849 0.012 0.118
Type ×G oup 1,51 2.131 0.151 0.040
Hemisphe e ×G oup 1,51 0.743 0.393 0.014
Type ×Hemisphe e 1,51 0.253 0.617 0.005
Type ×Hemisphe e ×G oup 1,51 0.190 0.665 0.004
The a –In e io - on al egion Main e ec s and in e ac ions d F alue p alue pa ial η2
Type 1,51 50.638 0.000 0.498
Hemisphe e 1,51 0.540 0.465 0.011
G oup 1,51 14.688 0.000 0.224
Type ×G oup 1,51 0.001 0.977 0.000
Hemisphe e ×G oup 1,51 0.865 0.357 0.017
Type ×Hemisphe e 1,51 0.398 0.531 0.008
Type ×Hemisphe e ×G oup 1,51 0.003 0.960 0.000
No e. Bold alues emained signi ican a e alse disco e y a e co ec ion.
Neu obiology o Language 240
Age- ela ed di e ences in cohe ence
Downloaded om h p://di ec .mi .edu/nol/a icle-pd /2/2/226/1915943/nol_a_00033.pd by JYVASKYLAN YLIOPISTO KIRJASTO use on 12 May 2021

agains he N1m esponses in he child and adul g oups (see Figu e 8). No signi ican co e-
la ions we e ound be ween N1m ampli ude o syllables and del a and he a cohe ence alues
o wo ds and sen ences in ei he he le o igh hemisphe es a e co ec ion o mul iple com-
pa isons. (Fo he able o co ela ion coe icien s and p alues, see Supplemen a y Ma e ial 7,
Table 7.2.)
Figu e 5. Le panels: G and a e age o sou ce le el cohe ence alues o child en and adul s o wo ds and sen ences. (A) In he del a
equency band in he empo al egion o in e es . (B) In he del a equency band in he in e io - on al egion o in e es ; op ow: g and
a e ages o child en and adul s; bo om ow: g and a e ages o wo ds and sen ences. Wa me colou s e lec highe cohe ence be ween
he s imuli en elope and he b ain da a. Righ panels: Region o in e es highligh ed in pu ple (as de ined in he Desikan-Killiany A las;
Desikan e al., 2006). (C) Box plo s o a e aged cohe ence alues in he del a equency band in he empo al egion collapsed ac oss hemi-
sphe es. (D) Box plo s o a e aged cohe ence alues in he del a equency band in he in e io - on al egion collapsed ac oss hemisphe es
and ages ( op) o s imulus ypes (bo om). Bold lines deno e he median o he cohe ence alues; he bo om and op edges o he box indica e
he 25 h and 75 h pe cen iles, espec i ely. W = wo ds, S = sen ences, C = child en, A = adul s. (*** < 0.001, ** < 0.01)
Neu obiology o Language 241
Age- ela ed di e ences in cohe ence
Downloaded om h p://di ec .mi .edu/nol/a icle-pd /2/2/226/1915943/nol_a_00033.pd by JYVASKYLAN YLIOPISTO KIRJASTO use on 12 May 2021
Figu e 6. Le : G and a e age o sou ce le el cohe ence alues in he he a equency band. Wa me colou s e lec highe cohe ence be-
ween he s imuli en elope and he b ain da a. Top ow: G and a e ages o child en and adul s. Bo om ow: G and a e ages o wo ds and
sen ences. Righ : Region o in e es highligh ed in pu ple (as de ined in he Desikan-Killiany A las; Desikan e al., 2006) and box plo s o
a e aged cohe ences o he ROIs in he he a equency band collapsed ac oss hemisphe es and ages ( op) o s imulus ypes (bo om).
Bold lines deno e he median o he cohe ence alues; he bo om and op edges o he box indica e he 25 h and 75 h pe cen iles, espec-
i ely. Top plo : Ligh blue boxes show a e age cohe ence o child en (C), and da k blue boxes o he adul s (A). Bo om plo : Red boxes
ep esen a e age cohe ence alues o wo ds (W), and g een boxes o sen ences (S). (*** < 0.001, * < 0.05)
Table 8. Resul s o co ela ions be ween he cohe ence alues a sou ce le el and age in he
child en g oup
Co ela ion coe icien Sig N
Del a Temp 0.049 0.785 34
In - on −0.036 0.840 34
The a Temp 0.165 0.352 34
In - on 0.056 0.752 34
No e. Sig = signi icance. Temp = Tempo al. In - on = In e io - on al.
Figu e 7. Blue bu e ly plo s o he g oup-a e aged magne ome e s wi h he global mean ield powe (GMFP) ( ed line) and opog aphic
maps o he e oked esponses o he syllable s imuli o he wo age g oups (Child en, N= 34; Adul s, N= 19). The yellow boxes highligh
whe e he audi o y esponse was expec ed in he g oups based on he GMFP.
Neu obiology o Language 242
Age- ela ed di e ences in cohe ence
Downloaded om h p://di ec .mi .edu/nol/a icle-pd /2/2/226/1915943/nol_a_00033.pd by JYVASKYLAN YLIOPISTO KIRJASTO use on 12 May 2021
Sou ce le el
The esponses we e compa ed in a 2 (Hemisphe e: le , igh ) × 2 (G oup: Child en, Adul s)
epea ed measu es mixed ANOVA (see Table 10 and Figu e 9). No signi ican di e ences we e
ound.
The a e aged powe was hen co ela ed wi h he co esponding hemisphe e’s cohe ence
alues in he equency bands. No signi ican co ela ions we e ound. (Fo he able o co e-
la ion coe icien s and p alues, see Supplemen a y Ma e ial 7, Table 7.3.)
Table 9. Resul s o epea ed measu es mixed ANOVA o he a e aged squa ed esponses based
on he GMFP peaks a senso le el
Main e ec s and in e ac ions d F alue p alue pa ial η2
Hemisphe e 1,51 0.013 0.531 0.000
G oup 1,51 3.761 0.058 0.069
Hemisphe e ×G oup 1,51 0.414 0.523 0.008
Figu e 8. Line plo s o cohe ence alues in he le and igh hemisphe es o wo d and sen ence s imuli in compa ison o he squa ed am-
pli ude o he N1m (T
2
) in he pa icipan s who showed a P1m o N1m pa e n in hei ERF esponses. In each plo , he scale on he le side
shows squa ed ampli ude o N1m esponse, and he scale on he igh side shows cohe ence alues. Blue dashed line: cohe ence alues o
wo d s imuli; ed dashed line: cohe ence alues o sen ence s imuli; g een solid line: he squa ed N1m ampli ude. Top ow plo s show
cohe ence alues o he del a band, le and igh hemisphe es espec i ely; bo om ow plo s show cohe ence alues o he he a band,
le and igh hemisphe es espec i ely. Values a e o ganized in o de o he squa ed ampli udes.
Neu obiology o Language 243
Age- ela ed di e ences in cohe ence
Downloaded om h p://di ec .mi .edu/nol/a icle-pd /2/2/226/1915943/nol_a_00033.pd by JYVASKYLAN YLIOPISTO KIRJASTO use on 12 May 2021
Co ela ions o Sou ce Le el Cohe ence wi h Beha iou al Sco es
Beha iou al sco es in he Phonological p ocessing and Sen ence epe i ion asks did no co -
ela e wi h cohe ence alues om he del a and he a bands. RAN objec s did co ela e in-
e sely wi h bo h equency bands and bo h ROIs (see Table 11 and Figu e 10), bu when
age was con olled o , he co ela ion was no longe signi ican (see Table 12).
Cohe ence alues we e co ela ed wi h he child g oups sco es on NEPSY’s phonological
p ocessing ask and sen ence epe i ion ask (see Tables 13 and 14). No signi ican co ela ions
we e ound.
Table 10. Resul s o epea ed measu es mixed ANOVA o he a e aged squa ed esponses based
on he GMFP peaks a sou ce le el
Main e ec s and in e ac ions d F alue p alue pa ial η2
Hemisphe e 1,51 0.3976 0.531 0.008
G oup 1,51 0.105 0.747 0.002
Hemisphe e ×G oup 1,51 3.469 0.068 0.064
Figu e 9. G and a e aged sou ce le el ERFs o syllables. Wa me colou s e lec highe sou ce powe o he e en - ela ed ield. The ed a eas
highligh ed we e included in he egion o in e es (as de ined in he Desikan-Killiany A las; Desikan e al., 2006). Righ panels show he
a e age sou ce wa e o m (MNE es ima e) ex ac ed om he b ain egions. The blue shading ep esen s he s anda d e o o he mean,
and he yellow shading shows he ime-windows used o he N1m esponse.
Neu obiology o Language 244
Age- ela ed di e ences in cohe ence
Downloaded om h p://di ec .mi .edu/nol/a icle-pd /2/2/226/1915943/nol_a_00033.pd by JYVASKYLAN YLIOPISTO KIRJASTO use on 12 May 2021
Table 11. Co ela ions be ween pe o mance on RAN o objec s ( ime in seconds) and he cohe ence
alues om he wo egions o in e es in he del a and he a equency bands a sou ce le el
Co ela ion coe icien Sig N
Del a Temp −0.377 0.005 53
In - on −0.350 0.010 53
The a Temp −0.292 0.034 53
In - on −0.330 0.016 53
No e. Sig = signi icance. Temp = Tempo al. In - on = In e io - on al.
Table 12. Co ela ions be ween pe o mance on RAN o objec s and he cohe ence alues a
sou ce le el con olled o age
Co ela ion coe icien Sig N
Del a Temp −0.070 0.624 53
In - on 0.020 0.888 53
The a Temp −0.005 0.971 53
In - on 0.055 0.699 53
No e. Sig = signi icance. Temp = Tempo al. In - on = In e io - on al.
Figu e 10. Sca e plo o co ela ion be ween pe o mance on RAN objec s and cohe ence alues
a sou ce le el. Blue do s and line ep esen cohe ence alues om he empo al egion o in e es ,
and ed do s and line ep esen alues om he in e io - on al egion o in e es in he del a equency
band. G een do s and line ep esen alues om he empo al egion o in e es , and o ange do s and
line ep esen alues om he in e io - on al egion o in e es in he he a equency band.
Neu obiology o Language 245
Age- ela ed di e ences in cohe ence
Downloaded om h p://di ec .mi .edu/nol/a icle-pd /2/2/226/1915943/nol_a_00033.pd by JYVASKYLAN YLIOPISTO KIRJASTO use on 12 May 2021

DISCUSSION
This s udy in es iga ed whe he child en and adul s di e in o e all b ain ac i i y as well as in
le and igh audi o y co ex ac i i y while lis ening o a ious speech uni s ha a e essen ial
o spoken language p ocessing. Mo e speci ically, we examined how audi o y p ocessing o
wo ds and sen ences is e lec ed in he le el o cohe ence and hemisphe ic la e aliza ion
ac oss de elopmen , and how hese a e ela ed o he p ocessing o syllables, a bo h he sen-
so and sou ce le els. To his end, wo age g oups we e es ed o compa ison: child en be-
ween he ages o 4.7 and 9.3 yea s and adul s. Cohe ence is an in e es ing and use ul measu e
o he b ain’s abili y o ack he speech en elope ac oss di e en equency bands by quan-
i ying he simila i y in equency con en be ween b ain ac i i y and he speech en elope. The
highe he cohe ence be ween b ain ac i i y and he speech en elope, he be e he speech
acking.
Fi s , we ound an imp o emen wi h age in he b ain’s abili y o ack speech e idenced as
inc eased cohe ence alues in he del a and he a equency bands be ween b ain signal and
he speech en elope. Second, a he senso le el, whe e he whole hemisphe es we e exam-
ined, we ound an in e ac ion be ween hemisphe es and age g oups in he del a band wi h
adul s showing la ge alues a he le han he igh hemisphe e. Howe e , a he sou ce le el,
hemisphe ic di e ences in cohe ence alues did no in e ac wi h age, which sugges s no di -
e ences in ma u a ion a es o he le and igh audi o y and on al co ices in he deg ee o
which he b ain can synch onize o he speech en elope. Thi d, we also ound di e ences in
he cohe ence alues obse ed o he wo d and sen ence s imuli independen o age, al-
hough his was a ibu ed o physical s imulus leng h a he han linguis ic uni size, sugges ing
ha he me hodological app oach should be aken in o accoun when in e p e ing indings
Table 14. Co ela ions be ween pe o mance on he epe i ion o sen ences ask and he
cohe ence alues a sou ce le el
Co ela ion coe icien Sig N
Del a Temp 0.097 0.585 34
In - on 0.200 0.256 34
The a Temp 0.076 0.670 34
In - on −0.055 0.757 34
No e. Sig = signi icance. Temp = Tempo al. In - on = In e io - on al.
Table 13. Co ela ions be ween pe o mance on he phonological p ocessing ask and he
cohe ence alues a sou ce le el
Co ela ion coe icien Sig N
Del a Temp 0.037 0.836 34
In - on −0.058 0.743 34
The a Temp 0.234 0.182 34
In - on 0.131 0.462 34
No e. Sig = signi icance. Temp = Tempo al. In - on = In e io - on al.
Neu obiology o Language 246
Age- ela ed di e ences in cohe ence
Downloaded om h p://di ec .mi .edu/nol/a icle-pd /2/2/226/1915943/nol_a_00033.pd by JYVASKYLAN YLIOPISTO KIRJASTO use on 12 May 2021
abou speech pe cep ion. Las , we ound no ela ionship be ween he gene al ma u a ion o
audi o y p ocessing and speech acking as indica ed by ea ly ERFs o syllables.
De elopmen al di e ences we e ound as an o e all inc ease in he cohe ence alues in
bo h he del a and he a equency bands be ween adul s and child en, wi h adul s showing
la ges alues compa ed o child en. Fu he , he opog aphy o he cohe ence o hese e-
quencies exhibi ed a clea pa e n o audi o y co ex ac i a ion a he senso le el. This was
mos ly con i med by he sou ce le el analysis. The cohe ence alues e lec how simila he
equency con en s be ween he b ain signal and he speech en elope a e; he e o e, ou ind-
ings could be in e p e ed as inc eased p ecision o he audi o y sys em o ack he speech om
childhood o adul hood. Howe e , when examining he cohe ence alues as a con inuous
a iable wi hin he child g oups, we did no ind any co ela ion be ween age and cohe ence
alues. The e may be se e al easons o he obse ed di e ences be ween adul s and
child en.
Fi s , basic audi o y p ocessing ma u es slowly wi h majo changes in, o example, ERP
esponses no ed a a ound ages 8–9 yea s wi h u he changes un il la e adolescence
(Pon on e al., 2000). This slow ma u a ion o basic p ocessing could a ec he p ecision o
speech p ocessing in a bo om-up manne .
Second, he bo om-up p ocess could be a ec ed by gene ically d i en ma u a ion o con-
inued exposu e o speech ha e ines he bo om-up pa hway o he audi o y sys em (Kuhl,
2000;Pon on e al., 2000). A he same ime, con inuous exposu e o speech e ines and
changes he b ain’s abili y o pe cei e speech in a op-down manne as well (Kuhl, 2000).
This en i onmen al inpu would shape long- e m memo y ep esen a ions, he e o e a ec ing
speech p ocessing.
Las , he de elopmen o speech acking may in e ac wi h o he co-de eloping cogni i e
and language- ela ed abili ies (e.g., ecep i e and exp essi e ocabula y, speech mo o and
phonological de elopmen s) in addi ion o ma u a ional ac o s such as age. The e is o in-
s ance e idence ha child en ini ially p ocess la ge uni s ha a e lexically based (e.g., wo ds)
be o e de eloping ep esen a ions o smalle uni s (syllables, indi idual phonemes; o a e-
iew, see Vihman, 2017). This p ocess may also be a ec ed by eading acquisi ion ha places
emphasis on phonemes (e.g., B ennan e al., 2013;Popescu & Noi ay, 2019;Ziegle &
Goswami, 2005). Fu he , in speech p oduc ion esea ch in es iga ing he size o coa icula-
o y uni s ac oss age, Noi ay and colleagues (Noi ay e al., 2018;Noi ay, Wieling e al., 2019)
no ed ha child en do no ma u e hei coa icula o y pa e ns in a linea ashion. Ins ead, hey
ound ha p eschoole s a he age o 3, 4, and 5 o ganised hei speech in la ge chunks com-
pa ed o p ima y school child en a he age o 7 and adul s (Noi ay e al., 2018;Noi ay,
Wieling e al., 2019).
In a subsequen s udy, Noi ay and colleagues u he demons a ed ha he de elopmen o
child en’s phonological awa eness, ha is, he awa eness ha he na i e language, is composed
o a ious size compounds (e.g., syllables, hyme, and indi idual phonemes) and he abili y o
manipula e hose uni s in e ac s wi h child en’s speech mo o o ganisa ion (Noi ay, Popescu
e al., 2019). G ea e awa eness o indi idual phonemes was associa ed wi h g ea e phonemic
di e en ia ion o a icula o y ges u es. To summa ise, ela ionships be ween se e al cogni i e
and language- ela ed abili ies occu in he cou se o language acquisi ion, and hey seem o
e ol e dynamically o e ime (e.g., Noi ay, Popescu e al., 2019;Noi ay, Wieling e al.,
2019;Vihman, 2017). In u u e esea ch, i will be impo an o in es iga e la ge samples o
child en spanning kinde ga en o p ima y school o be e unde s and he dynamics o hese
ela ionships and how hey con ibu e o he de elopmen o speech acking speci ically.
Neu obiology o Language 247
Age- ela ed di e ences in cohe ence
Downloaded om h p://di ec .mi .edu/nol/a icle-pd /2/2/226/1915943/nol_a_00033.pd by JYVASKYLAN YLIOPISTO KIRJASTO use on 12 May 2021
Ou esea ch p o ides supplemen a y in o ma ion abou he p ocessing o a ious speech-
sized uni s. Mo e speci ically, we con i med he ole o he lowe equency bands o sen ence
p ocessing (Molina o & Liza azu, 2018;Ríos-López e al., 2020) and ex ended his inding o
he p ocessing o wo ds. Indeed, he e is e idence ha he a and del a bands play a main ole
o pa sing he con inuous speech signal in o linguis ic and p osodic uni s (Poeppel, 2014;
Poeppel & Assaneo, 2020). Thus, a de elopmen al inc ease in b ain cohe ence in hese
equency bands could be associa ed wi h a de elopmen in he p ocessing and awa eness
o hose dis inc speech uni s. While we did no ind any signi ican co ela ion be ween chil-
d en’s cohe ence alues and hei pe o mance on phonological p ocessing o sen ence ep-
e i ion, u u e s udies should u he examine he ela ionship among phonological awa eness,
eading, and speech acking in he b ain wi h la ge samples o child en and longi udinally.
We also ound ha cohe ence was highe o wo ds han sen ences o all equency bands.
This was somewha su p ising gi en longe s imuli should p o ide oppo uni ies o b ain ac i i y
o lock o he ongoing audi o y signal. I is impo an , howe e , o no e ha a e checking he
cohe ence a he beginning o sen ences ials (wi h he same leng h as used o wo ds), we no ed
ha cohe ence inc eased compa ed o he o iginal alues o sen ences. This sugges s ha he
highe cohe ence o wo ds han sen ences does no e lec di e ences ha would be di ec ly
ele an o neu al compu a ion o linguis ic uni s, bu mo e he cha ac e is ics o calcula ing he
cohe ence measu e o sho e sus long s imuli. Fo example, longe s imuli p o ide g ea e
chances o b ain ac i i y un ela ed o s imula ion o occu , wi h highe likelihood o his noise
in he b ain ac i i y in e ac ing wi h he cohe ence measu e. The e o e, compa ison o cohe -
ence measu es ac oss di e en leng h s imuli should be done wi h ca e, as pu e physical leng h
o he s imulus migh ha e an e ec on he esul s. In gene al, ou indings con i m ha speech
acking can happen a a sho e leng h, such as wo ds, as well as a sen ence le el.
While we expec ed o ind a signi ican in e ac ion be ween he cohe ence alues in he le
and igh hemisphe es and he wo age g oups in he del a band, his di e ence was in he
opposi e di ec ion om ou p edic ions, whe e we expec ed la ge alues in he igh hemi-
sphe e compa ed o he le , pa icula ly in he adul s (Luo & Poeppel, 2007). We obse ed
signi ican ly la ge cohe ence alues in he le hemisphe e han igh o adul s in he del a
band only a he senso le el— his was no obse ed a he sou ce le el. One possible eason
o his di e ence be ween senso and sou ce le el indings could be he selec ion o channels
o egions used in he analysis. The senso le el compa ison used an o e all a e age o he
hemisphe es, while he sou ce le el ocused on he empo al egions. A he sou ce le el,
when ocusing on he empo al egions, a hemisphe ic di e ence was ound in he expec ed
di ec ion o la ge igh side ac i a ion compa ed o he le ; howe e , he di e ence was no
longe signi ican a e FDR co ec ion.
Finally, we in es iga ed he o e lap o di e en ma u a ional p ocesses ac oss linguis ic
uni s by examining he age- ela ed changes in b ain ac i i y a ound 125 ms ( he ime-window
o he N1m esponse in adul s o he syllables) and compa ed hose o he cohe ence alues o
wo ds and sen ences. The ampli udes o e oked esponses and he cohe ence alues likely
ep esen di e en neu onal mechanisms. The i s p esumably ep esen s a mo e gene al ma -
u a ion o he audi o y and speech pe cep ion sys em, and he la e is likely linked o op-
down p ocesses such as comp ehension o speech (Luo & Poeppel, 2007;Peelle e al.,
2013;Pon on e al., 2000,2002). We compa ed he de elopmen o he e oked esponses
and cohe ence using wo app oaches. In ou i s app oach using GMFP, which included bo h
he P1m and N1m esponses, we ound no signi ican di e ence be ween he g oups.
Howe e , he P1m and N1m likely ep esen di e en compu a ional p ocesses o he sounds.
In ea ly childhood he audi o y ERPs show p ominen P1 and N2 peaks. Du ing de elopmen ,
Neu obiology o Language 248
Age- ela ed di e ences in cohe ence
Downloaded om h p://di ec .mi .edu/nol/a icle-pd /2/2/226/1915943/nol_a_00033.pd by JYVASKYLAN YLIOPISTO KIRJASTO use on 12 May 2021
he P1 esponse shi s o ea lie la encies accompanied by a dec ease in ampli udes, and he
p ominen N1 esponse eme ges o he wa e o m a a ound ea ly school yea s (Alb ech e al.,
2000;Pon on e al., 2000). Impo an ly, P1m o young child en and N1m o olde child en and
adul s show e y simila iming, obscu ing he in e p e a ion o pu ely GMFP-based in e p e-
a ions (Pa iainen e al., 2019).
The e o e, as a nex s ep, we checked whe he he spa ial pa e ns and imings o esponses
in he le and igh hemisphe es o each indi idual ma ched wi h he expec ed N1m pa e n
based on he cu en di ec ion in he magne ome e opog aphy. We ound no sys ema ic co -
ela ion be ween he esponses in he ime-window o he i s p ominen e oked ield ( he
N1m o P1m) and cohe ence be ween he speech en elope in he del a and he a bands.
Al hough he co ela ions we e no signi ican i should be no ed ha se e al ac o s migh
a ec he esul , such as sample size and me hodology used. The ERFs and cohe ence we e
examined using di e en app oaches (ERFs in he ime domain and cohe ence in he equency
domain). I is possible ha he use o hese di e en app oaches makes he measu es di icul o
compa e di ec ly. Taking his in o conside a ion, ou esul s sugges ha he e oked esponse
o syllables and he speech acking migh de elop independen ly o each o he and no sha e
obus ma u a ional mechanisms. I his is he case, he ERF ampli udes could e lec mo e
bo om-up p ocesses while he cohe ence alues mo e op-down p ocesses. P e ious li e a u e
shows ha ERFs a e clea ly modula ed by he physical ea u es o he sounds (Nää änen &
Pic on, 1986;Nää änen e al., 1997), and he speech en elope ollowing seems o be linked
o speech in elligibili y and a en ion (Peelle e al., 2013).
Fu he mo e, especially in he case o younge child en, while he GMFP did show a e-
sponse a ound 100 ms, indi idual inspec ion o he esponses showed ha he esponse a he
ime was no ac ually an N1m esponse, bu a he P1m. Taking his in o accoun , i could
pe haps explain why we ound no di e ences be ween he g oups when compa ing he e-
sponses based on ime-windows de ined by only he peak in he GMFP.
Likewise, no hemisphe ic di e ences we e obse ed in any o he age g oups o he GMFP-
based alues o he e oked esponse, in con as o a p e ious MEG s udy (Pa iainen e al.,
2019). Howe e , he di e ence be ween hese s udies mos likely e lec s he chosen analysis
app oach. While he GMFP e lec s he o e all esponse s eng h a he senso le el, a he
sou ce le el, measu es o equi alen cu en dipoles depic he spa ially speci ic ampli ude
alues a di e en ime poin s. Indeed, ou da a demons a ed a simila ly delayed pa e n o
N1m opog aphy in child en, wi h mo e clea esponse in he igh han le hemisphe e, as
was implied by Pa iainen and colleagues (2011,2019). Inspec ion o he esponses hem-
sel es e ealed ha only abou one hi d o he child en ac ually had he N1m e oked
esponse. Due o ou sample size, compa ison using b ain esponses o child en who indeed
p oduced he N1m esponse would be unbalanced, and u u e esea ch should look in o his
compa ison wi h a la ge sample size o bo h g oups.
One o he limi a ions o ou s udy is ela ed o he ype o s imulus we used, as wo ds u -
e ed in isola ion a e mo e p onounced han hose in a sen ence. Howe e , ou pos hoc com-
pa ison o he cohe ence alues o sen ences a wo d-leng h ials s. sen ence-leng h ials
e ealed ha highe cohe ence was ound a he beginning o he sen ence ega dless o s im-
ulus ype. I is possible he wo d le el s imuli could be a ec ed by hei sho leng h in he
es ima ion o low equencies, and he e o e he esul s o he wo d s imuli should be ega ded
wi h cau ion. Ano he po en ial limi a ion is he numbe o pa icipan s limi ing he powe o
he s udy o de ec mo e sub le di e ences ela ed o de elopmen o hemisphe ic p ocessing
o he di e en s imuli.
Neu obiology o Language 249
Age- ela ed di e ences in cohe ence
Downloaded om h p://di ec .mi .edu/nol/a icle-pd /2/2/226/1915943/nol_a_00033.pd by JYVASKYLAN YLIOPISTO KIRJASTO use on 12 May 2021