Listening effort in children and adults in classroom noise

Lis ening e o in child en and
adul s in class oom noise
Julia Sei z, Ka in Loh & Janina Fels
I is well known ha hea ing in noisy si ua ions is mo e challenging han in quie en i onmen s.
This holds ue o adul s and especially o child en. This s udy employed a child-app op ia e dual-
ask pa adigm o in es iga e lis ening e o in child en aged six o en yea s and young adul s. The
p ima y ask in ol ed wo d ecogni ion, while he seconda y ask e alua ed digi ecall. Addi ionally,
subjec i e pe cep ion o lis ening e o was assessed using a child-app op ia e ques ionnai e. This
s udy inco po a ed plausible sound ep oduc ion and examined class oom scena ios including mul i-
alke babble noise wi h wo signal- o-noise a ios (0 dB and −3 dB) in an anechoic and an acous ically
simula ed class oom en i onmen . Fo y- ou p ima y school child en aged six o en (17 i s - o
second-g ade s and 18 hi d- o ou h-g ade s) and 25 young adul s pa icipa ed in his s udy. The
esul s e ealed di e ences in lis ening e o be ween he noise condi ions in hi d- o ou h-g ade s
and suppo ed using he dual- ask pa adigm o ha age g oup. Fo all h ee age g oups, a g ea e
subjec i e pe cep ion o lis ening e o in noise was ound. Fu he mo e, a co ela ion be ween he
subjec i e pe cep ion o lis ening e o and beha iou al lis ening e o based on he expe imen al
esul s was ound o hi d- o ou h-g ade s and adul s.
Keywo ds Lis ening e o , Dual- ask pa adigm, Audi o y cogni ion, Hea ing in child en
Child en spend a la ge pa o hei day a school, especially wi h he inc easing a ailabili y o a e -school
ca e. All ac i i ies in school, lessons, and b eaks, in ol e communica ion and lis ening o pee s o eache s.
Howe e , communica ion in schools is o en dis u bed by noise1. I is al eady known ha noise in schools a ec s
child en’s heal h, de elopmen , and school pe o mance1. Addi ionally, esea ch has shown ha child en a e
much mo e impai ed by noise du ing speech pe cep ion han adul s which can a ec hei abili y o p ocess
spoken in o ma ion e ec i ely2,3. This heigh ened suscep ibili y o noise aises he issue o inc eased p ocessing
esou ces being demanded o child en in a noisy en i onmen . These p ocessing esou ces can be in e p e ed as
e o , leading o he ollowing ques ion: How e o ul is lis ening o child en a school?
Gagné e al.4 de ined lis ening e o as he amoun o p ocessing esou ces (pe cep ual, a en ional, cogni i e,
e c.) alloca ed o a speci ic audi o y ask, when he ask demands a e high and when he lis ene s i es o each a
high le el o pe o mance on he lis ening ask. Cu en ly, he e a e h ee app oaches o measu e lis ening e o 4:
subjec i e (e.g. by ques ionnai es, see5), psychophysiological me hods (e.g., by elec oencephalog aphy (EEG)6),
and beha iou al me hods (e.g., by lis ening expe imen s7,8).
Two beha iou al me hods a e well-es ablished in he li e a u e o measu ing lis ening e o : single- ask
and dual- ask pa adigms9. Single- ask pa adigms o en u ilize speech ecogni ion asks assessing accu acy and
esponse imes o di e en noise scena ios. As demons a ed in s udies by Gus a son e al.8 wi h child en aged
se en o wel e, P odi and Visen in10 wi h child en aged eigh o en, and P odi e al.11 wi h child en aged i e
o se en, hese esponse imes a e hen used o p edic lis ening e o o he di e en noise scena ios. Dual-
ask pa adigms o measu e lis ening e o use a speech ecogni ion ask as p ima y ask and di e en ypes o
seconda y asks. Fo example, Hicks and Tha pe12 used a eac ion ask o i e- o ele en-yea -olds and Choi
e al.13 used a digi ecall o child en aged se en o ou een. The dual- ask pa adigm is based on he heo y o
limi ed cogni i e esou ces by Kahneman14 s a ing ha only limi ed esou ces a e a ailable o addi ional asks
i inc eased e o is spen on lis ening. Kahneman’s heo y was u he de eloped in o he amewo k o ease
o language unde s anding (ELU)15 and la e e ined in o he F amewo k o Unde s anding E o ul Lis ening
(FUEL)16.
To shed ligh on which beha iou al me hod o use, McGa igle e al.17 compa ed whe he dual- ask
pa adigms including isual esponse imes (measu ed in a isual moni o ing ask as seconda y ask) o single
ask pa adigms including e bal esponse imes (measu ed in a wo d ecogni ion ask) we e mo e sensi i e o
de ec di e ences in lis ening e o o child en aged six o hi een yea s. They sugges ed ha single ask e bal
esponse imes we e mo e sensi i e o he e ec s o backg ound noise han dual- ask isual esponse imes. This
Ins i u e o Hea ing Technology and Acous ics, RWTH Aachen Uni e si y, 52062 Aachen, Ge many. email:
[email p o ec ed]
OPEN
Scien i ic Repo s | (2024) 14:25200 1
| h ps://doi.o g/10.1038/s41598-024-76932-7
www.na u e.com/scien i ic epo s
aligns wi h he ecommenda ions summa ised in he li e a u e e iew by F ancis and Lo e9 who s a ed ha he
seconda y asks should d aw equally om he same pool o cogni i e esou ces as he p ima y ask in a dual- ask
pa adigm measu ing lis ening e o .
Since a majo in luence on lis ening e o is he ype and le el o backg ound noise, se e al dual- ask
app oaches esul ed in ambiguous esul s. Hicks and Tha pe12 we e among he i s o in es iga e lis ening e o
in child en (5- o 11-yea -olds) using a dual- ask pa adigm wi h babble speech as backg ound noise. Howe e ,
he e we e no di e ences in lis ening e o among no mal-hea ing child en o he di e en noise le els. They
assumed, ha his was because o he high signal- o-noise a ios (SNRs) applied (
⩾
10dB)12. While Hicks and
Tha pe12 used a eac ion ask as he seconda y ask o he dual- ask pa adigm, S elmachowicz e al.18 and Choi
e al.13 es ablished a ecall o i e digi s as he seconda y ask o he dual- ask pa adigm o child en aged se en
o ou een. Bo h s udies we e conduc ed wi h wo noise scena ios: silence and speech-shaped noise a an SNR
o 8dB. Then, Howa d e al.19 used his seconda y ask in a dual- ask pa adigm o es child en’s lis ening e o
in ealis ic class oom SNRs (
⩽
4dB). They ound signi ican di e ences be ween digi ecall pe o mance in he
single ask and he dual- ask condi ion indica ing ha he pa adigm was sui able o measu e lis ening e o 19.
Howa d e al.19 ecommended u u e s udies o ex end he pa adigm o include spa ial speake dis ibu ion,
and noise conside ing e e be a ion19. The la e was al eady espec ed by Picou e al.20 who in es iga ed he
compa ison be ween a e e be an (
T30 = 834 ms
) and non- e e be an (
T30 <100 ms
) noise scena io. In
addi ion o measu ing he in luence o e e be a ion on beha iou al lis ening e o , Picou e al.20 in es iga ed
subjec i e lis ening e o , which was assessed by he child en using ques ionnai es. Howe e , hey ound no
signi ican di e ence be ween hese wo e e be a ion scena ios in he esul s o he dual- ask pa adigm o in
subjec i e lis ening e o .
Inspi ed by he indings o Howa d e al.19, his s udy aimed o assess lis ening e o wi h mo e acous ic
ealism, aking in o accoun he spa ial dis ibu ion o he speake s and an acous ically ypical Ge man class oom
scena io. The dual- ask app oach wi h digi ecall as seconda y ask was chosen o in es iga e lis ening e o due
o i s simila i y o class oom si ua ions, whe e lis ening and lea ning a e o en combined wi h o he asks, e.g.,
aking no es. Leis e al.21 showed ha child en memo ise digi s e bally, which implies ha he p ima y ask
(wo d ecogni ion) and he seconda y ask (digi ecall) d aw om he same pool o cogni i e esou ces9, hus
he digi ecall was conside ed sui able as seconda y ask. In addi ion o he dual- ask pa adigm, his s udy
applied he subjec i e measu emen app oach using a ques ionnai e on child en’s lis ening e o .
The dual- ask pa adigm assessed lis ening e o wi h a combina ion o wo app oaches, as sugges ed by
F ancis and Lo e9. Fi s , he compa ison o dual- ask and single ask pe o mance in silence was s a is ically
analysed as p oposed by Gagné e al.4. By compa ing he single seconda y ask pe o mance o he dual- ask
pe o mance, he speci ic con ibu ion o he lis ening ask o he o e all cogni i e load could be de e mined
p o iding insigh s in o he equi ed e o . Second, lis ening e o was measu ed as he compa ison o seconda y
ask pe o mance unde di e en noise condi ions22. This me hod e ealed an assessmen o how di e en
plausible acous ic class oom noise scena ios impac ed cogni i e esou ces in o de o iden i y any possible
di e ences in lis ening e o .
As ecen esea ch measu ing lis ening e o wi h dual- ask pa adigms in child en mos ly in es iga ed la ge
age g oups spanning se e al yea s (e.g., six12 o eigh yea s7), i was decided o in es iga e wo smalle age g oups
based on Ge man p ima y school classes o espec child en’s de elopmen be e . One age g oup included i s -
o second-g ade s (aged six o se en yea s), and he o he age g oup included hi d- o ou h-g ade s (aged
eigh o en yea s). Addi ionally, young adul s we e in es iga ed as an age g oup because Picou e al.7 ound
di e ences in dual- ask pe o mance be ween child en (aged nine o se en een) and young adul s.
Fou hypo heses we e o mula ed and de i ed om he exis ing li e a u e o his s udy. Fi s , i was
hypo hesised ha lis ening e o could be measu ed using he p oposed dual- ask app oach meaning ha
he esul s by Howa d e al.19 could be ep oduced wi h he ex ended pa adigm. This would be e lec ed in
pa icipan s showing wo se pe o mance in he seconda y ask (digi ecall) when pe o med in he dual- ask
compa ed o he single- ask condi ion, signalling a highe lis ening e o . Second, i was assumed ha he
pe o mance in he dual- ask condi ion (measu ed as he seconda y ask’s ER and RT, and he p ima y ask’s
RT) would be a ec ed by he di e en backg ound noise scena ios, i.e. pe o mance would be wo se in any noise
scena io compa ed o quie , a an SNR o −3dB compa ed o an SNR o 0dB, and wi h addi ional oom e ec s
compa ed o anechoic condi ions. The hi d hypo hesis was ha he e o a e in he wo d ecogni ion ask
would inc ease wi h each noise condi ion compa ed o quie , wi h an SNR o −3dB compa ed o an SNR o 0dB,
and wi h addi ional oom e ec s compa ed o anechoic condi ions. Finally, i was hypo hesised ha a signi ican
co ela ion exis s be ween he beha iou al lis ening e o assessed by he dual- ask pa adigm and he subjec i e
lis ening e o assessed by ques ionnai es.
Me hods
Pa icipan s
The lis ening expe imen was conduc ed wi h 44 p ima y school child en aged six o en (mean age 7.5 yea s,
42.86% emale) and 25 adul s aged 21 o 37 yea s (mean age 26, 28% emale). The child en we e ec ui ed
h ough coope a ion wi h p ima y schools in Aachen, Ge many, and ia e-mail no i ica ions in which he
gua dians had p e iously exp essed hei in e es in pa icipa ing in lis ening es s. Adul s we e ec ui ed
h ough he uni e si y. The inclusion c i e ia o pa icipan s we e Ge man-speaking, no mal hea ing (HL
⩽
20dB o
= [250,500,1000,2000,4000]Hz
23), and no diagnosed wi h a en ion de ici hype ac i i y
diso de (ADHD) o epilepsy. The expe imen al da a om nine child en we e excluded om he da a analysis:
one child had o cancel du ing pa icipa ion due o illness, ano he child due o echnical p oblems, wo child en
we e excluded because hey did no ha e no mal hea ing, wo child en due o su e ing om ADHD and h ee
child en had o be excluded due o no ha ing su icien Ge man knowledge (based on pa en epo ing). No
Scien i ic Repo s | (2024) 14:25200 2
| h ps://doi.o g/10.1038/s41598-024-76932-7
www.na u e.com/scien i ic epo s/
adul s needed o be excluded; howe e , 11 ou o 25 adul s pa icipa ed in he p es udy es ing he expe imen al
design. E hical app o al by he Medical E hics Commi ee a RWTH Aachen Uni e si y was ob ained p io o
he ec ui men o he pa icipan s (EK 23-242) and he s udy was pe o med in compliance wi h he Decla a ion
o Helsinki guidelines. As compensa ion o hei pa icipa ion, child en and adul s ecei ed a 10€- ouche om
a local book and oy s o e ha also o e s online dispa ch.
Expe imen ask
A dual- ask pa adigm was chosen as he expe imen ask o measu e lis ening e o . This pa adigm consis ed
o wo asks pe o med simul aneously, he p ima y ask and he seconda y ask. The pa icipan s we e old
o concen a e p ima ily on he p ima y ask. I in ol ed wo d ecogni ion using a wo d- o-pic u e ma ching
ask3,24. Each ial o his ask con ained one spoken a ge wo d o be iden i ied and ou phonologically simila
esponse choices ha we e ep esen ed by child-app op ia e pic u es. The ask consis ed o h ee aining wo d
lis s and i e expe imen wo d lis s balanced ac oss pa icipan s. The a ge wo ds we e eco ded by a ained,
na i e Ge man-speaking emale wi h a Sennheise MD 421-II-4 Dynamic S udio Mic ophone in a sound-
a enua ed boo h a a sampling a e o 44.1kHz and 16-bi esolu ion. The eco dings we e loudness-no malized
acco ding o EBU R-12825 using he audio so wa e Audaci y26.
The seconda y ask was a digi ecall ask in which adul s had o emembe se en digi s ( om one o se en)
and child en i e digi s ( om one o i e). The algo i hm ensu ed ha no digi was epea ed and no mo e han wo
consecu i e s eps, up o down, occu ed (e.g., “1-2-3-4” o “7-6-5-4” would no be allowed). The expe imen al
p ocedu e o one ial uni (see Fig.1) was as ollows: Fi s , he digi sequences o be emembe ed we e shown
one a e ano he displaying each digi o 1s. They had o be emembe ed h oughou ou wo d ecogni ion
ials. In each o hese ou wo d ecogni ion ials, he pa icipan s had o lis en o he spoken a ge wo d and
inse hei esponses by clicking on he co esponding colou ed d awing. Subsequen ly, he pa icipan s had o
espond o a mixed-numbe a ay inse ing he memo ised digi s.
Assessing subjec i e lis ening e o
A ecen li e a u e e iew27 showed ha di e en measu es o lis ening e o a e poo ly co ela ed. This
inding suppo ed he mul idimensional concep o lis ening e o meaning ha di e en measu es assess
di e en aspec s o e o on he pe cep ion pa hway27. Thus, in his s udy, lis ening e o was measu ed no
only beha io ally bu also subjec i ely using a pa icipan ques ionnai e. In hei e iew pape , F ancis and
Lo e9 summa ised ha by answe ing he ques ion o how e o ul lis ening was, pa icipan s migh also answe
wo o he ques ions such as “How do you eel?” and “How did you pe o m?”. Based on hei summa y, h ee
ques ions we e chosen o ca ego ise he e o exe ed by pa icipan s in he di e en noise condi ions in he
lis ening expe imen . The pa icipan s we e asked o answe hese h ee ques ions a e e e y expe imen block:
1. Wie geh es Ihnen/di ge ade in diesen Momen bei dem Hö e such?
Seh gu , gu , mi elmäßig, schlech , seh schlech
Engl.: How a e you eeling a his momen du ing he lis ening es ?
Ve y good, good, a e age, bad, e y bad
2. Was denken Sie/ denks du, wie gu haben Sie/ has du die Au gabe gelös ? Seh gu , gu , mi elmäßig,
schlech , seh schlech
Fig. 1. The s uc u e o one ial uni wi h he digi ecall (seconda y ask) enclosing ou wo d ecogni ion
ials (p ima y ask). Fo he wo d ecogni ion ials, he Ge man wo ds, “Fuß” (Engl. oo ) and “Haus” (Engl.
house) a e depic ed. The child en’s e sion o he digi ecall comp ising only i e digi s is shown he e.
Scien i ic Repo s | (2024) 14:25200 3
| h ps://doi.o g/10.1038/s41598-024-76932-7
www.na u e.com/scien i ic epo s/
Engl.: How well do you hink you sol ed he ask?
Ve y good, good, a e age, bad, e y bad
3. Wie ans engend wa es ü Sie/ dich, diese Au gabe zu bea bei en? Ga nich ans engend, wenig ans en-
gend, mi elmäßig ans engend, ehe ans engend, seh ans engend
Engl.: How e o ul was i o you o do his ask?
No a all e o ul, li le e o ul, mode a ely e o ul, a he e o ul, e y e o ul
The ques ions we e answe ed on a i e-poin Like scale wi h he answe op ions depic ed abo e. The answe
op ions o each ques ion we e coded wi h a numbe om ze o (e.g., “ e y good”) o ou (e.g., “ e y bad”). To
c ea e a subjec i e sco e on lis ening e o , he h ee ques ions’ sco es we e summed o each noise condi ion
and each pa icipan leading o con inuous alues be ween ze o (no e o ) and wel e ( he mos e o ).
Expe imen al p ocedu e
The lis ening expe imen wi h adul s was conduc ed in Ap il and May 2023 and he lis ening expe imen wi h
child en ook place in Oc obe and No embe 2023. The lis ening expe imen was ca ied ou in he mobile
hea ing labo a o y, a ca a an wi h an in eg a ed hea ing boo h28 (see Fig. 2). Wi h he ca a an, he lis ening
expe imen could be ca ied ou on-si e in he p ima y school.
Be o e he s a o he expe imen , in o med consen and he ques ionnai e on he pa icipan s’ social
backg ound we e collec ed om he adul pa icipan s and he child en’s gua dians espec i ely. Then, he
pa icipan s pe o med a ision es (Snellen29) and a pu e- one audiome y (250Hz - 4000Hz). Fu he mo e,
child en we e asked o name selec ed images om he wo d- o-pic u e ma ching ask3,24 o ensu e ask
unde s anding. As he expe imen was designed child-app op ia ely including gami ica ion, he pa icipan s
we e in oduced o a s o y leading hem h ough he lis ening expe imen . The s o y was abou M s. Fox, a
eache , who has a ou i e wo ds o be ecognised and likes o play memo y games oge he wi h he pa icipan s.
The expe imen al in oduc ion comp ised h ee pa s: Fi s , pa icipan s p ac ised he wo d ecogni ion
(p ima y) ask (32 ials), hen he digi ecall (seconda y) ask (9 ials) and inally bo h asks concu en ly in
he o m o he dual- ask (4 ial uni s). The i s wo in oduc ion pa s o p ac ising he wo d ecogni ion ask
and he digi ecall ask also se ed as a baseline o conduc ing he expe imen al ask as a single ask wi hou
backg ound noise. Thus, he i s ou ials o he wo d ecogni ion ask we e aken as aining and he ollowing
28 ials we e coun ed as he baseline. The same applies o digi ecall: he i s wo ials we e used o p ac ising
and he ollowing se en ials o med he baseline. Ha ing he aining and he baseline a he beginning o he
expe imen is common p ac ice in dual- ask designs o educe he du a ion o he expe imen 7,13 and i is s ill
applied in up- o-da e esea ch30. The impo ance o he p ima y ask (wo d ecogni ion) was highligh ed in he
ins uc ions and suppo ed by eedback occu ing a e each wo d.
Following he in oduc ion, he pa icipan s could ask comp ehension ques ions and hen he expe imen ’s
main pa s a ed. The expe imen consis ed o i e blocks, each con aining one o he i e noise condi ions.
Each expe imen block consis ed o se en ial uni s (7 digi ecall and 28 wo d ecogni ion asks). The noise
condi ions we e balanced ac oss pa icipan s o each pa icipan ID. Be ween each block, pa icipan s had he
oppo uni y o ake a b eak. The pa icipan s en e ed he expe imen ’s esponses on a able . The e o a es
(ERs), which ange om ze o o one, and esponse imes (RTs) in seconds we e measu ed o bo h asks. A e
each expe imen al block, pa icipan s comple ed he ques ionnai e on subjec i e lis ening e o .
Fig. 2. A child pa icipan du ing he lis ening expe imen .
Scien i ic Repo s | (2024) 14:25200 4
| h ps://doi.o g/10.1038/s41598-024-76932-7
www.na u e.com/scien i ic epo s/
Plausible i ual sound ep oduc ion
The lis ening expe imen aimed o c ea e an acous ic scena io ha was as close o eali y as possible. To achie e
his goal, he sound ep oduc ion was spa ialised and indi idualised, and he speake posi ions and backg ound
noises we e chosen as in he class oom. The sounds we e ep oduced binau ally, which co esponds o he human
way o hea ing and enables, o example, spa ially dis ibu ed sound sou ces. Fo binau al ep oduc ion, a gene ic
head- ela ed ans e unc ion o he a i icial head o he Ins i u e o Hea ing Technology and Acous ics31
was used. This gene ic head- ela ed ans e unc ion was indi idualised o espec he indi idual di e ences
be ween pa icipan s’ hea ing. The e o e, a he beginning o he expe imen , he head sizes (head wid h, dep h,
and heigh ) o he pa icipan s we e measu ed. Based on hese, in e au al- ime di e ence cues we e adjus ed32.
This is especially impo an o child en due o hei smalle head sizes. Fo headphone equalisa ion o he open
headphones used (HD 650 by Sennheise ), he headphone- ela ed ans e unc ion o each pa icipan was
measu ed a he beginning o he expe imen and espec ed o audio ep oduc ion33.
Addi ionally, wi h ega d o he speake posi ions and backg ound noise, he lis ening expe imen aimed o
c ea e a scena io ha was as close o being ealis ic as possible. I was a challenge o design a plausible class oom
noise as he e a e many di e en noise sou ces in a class oom ha in luence child en di e en ly. In his s udy, he
main noise sou ce was conside ed o be he child en hemsel es; hence, Ge man child en’s mul i- alke babble
noise (MTB) was chosen as a plausible noise signal o his lis ening expe imen . The MTB was eco ded in an
anechoic oom consis ing o ou Ge man gi ls (aged eigh o nine) eading a sho ai y ale in a compe ing
speech manne 34. Vi ual acous ic scenes we e c ea ed using MATLAB R2022a35 and he MATLAB in eg a ion
Vi ual Acous ics 2022a36 using he oom-acous ic ende e . The i ual spa ial speake posi ions we e a anged
o simula e a ypical class oom se ing o g oup wo k, wi h spoken a ge wo ds always o igina ing om he on
a
0◦
as i hey came om a eache . In addi ion, he mul i- alke babble noise was played back om each o he
ou posi ions su ounding he lis ene : on - igh (
315◦
), back- igh (
225◦
), back-le (
135◦
), and on -le
(
45◦
) c ea ing he imp ession as i o he child en we e babbling a ound he lis ene . All speake posi ions we e
a a dis ance o 2m om he lis ene . Depending on he pa icipan s, whe he hey we e child en o adul s, he
ecei e and speake heigh s we e adjus ed acco ding o hei sizes o child en (.8m) and adul s (1.2m).
In hei ou look, Howa d e al.19 mo i a ed u u e s udies o conside no only he spa ial dis ibu ion o noise
and speake s bu also he e e be an cha ac e is ics o a class oom. Bo h we e aken in o accoun in his lis ening
expe imen . Two di e en noise scena ios we e chosen: anechoic and wi h oom e ec s o a ypical class oom.
The la e was achie ed by c ea ing a i ual acous ic scene ha made he pa icipan s acous ically eel as i hey
we e si ing in a eal class oom, aking in o accoun oom e ec s such as e e be a ion. The simula ed oom wi h
he espec i e ecei e posi ion was chosen om he EduRA Da abase37 (Clus e 1) and was based on acous ic
measu emen s in eal class ooms. The simula ed class oom had a loo a ea o
9.1m×7.5m= 68.25 m2
, a
olume o
68.25 m2×3.22 m= 219.77 m3
, and a e e be a ion ime
T30 =0.63 s
(a e aged o e he oc a e
bands
=0.5kHz, =1kHz
and
=2kHz
). Mealings and Buchholz38 ound ha
T30 =0.63 s
is a ypical
e e be a ion ime o p ima y school class ooms. In he simula ed class oom, binau al oom impulse esponses
o all speake posi ions we e simula ed using RAVEN39. In addi ion o he wo noise scena ios, wo di e en
SNRs, 0dB and
−3
dB, we e chosen acco ding o he s udy by Kla e e al.2. Be o e he expe imen conduc ion,
he loudness was calib a ed. The sound p essu e le el o he wo d o be pe cei ed was ixed a 60dB. The noises
we e calib a ed o SNRs o 0dB and −3dB espec i ely. Hence, in o al, he e we e i e di e en noise condi ions:
no noise (Q), anechoic MTB wi h SNRs o 0dB (A1) and −3dB (A2) and MTB wi h oom e ec s wi h SNRs o
0dB (R1) and −3dB (R2). The abb e ia ions in b acke s a e used o p esen he esul s mo e clea ly.
Resul s
Fo he da a analysis o he expe imen and he ques ionnai e esul s, he child en we e g ouped in o wo age
g oups acco ding o hei g ades. This led o h ee age g oups: 25 adul s (mean age: 26
±3.7
yea s; 28% emale),
17 i s - o second-g ade s aged six o se en yea s (mean age: 6.47
±0.5
yea s; 58.82% emale), and 18 hi d-
o ou h-g ade s aged eigh o en yea s (mean age: 8.56
±0.7
yea s; 27.78% emale). The da a analysis was
conduc ed wi h IBM SPSS S a is ics40 and plo s we e gene a ed using he R oolbox ggplo 241. Fo all analyses,
he signi icance le el
α=0.05
was applied.
Fo each expe imen al ask, wo d ecogni ion and digi ecall, e o a es (ERs) and esponse imes (RTs) we e
e alua ed as dependen a iables. Fo he wo d ecogni ion ask, esponse ime was measu ed om he onse
o wo d playback un il esponse inse ion and he e o a e was calcula ed by di iding he numbe o inco ec
esponses pe block by he o al numbe o ials in ha block. Fo digi ecall, esponse ime was measu ed om
he momen he digi s we e isible un il he inse ion o he las digi , and he e o a e was calcula ed pe ial
uni by di iding he numbe o inco ec ly ecalled digi s by he o al numbe o digi s. Fo pa adigm alida ion,
esponse imes and e o a es o he seconda y ask we e e alua ed using wo- ac o epea ed measu es analysis
o a iance (ANOVA). Task ype (single o dual- ask) ac ed as a wi hin-subjec s a iable, and age g oup was
a be ween-subjec s a iable. To in es iga e he noise-induced di e ences be ween he asks, ou wo- ac o
epea ed-measu es ANOVAs, one o each ask (p ima y and seconda y) and one o each dependen a iable
(ER and RT) we e conduc ed. The di e en noise condi ions ac ed as wi hin-subjec s a iables and age as a
be ween-subjec s a iable. The ques ionnai e sco es we e analysed wi h a wo- ac o ANOVA wi h noise block
ac ing as he wi hin-subjec s a iable, and age ac ing as he be ween-subjec s a iable.
Fo he ANOVAs on he expe imen and he ques ionnai e esul s, no mali y was checked based on QQ
plo s, and he da a we e mainly no mally dis ibu ed. The only noise condi ion ha s ood ou was he no noise
condi ion, which was no no mally dis ibu ed. The ANOVAs we e s ill pe o med because ha analysis is obus
agains his non-no mal dis ibu ion42,43. In addi ion, Mauchly’s es o sphe ici y was conduc ed be o e each
analysis and he G eenhouse-Geisse co ec ion was applied whe e he assump ion o sphe ici y was iola ed.
Scien i ic Repo s | (2024) 14:25200 5
| h ps://doi.o g/10.1038/s41598-024-76932-7
www.na u e.com/scien i ic epo s/

Be o e he analysis, he i s ou wo d ecogni ion ials and he i s memo y ial we e il e ed ou o
each noise condi ion. This was done o a oid pa icipan adap a ion e ec s so ha he da a would be e e lec
hei pe o mance wi hou ini ial luc ua ions. Fo he in oduc ion whe e he la e ask pa was he single
ask baseline, he i s ou wo d ecogni ion ials o he p ima y ask and he i s wo memo y ials o digi
ecall we e il e ed ou . Then, he e o a e was compu ed o he wo d ecogni ion ask pe block. P ima y o
seconda y ask ials exceeding h ee imes he s anda d de ia ion pe pa icipan o esponse ime we e il e ed
ou o he esponse ime and he e o a e44. Fo he ANOVA on he ques ionnai e sco es, no il e ing was
applied because only one da a poin pe pa icipan was collec ed o each noise condi ion.
Pa adigm alida ion: compa ison be ween single and dual- ask esul s o he seconda y ask
The di e ence be ween dual- ask and single- ask ER o digi ecall showed dec easing alues wi h age in he
quie en i onmen :
ERdi =0.296
o i s - o second-g ade s,
ERdi =0.158
o hi d- o ou h-g ade s,
and
ERdi =0.037
o adul s. A wo- ac o epea ed-measu es ANOVA demons a ed signi ican main
e ec s o ask ype (F(1, 57) = 38.190,
p<0.001
,
η2
p=0
.
401
), age (F(2, 57) = 37.216,
p<0.001
,
η2
p=0
.
566
),
and he in e ac ion be ween ask ype and age (F(2, 57) = 8.324,
p<0.001
,
η2
p=0
.
226
) on he e o a e in he
digi ecall ask. The pos -hoc es o he in e ac ion o age g oup and ask ype e ealed s a is ically signi ican
compa isons o single and dual- ask o he i s - o second-g ade s (
p<0.001
,
Mdi =0.296
) and he
hi d- o ou h-g ade s (
p=0.002
,
Mdi =0.158
), whe eas he e was no signi ican di e ence o he adul s.
The compa ison o he di e ences in esponse ime be ween single and dual- ask condi ions o digi ecall in
quie showed su p isingly g ea e esponse imes o he single ask namely, 61ms o adul s, 465ms o hi d- o
ou h-g ade s and 1733ms o i s - o second-g ade s. A wo- ac o epea ed-measu es ANOVA on esponse
imes o he digi ecall ask e ealed a s a is ically signi ican e ec o ask ype (F(1, 57) = 5.616,
p=0.021
,
η2
p=0
.
090
) and g ade (F(2, 57) = 4.485,
p=0.016
,
η2
p=0
.
136
). The in e ac ion be ween ask ype and g ade
did no show signi ican di e ences (F(2, 57) = 2.390,
p=0.101
,
η2
p=0
.
077
) Bon e oni co ec ed pos -hoc
es s e ealed signi ican ly g ea e esponse imes o i s - o second-g ade s compa ed o adul s (
p=0.046
,
Mdi =2.02 s
) as well as compa ed o hi d- o ou h-g ade s (
p=0.023
,
Mdi =1.96 s
). The Bon e oni-
co ec ed pos -hoc compa ison be ween single and dual- ask was also s a is ically signi ican , wi h, on a e age,
0.76s g ea e esponse imes o he single ask (
p=.021
). Bo h compa isons, o esponse ime and e o a e,
a e isualized in Fig. 3.
Compa ison be ween noise blocks
A wo- ac o epea ed-measu es ANOVA was conduc ed o he p ima y and seconda y asks o he e o
a e and esponse ime, espec i ely, o in es iga e he noise-induced di e ences be ween he di e en noise
condi ions.
Seconda y ask—e o a es
A wo- ac o epea ed-measu es ANOVA demons a ed signi ican main e ec s o noise condi ion (F(4, 228)
= 5.075,
p<0.001
,
η2
p=0
.
082
), age (F(2, 57) = 77.918,
p<0.001
,
η2
p=0
.
732
), and he in e ac ion o noise
condi ion and age (F(8, 228) = 2.236,
p=0.026
,
η2
p=0
.
073
) o he e o a es o he digi ecall ask. Figu e
Fig. 3. Single and dual- ask compa isons o e o a e (le ) and esponse ime ( igh ) o he seconda y ask
(digi ecall) o he h ee age g oups (adul s, i s - o second-g ade s, and hi d- o ou h-g ade s). The e o
a e/ esponse ime is depic ed on he y-axis, and he ask ype is shown on he x-axis. Box plo s show he
median (middle line), and in e qua ile ange o he da a dis ibu ions.
Scien i ic Repo s | (2024) 14:25200 6
| h ps://doi.o g/10.1038/s41598-024-76932-7
www.na u e.com/scien i ic epo s/
4 shows he esul s. Pos -hoc es ing using Bon e oni co ec ion indica ed ha o hi d- o ou h-g ade s he
ER was signi ican ly highe o he noise blocks R1 (
p=0.005
) and R2 (
p<0.001
) han hey we e o he Q
condi ion. The Bon e oni-co ec ed pos -hoc compa ison o ERs be ween he noise condi ions Q and A2 was
ma ginally signi ican wi h
p=0.05
o hi d- o ou h-g ade s. Howe e , he e was no signi ican di e ence in
he ERs be ween he noise condi ions o i s - o second-g ade s (
p>0.05
) o adul s (
p>0.05
).
Seconda y ask— esponse imes a es
The wo- ac o epea ed-measu es ANOVA (G eenhouse-Geisse co ec ed) on digi ecall esponse imes did
no e eal any signi ican main e ec o he noise condi ion (F(3.047, 173.689) = 0.121,
p=0.949
,
η2
p=0
.
002
), age g oup (F(2,57) = 0.686,
p=0.508
,
η2
p=0
.
024
), o he in e ac ion be ween age g oup and noise condi ion
(F(6.094, 173.689) = 0.950,
p=0.462
,
η2
p=0
.
032
).
P ima y ask—e o a es
A wo- ac o epea ed-measu es ANOVA (G eenhouse-Geisse co ec ed) o compa e e o a es in he wo d
ecogni ion ask showed signi ican di e ences o he main e ec o age g oup (F(2, 57) = 251.931,
p<0.001
,
η2
p=0
.
898
), noise condi ion (F(3.333, 189.961) = 235.923,
p<0.001
,
η2
p=0
.
805
), and he in e ac ion o he
wo (F(6.665, 189.961) = 35.132,
p<0.001
,
η2
p=
.
552
) (see Fig.5).
Bon e oni-co ec ed pos -hoc es s o he in e ac ion o age g oup and noise condi ion we e signi ican
o all compa isons be ween blocks o i s - o second-g ade s (
p⩽0.001
) and o hi d- o ou h-g ade s
(
p⩽0.001
). Addi ionally, he compa isons be ween blocks o adul s we e signi ican o he compa ison o he
blocks Q and A2 (
p<0.001
), Q compa ed o R1 (
p<0.001
), Q compa ed o R2 (
p<0.001
), and ma ginally
signi ican o he compa ison A1 compa ed o R2 (
p=0.005
). Looking a he compa ison o i s - o second-
g ade s and hi d- o ou h-g ade s o speci ic noise blocks, he e we e signi ican di e ences be ween hese
wo age g oups o blocks A1 (
p=0.003
) and R1 (
p<0.001
).
P ima y ask— esponse imes
A wo- ac o epea ed-measu es ANOVA (G eenhouse-Geisse co ec ed) compa ing esponse imes in he
wo d ecogni ion ask showed signi ican esul s o he main e ec s o age g oup (F(2, 57) = 38.098,
p<0.001
,
η2
p=0
.
572
), noise condi ion (F(2.509,143.011) = 59.037,
p<0.001
,
η2
p=0
.
509
), and he in e ac ion o noise
condi ion and age g oup (F(5.018,143.011) = 12.887,
p<0.001
,
η2
p=0
.
311
), The Bon e oni-co ec ed pos -
hoc es s o he in e ac ion be ween noise condi ion and age g oup showed no signi ican di e ence be ween
he noise condi ions o adul s (
p>0.05
). Howe e , all compa isons o noise condi ions we e signi ican o
hi d- o ou h-g ade s (
p⩽0.019
). Addi ionally, o he g oup o i s - o second-g ade s, all compa isons o
noise condi ions we e signi ican (
p⩽0.033
) excep o he compa ison be ween A2 and R1 (
p>0.05
). The
in e ac ions a e plo ed in Fig.6.
Fig. 4. E o a e o he seconda y ask (digi ecall) o he h ee age g oups (adul s, i s - o second-g ade s,
and hi d- o ou h-g ade s) o he i e noise condi ions: no noise (Q), anechoic MTB wi h SNRs o 0dB (A1)
and −3dB (A2) and MTB wi h oom e ec s wi h SNRs o 0dB (R1) and −3dB (R2). The e o a e is depic ed
on he y-axis and he noise condi ions a e ep esen ed on he x-axis. Box plo s show he median (middle line),
and in e qua ile ange o he da a dis ibu ions.
Scien i ic Repo s | (2024) 14:25200 7
| h ps://doi.o g/10.1038/s41598-024-76932-7
www.na u e.com/scien i ic epo s/
Valida ion o he child-app op ia e ques ionnai e on subjec i e lis ening e o
The ques ionnai e on subjec i e lis ening e o consis ed o h ee i ems: ques ions on well-being, ask pe o mance
and ask e o . C onbach’s alpha45 was used o es he in e nal consis ency o he ques ionnai e. C onbach’s
alpha o he h ee i ems was
α=0.653
o adul s,
α=0.623
o i s - o second-g ade s,
α=0.626
o hi d-
o ou h-g ade s, and
α=0.607
o all age g oups combined.
Fig. 6. Response ime o he p ima y ask o he h ee age g oups (adul s, i s - o second-g ade s, and
hi d- o ou h-g ade s) o he i e noise condi ions: no noise (Q), anechoic MTB wi h SNRs o 0dB (A1) and
−3dB (A2) and MTB wi h oom e ec s wi h SNRs o 0dB (R1) and −3dB (R2). The esponse ime is depic ed
on he y-axis and he noise condi ions a e depic ed on he x-axis. Box plo s show he median (middle line), and
in e qua ile ange o he da a dis ibu ions.
Fig. 5. E o a e o he p ima y ask (wo d ecogni ion) o he h ee age g oups (adul s, i s - o second-
g ade s, and hi d- o ou h-g ade s) o he i e noise condi ions: no noise (Q), anechoic MTB wi h SNRs o
0dB (A1) and −3dB (A2) and MTB wi h oom e ec s wi h SNRs o 0dB (R1) and −3dB (R2). The e o a e
is ep esen ed on he y-axis and he noise condi ions a e depic ed on he x-axis. Box plo s show he median
(middle line), and in e qua ile ange o he da a dis ibu ions.
Scien i ic Repo s | (2024) 14:25200 8
| h ps://doi.o g/10.1038/s41598-024-76932-7
www.na u e.com/scien i ic epo s/
Subjec i e lis ening e o ANOVA
The wo- ac o ANOVA (G eenhouse-Geisse co ec ed) compa ing subjec i e lis ening e o sco es e ealed
signi ican di e ences o he main e ec o noise condi ion (F(3.426, 195.260) = 30.006,
p<0.001
,
η2
p=0
.
345
) and he in e ac ion be ween noise condi ion and age g oup (F(6.851, 195.260) = 3.041,
p=0.005
,
η2
p=0
.
096
). The main e ec o age g oup (F(2, 57) = 2.241,
p>0.05
,
η2
p=0
.
116
) was no signi ican . The esul s a e
depic ed in Fig.7.
The Bon e oni-co ec ed pos -hoc analyses o he in e ac ion be ween noise condi ion and age g oup
e ealed ha he i s - o second-g ade s had highe sco es o he subjec i e es ima ion o lis ening e o
han did he hi d- o ou h-g ade s o he Q condi ion (
p=0.011
). Howe e , i s - o second-g ade s had
signi ican ly lowe alues han adul s o he R2 condi ion (
p=0.004
). Conside ing he esul s o he pos -hoc
es s wi h Bon e oni co ec ion o he in e ac ion be ween noise condi ion and age g oup om he pe spec i e
o he age g oup, i was ound ha
• o adul s, all noise condi ions di e ed signi ican ly om he Q condi ion (
p<0.001
)
• o i s - o second-g ade s,
– Q di e ed signi ican ly om R2 (
p=0.018
)
– A2 di e ed signi ican ly om R1 (
p=0.022
)
– A1 di e ed signi ican ly om R2 (
p=0.033
)
– R1 di e ed signi ican ly om R2 (
p=0.008
)
• o hi d- o ou h-g ade s, he Q condi ion di e ed signi ican ly om all o he noise condi ions (
p⩽0.001
) and he A1 condi ion di e ed signi ican ly om he R2 condi ion (
p<0.001
).
Co ela ion be ween beha iou al and subjec i e lis ening e o
To check o any co ela ion be ween he o e all sco e on subjec i e lis ening e o and he beha iou al esul s
o he lis ening expe imen , a Spea man co ela ion
s
was es ed be ween he esul s o he ERs and he RTs
o he p ima y and seconda y ask and he co esponding o e all ques ionnai e sco e. This means ha o each
pa icipan i e da a poin s, one o each noise condi ion, we e analysed esul ing in a sample size o
n= 85
o
i s - o second-g ade s,
n= 90
o hi d- o ou h-g ade s,
n= 125
o adul s and an o e all sample size o
all g oups o
n= 300
. The co ela ion esul s a e p esen ed in Table 1.
Discussion
This s udy aimed o answe he ques ion o how e o ul lis ening is in class oom scena ios. To his ex en , ou
main hypo heses we e o mula ed. Fi s , he objec i e was o e i y i he dual- ask pa adigm could de ec di e eces
in lis ening e o by checking o signi ican di e ences be ween single and he dual- ask o he seconda y ask.
Then, i was hypo hesised ha di e en noise condi ions would a ec wo d ecogni ion pe o mance as well
Fig. 7. Ques ionnai e sco e o he h ee age g oups (adul s, i s - o second-g ade s, and hi d- o ou h-
g ade s) o he i e noise condi ions: no noise (Q), anechoic MTB wi h SNRs o 0dB (A1) and −3dB (A2)
and MTB wi h oom e ec s wi h SNRs o 0dB (R1) and −3dB (R2). The ques ionnai e sco e is depic ed on
he y-axis and he noise condi ions a e shown on he x-axis. Box plo s show he median (middle line), and
in e qua ile ange o he da a dis ibu ions.
Scien i ic Repo s | (2024) 14:25200 9
| h ps://doi.o g/10.1038/s41598-024-76932-7
www.na u e.com/scien i ic epo s/