Working memory supports rapid talker and accent accommodation: An individual differences investigation

Vol.:(0123456789)
Psychonomic Bulle in & Re iew (2026) 33:71
h ps://doi.o g/10.3758/s13423-025-02851-x
BRIEF REPORT
Wo king memo y suppo s apid alke andaccen accommoda ion:
Anindi idual di e ences in es iga ion
D ewJ.McLaughlin1,2 · G aceE.Teusche 1· MelissaM.Baese‑Be k3,4· K is inJ.VanEngen1
Recei ed: 9 Augus 2023 / Accep ed: 8 Oc obe 2025
© The Au ho (s) 2026
Abs ac
Unde mul i alke lis ening condi ions, lis ene s appea o apidly accommoda e a iabili y in speake p oduc ions. Howe e ,
e idence indica es ha his ial- o- ial accommoda ion incu s a p ocessing cos and is ampli ied by accen di e ences among
alke s. The p esen s udy in es iga ed how indi idual lis ene di e ences in wo king memo y capaci y and a en ional con ol
may p edic he deg ee o p ocessing cos s (as measu ed by esponse ime and accu acy) incu ed by ial- o- ial swi ches
in English alke s o i s language (L1) Ame ican accen and second language (L2) Manda in Chinese accen . Lis ene s
comple ed a dual- ask expe imen wi h a p ima y lis ening ask and seconda y non-linguis ic ask as well as measu es o
wo king memo y capaci y ( he Wo d Audi o y Recogni ion and Recall Measu e) and a en ional con ol ( he T ail Making
Tes and a S oop Tes ). Resul s indica ed ha , on ials wi h a alke swi ch, lis ene s wi h smalle wo king memo y capaci ies
we e slowe o espond o he seconda y ask and had poo e ecogni ion accu acy o he p ima y ask. The measu es o
a en ional con ol did no p edic swi ching cos s. One ma ginal in e ac ion eme ged be ween S oop es sco es ( ypically
used o assess selec i e a en ion and inhibi o y con ol) and he accen manipula ion, such ha lis ene s wi h be e a en ional
con ol had ma ginally be e o e all ecogni ion accu acy o L2 accen . Wo king memo y capaci y, on he o he hand, did
no p edic di e ences in ecogni ion accu acy be ween accen condi ions. We conclude ha wo king memo y plays a c i ical
ole in suppo ing apid ( ial- o- ial) alke accommoda ion. A en ional con ol may suppo p ocessing o L2 accen mo e
gene ally bu no be c i ical o apid accommoda ion.
Keywo ds Speech pe cep ion· Wo king memo y· Talke accommoda ion· Accen
Al hough speech p ocessing o en appea s e o less, p io
e idence indica es ha he unique quali ies o indi idual
alke s’ p oduc ions can complica e speech p ocessing
(Lu h a, 2024). Fo example, expe imen blocks wi h mul-
iple speake s, changing om ial- o- ial, end o inc ease
p ocessing cos s (as e idenced by poo e accu acy and
inc eased esponse imes) as compa ed wi h blocks wi h
a single speake (Choi e al., 2018; Choi & Pe achione,
2019; Heald & Nusbaum, 2014; Ma in e al., 1989; Mul-
lennix e al., 1989; howe e , see Sal zman e al.,(2021), o
some limi a ions o p io s udies ha ecycled ma e ials).
E idence om McLaughlin, Col e , and colleagues (2024b)
u he indica es ha he p ocessing cos s associa ed wi h
accommoda ing alke a ia ion can be examined ial- o-
ial; esul s o hei analyses e ealed consis en cos s o
ials in which a swi ch in alke s had occu ed, and hese
cos s we e g ea es when swi ching om a speake wi h i s
language (L1) accen o a speake wi h second language (L2)
accen . In he p esen s udy, we sough o build on p io
wo k by in es iga ing he po en ial oles o wo king memo y
and a en ional con ol in managing ial- o- ial changes in
alke and accen .
The e m wo king memo y was i s coined by Mille
e al. (1960) o dis inguish be ween he passi e holding o
in o ma ion (sho - e m memo y) e sus holding in o ma-
ion in memo y in he se ice o ongoing men al ac i i y
* D ew J. McLaughlin
dmclaughlin. esea c[email p o ec ed]
1 Depa men o Psychological & B ain Sciences, Washing on
Uni e si y inS . Louis, S .Louis, MO, USA
2 Basque Cen e onCogni ion, B ain andLanguage, Paseo
Mikele egi, 69, 20009Donos ia-SanSebas ián, Gipuzkoa,
Spain
3 Depa men o Linguis ics, Uni e si y o O egon, Eugene,
OR, USA
4 Depa men o Linguis ics, Uni e si y o Chicago, Chicago,
IL, USA
Psychonomic Bulle in & Re iew (2026) 33:71 71 Page 2 o 15
( o example, language p ocessing). Co espondingly,
wo king memo y capaci y e e s o he unc ional capabil-
i y o an indi idual’s sys em and is p ac ically es ima ed
in e ms o he numbe o uni s o in o ma ion he indi-
idual can hold in memo y (while comple ing a cogni i e
p ocess) a a gi en ime. A en ional con ol, on he o he
hand, is de ined as he indi idual’s abili y o engage in
goal-di ec ed beha io (Posne , 1982). A a ie y o clas-
sic asks, including he S oop (1935)and E iksen lanke
(E iksen & E iksen, (1974), a e commonly used o examine
indi idual di e ences in a en ional con ol abili y.
Howe e , sepa a ing wo king memo y and a en ional
con ol in o disc e e cogni i e componen s poses bo h a
heo e ical dilemma as well as me hodological di icul-
ies. I emains an ac i e opic o deba e as o whe he
wo king memo y capaci y and a en ional con ol ep e-
sen unique cogni i e componen s. Using a ac o ana-
ly ic app oach in a la ge sample o subjec s ac oss he
adul li espan, McCabe e al. (2010) ound e idence
o s ong co ela ions among se e al es s o wo king
memo y and execu i e unc ion (including a en ional
con ol). Indeed, i has been shown ha a en ion a ec s
pe o mance on wo king memo y es s, bo h in e ms o
a en ion alloca ed o main aining i ems in memo y and
in e ms o luc ua ing a en ion ac oss ials (Unswo h
& Robison,(2015). Con e sely, pa icipan s wi h la ge
wo king memo y capaci ies ha e been ound o be be e
a sus ained a en ion (Unswo h e al., (2010), E iksen
lanke (Hei z & Engle,2007), dicho ic lis ening (Conway
e al.,2001), and S oop asks (Kane e al.,2001). In la -
ing his issue, no all wo king memo y and a en ional
con ol asks a e equal in e ms o hei p edic i e capaci-
ies (D aheim e al.,2022), and co ela ions among asks
mean o index he same unde lying cons uc s, such as
he E iksen lanke and S oop ask, do no always co -
ela e (Hedge e al.,2018).
In he p esen s udy, we use an indi idual di e ences
app oach o examine whe he wo king memo y and/o
a en ional con ol p ocesses suppo speech p ocessing
du ing ansi ions be ween alke s and accen s. We p i-
ma ily in es iga e p edic ions s emming om he ac i e
con ol model (Magnuson & Nusbaum, 2007; Nusbaum
& Magnuson, 1997). This model p oposes ha he cos s
o ial- o- ial alke changes e lec he a en ional and
memo y load equi ed o achie e phone ic cons ancy
(Heald e al., 2016). Mo e speci ically, his accoun p o-
poses ha a en ional con ol is necessa y o suppo
ec ui men o a alke accommoda ion mechanism ha
can ei he :
(a) Map he speake ’s idiosync a ic p oduc ions o he lis-
ene ’s phonological space, o
(b) Ma ch he speake ’s idiosync a ic p oduc ions o
ep esen a ions in memo y (i.e., o amilia alke s;
see schema ic by Magnuson, 2018; howe e , see also
coun e e idence in Magnuson e al., 2021).
C i ically, he ac i e con ol model also posi s a key ole
o wo king memo y (Nusbaum & Mo in, 1992); when a
change in alke s is de ec ed, he speech signal is held in
wo king memo y while a mapping is compu ed (and/o a
cu en mapping is checked o e o s and e ined).
An al e na i e accoun , he audi o y s eaming amewo k,
p oposes ha he p ocessing cos s associa ed wi h speake
changes s em om dis up ion o audi o y a en ion (Shinn-
Cunningham, 2008). On his iew, p ocessing cos s a e
incu ed as a lis ene e ocuses hei a en ion om one
audi o y objec (a speake ) o ano he (Meh aei e al., 2018).
Unde his amewo k, all ansi ions be ween alke s—
ega dless o how p edic able an upcoming ansi ion
is—ough o impose a cos . Key e idence suppo ing his
accoun s ems om wo k by Kapadia and Pe achione
(2020), in which he au ho s ound ha he numbe o alke s
in a mul i alke block did no a ec he deg ee o p ocessing
cos s; a he , a la cos appea ed o be p esen o any block
wi h wo o mo e alke s. Addi ionally, e en when changes
in alke s we e comple ely p edic able (i.e., changing back
and o h be ween wo alke s consis en ly) p ocessing cos s
pe sis ed. F om hese indings, Kapadia and Pe achione
concluded ha ansi ions be ween alke s incu a ixed cos
because o he a en ion equi ed o change be ween audi o y
objec s.
C i ically, he audi o y s eaming amewo k does no
pose a ole o wo king memo y in accommoda ion o
alke changes (see also Choi e al., 2022). No e also ha
al hough he audi o y s eaming amewo k p oposes a
key ole o dis up ions o a en ion, his is no synonymous
wi h a en ional con ol. A en ional con ol e e s o an
indi idual’s abili y o engage in oli ional, goal-di ec ed
beha io (e.g., by inhibi ing i ele an in o ma ion), whe eas
he audi o y s eaming amewo k posi s non oli ional,
in olun a y cos s s emming om alke changes.
Fu he wo k sugges s ha a combina ion o hese wo
models may bes accoun o speake -swi ching cos s.
Resul s o McLaughlin, Col e e al. (2024b) suppo he
conclusion ha mul iple mechanisms may ope a e in pa allel
o suppo apid accommoda ion. Using pupillome y, a
physiological index o cogni i e p ocessing load (Bea y,
1982), he au ho s examined speech p ocessing by L1
English pa icipan s du ing ial- o- ial changes be ween
English speake s wi h L1 and L2 (Manda in Chinese)
accen s. T ials in which a change in alke occu ed elici ed
la ge pupil esponses du ing sen ence p ocessing (i.e.,
g ea e cogni i e p ocessing load) han ials in which
no change occu ed. Fu he , alke swi ches om an
Psychonomic Bulle in & Re iew (2026) 33:71 Page 3 o 15 71
L1-accen ed speake o an L2-accen ed speake (“ac oss-
accen ”) we e mo e challenging han swi ches be ween
wo L2-accen ed speake s (“wi hin-accen ”). C i ically,
he g ea e p ocessing cos s obse ed o ac oss-accen
swi ches as compa ed wi h wi hin-accen swi ches may
indica e ha wo sepa a e mechanisms suppo apid alke
accommoda ion: A mechanism ha induces a p ocessing
cos when any ype o change in alke s occu s (e.g., akin
o an audi o y s eaming mechanism), and a mechanism
ha induces an addi ional cos and is speci ically ec ui ed
o compu ing mappings o L2 accen (e.g., a alke ,
o accen , accommoda ion mechanism). While a i s
glance he audi o y s eaming amewo k (as desc ibed
by Shinn-Cunningham, 2008) seems o desc ibe he
o me mechanism, i mus be no ed ha he design o
McLaughlin, Col e and colleagues (2024b) con lic s wi h
he amewo k’s assump ion o con inuous audi o y inpu ;
in McLaughlin e al., he e we e no only pe iods o silence
be ween audi o y s imuli (i.e., be ween ials), bu also
e bal epe i ions by he pa icipan . Thus, al hough any ype
o ac oss- ial alke change did incu a cogni i e cos —as
would simila ly be expec ed om dis up ions o audi o y
s eaming— his pa icula mechanism appea s o ope a e
on a longe imescale. I emains o be de e mined whe he
his is a dis inc mechanism engaged o p ocessing alke
changes, o a mo e domain gene al p ocess (e.g., a en ional
con ol).
In he p esen s udy, we build on McLaughlin, Col e
e al. (2024b) by examining whe he indi idual di e ences
in wo king memo y capaci y and a en ional con ol a e
ela ed o he p ocessing cos s associa ed wi h ial- o- ial
changes be ween alke s o he same and di e en accen s.
O e longe (expe imen -leng h) imescales, a subs an ial
li e a u e indica es ha lis ene s can apidly accommo-
da e L2 accen , he eby imp o ing he accu acy (B adlow
& Ben , 2008) and educing he cos s (B own e al., 2020;
Cla ke & Ga e , 2004) o speech p ocessing (see e iew by
Ben & Baese-Be k, 2021). Fu he e idence also sugges s
ha wo king memo y and/o a en ional con ol may sup-
po accommoda ion o e longe imescales. Fo example,
wo king memo y is likely c i ical o empo a y s o age o
he incoming speech s eam, such ha L2-accen ed p oduc-
ions—which po en ially esul in a g ea e numbe o one-
o-many mappings and/o e o s—can be eanalyzed (Heald
& Nusbaum, 2014). A en ional con ol, on he o he hand,
may aid he lis ene in selec ion and enhancemen o he
mos c i ical aspec s o he speech signal, o in il e ing ou
unin o ma i e cues ha occu equen ly in L2 speech (e.g.,
a ypical pauses wi hin ph ases; Ami ay, 2009).
E idence linking wo king memo y and a en ional con ol
wi h L2 accen p ocessing and/o adap a ion emain mixed.
Banks e al. (2015) examined L1 lis ene ’s ecogni ion
accu acy o sen ences p oduced in a no el (a i icially
cons uc ed) accen ac oss a six-block expe imen al
session. Lis ene pe o mance imp o ed ac oss he ask a
a as e a e, and o a g ea e deg ee, o indi iduals wi h
be e a en ional con ol (as measu ed by a S oop es ).
Wo king memo y capaci y (as measu ed wi h a eading span
ask), on he o he hand, was no di ec ly ela ed o o e all
pe o mance o adap a ion. In con as , e idence om
McLaughlin e al. (2018) indica es a key ole o wo king
memo y—bu no necessa ily a en ional con ol—in L2
accen pe cep ion. McLaughlin e al. (2018) examined
ansc ip ion accu acy by L1 lis ene s o seman ically
anomalous (bu syn ac ically no mal) sen ences spoken
by an L1- and an L2-accen ed speake o English. The
L1-accen ed s imuli we e p esen ed in ei he speech-shaped
o ca e e ia noise and he L2-accen ed s imuli we e p esen ed
in ei he quie o speech-shaped noise (i.e., ou condi ions
o al), and addi ional measu es o cogni i e abili y we e
collec ed. Unlike Banks e al. (2015), he condi ions in
McLaughlin e al. (2018) we e andomly in e mixed, and,
hus, only ela ionships be ween he cogni i e measu es and
o e all ask pe o mance (no adap a ion o he accen ) we e
examined. An analysis o indi idual di e ences indica ed
ha lis ene s wi h la ge wo king memo y capaci ies
pe o med be e o e all in bo h L2 accen condi ions,
bu no he L1 accen (in noise) condi ions; hus, wo king
memo y appea ed o be c i ical o L2 accen pe cep ion.
A en ional con ol (assessed wi h a S oop es ) was also
in es iga ed, bu did no co ela e wi h pe o mance in any
condi ions.
To summa ize, he cogni i e cos s associa ed wi h
p ocessing un amilia L2 accen in gene al (as compa ed
wi h L1 accen ; McLaughlin & Van Engen, 2020), and
adap ing o un amilia L2 accen o e he cou se o an
expe imen , may be due o inc eased engagemen o
a en ional con ol (Adank & Janse, 2010; Janse & Adank,
2012) and/o inc eased demands on wo king memo y
(McLaughlin e al., 2018). Likewise, models o he cogni i e
cos s associa ed wi h apid, ial- o- ial accommoda ion o
alke changes p opose ec ui men o hese same cogni i e
esou ces. By examining indi idual lis ene di e ences
in wo king memo y capaci y and a en ional con ol, he
p esen s udy aimed o un angle hese ela ed opics.
The p esen s udy
The p esen s udy used a dual- ask pa adigm p e iously used
by S and e al. (2018) and B own e al. (2020). Dual- ask
pa adigms le e age he concep o a limi ed-capaci y sys em
o in es iga e cogni i e p ocessing load o a gi en ask. Sen-
ences spoken by h ee L1-accen ed (Ame ican) alke s and
h ee L2-accen ed (Manda in Chinese) alke s we e p esen ed
o L1 English lis ene s in a andomized o de . By examining
Psychonomic Bulle in & Re iew (2026) 33:71 71 Page 4 o 15
he con ex o he p io ial ( ial N − 1), his design c ea es
h ee condi ions: No swi ch ( he same alke on ial N and ial
N − 1), wi hin-accen swi ch (di e en alke s on ial N and
ial N − 1, bu bo h o L1 o bo h o L2 accen ), and Ac oss-
Accen Swi ch (di e en alke s o di e en accen s on ial N
and ial N − 1). We p edic ed ha (1) esponse imes would be
slowe and speech ecogni ion accu acy would be poo e o i-
als wi h L2 accen han L1 accen ; (2) esponse imes would be
slowe and speech ecogni ion accu acy would be poo e o i-
als in which a swi ch in alke s had occu ed; and (3) esponse
imes would be slowe and speech ecogni ion accu acy would
be poo e o Ac oss-Accen Swi ch ials han Wi hin-Accen
Swi ch ials (as in McLaughlin, Col e e al., 2024b). No e
ha in he con ex o he cu en s udy, he measu e o esponse
ime comes om a dual- ask pa adigm designed o index cog-
ni i e p ocessing load (as opposed o a esponse ime index o
p ocessing speed, as in s udies such as Kapadia & Pe achione,
2020). Finally, we p edic ed an asymme y in ac oss-accen
swi ching cos s ma ching he esul s o McLaughlin, Col e
e al., such ha swi ches om L1 o L2 accen would be mo e
cogni i ely demanding han swi ches om L2 o L1 accen .
Indi idual di e ences in wo king memo y capac-
i y and a en ional con ol, and hei ela ionships wi h
o e all pe o mance as well as ial- o- ial swi ching
cos s, we e also examined. We p edic ed ha he e
would be ela ionships o wo king memo y capaci y
and a en ional con ol wi h o e all pe o mance, pa -
icula ly o L2-accen ed ials (i.e., in e ac ions). Addi-
ionally, we p edic ed ha indi idual di e ences in bo h
wo king memo y capaci y and a en ional con ol would
in e ac wi h he e ec o swi ching, indica ing a ole o
bo h in managing ial- o- ial apid alke and accen
accommoda ion.
The a o emen ioned p edic ions a e in line wi h an
ac i e con ol model o alke and accen swi ching
cos s. Speci ically, bo h asymme y in ac oss-accen
swi ching cos s (as in McLaughlin, Col e e al., 2024b)
and oles o wo king memo y capaci y and/o a en ional
con ol would suppo he ac i e con ol model. Howe e ,
he p esence o hese ela ionships would no nega e
a ole o a seconda y mechanism (i.e., one akin o an
audi o y s eaming mechanism ha ope a es on a longe
imescale). Thus, ou p ima y aim in he p esen s udy
was o de e mine he in ol emen o wo king memo y
and/o a en ional con ol du ing alke swi ching, as
e idence suppo ing he ac i e con ol model.
Me hods
The cu en s udy was app o ed by Washing on Uni e si y’s
Ins i u ional Re iew Boa d.
Pa icipan s
Young adul subjec s (Mage = 19.5 yea s; ange: 18–30 yea s;
36 men, 85 women) we e ec ui ed om Washing on Uni e -
si y in S . Louis’s Psychology Pa icipan s Pool. Inclusion c i-
e ia (se ia demog aphic il e s in SONA Sys ems) selec ed
o L1 English speake s wi h no mal hea ing and ision (o
co ec ed- o-no mal ision). Addi ional c i e ia on he SONA
lis ing indica ed ha subjec s should no sign up o he s udy
i hey had ex ensi e exposu e o Manda in Chinese (e.g., hey
should no speak Manda in Chinese, ha e s udied Manda in
Chinese, o ha e pa en s o oomma es who a e luen in Man-
da in Chinese). Pa icipan s who did no comple e all asks we e
excluded om analyses. Fo he dual- ask da ase , we se an
a p io i exclusion h eshold (o g ea e han 3,000 ms mean
esponse imes) wi h he aim o sc eening ou pa icipan s who
we e no ac i ely engaged; no pa icipan s exceeded his h esh-
old o we e excluded o his eason. The inal sample size was
N = 120.
Ma e ials
Sen ences om he Seman ically No mal Sen ence Tes
(SNST; Nye & Gai enby, 1974) we e used in he p esen
expe imen . The SNST includes seman ically anomalous
i ems wi h ou keywo ds each, such as “ he w ong sho led
he a m.” Reco dings o hese sen ences we e c ea ed in a
sound- educ ion boo h using MOTU Ul aLi e-mk3 Hyb id
mic ophone ha dwa e and Audaci y (Ve sion 2.4.2) un on iMac
(Ve sion 10.15.7). Six young adul women Manda in Chinese-
accen ed speake s we e eco ded eading all o he seman ically
anomalous i ems. These eco dings we e pilo ed wi h 253
pa icipan s, o a o al o app oxima ely 10 ansc ip ions pe
i em (i.e., each pa icipan lis ened o only 90 i ems). Th ee
Manda in-accen ed speake s who we e ai ly well-ma ched on
in elligibili y we e selec ed om his pilo o use in he p esen
s udy. The selec ed speake s we e es ima ed o be 51.8%, 53.1%,
and 55.8% in elligible; he o he h ee speake s we e es ima ed o
be 30.5%, 40.0%, and 40.3% in elligible. Gi en ha some s udies
ha e shown ha pa icipan s will “gi e up” when in elligibili y
h esholds a e below 50% (Zek eld & K ame , 2014), we op ed
o use he mo e in elligible speake s o he p esen s udy.
Fo he L1-accen ed condi ion, h ee women L1 speak-
e s o English om he Midwes e n Uni ed S a es we e
eco ded. The L1 speake s we e ins uc ed o p oduce
i ems a a ypical speed, a sligh ly slowe han no mal
speed, and a slowe han no mal speed. I ems we e hen
selec ed based on hei o al du a ion in o de o ma ch
he speaking a e ac oss he L2 and L1 speake s. The inal
a e age s imuli leng h o he L2 speake s was 1,790 ms,
and he inal a e age s imuli leng h o he L1 speake s
was 1,784 ms.
Psychonomic Bulle in & Re iew (2026) 33:71 Page 5 o 15 71
P ocedu es
O e iew
Da a o he p esen s udy comes om a la ge longi udinal s udy
ha examined pe cep ual accommoda ion o L2 accen ac oss
mul iple days (see McLaughlin, Baese-Be k e al., 2024a. Only
dual- ask da a om Day 1 and measu es o wo king memo y
capaci y and a en ional con ol collec ed on subsequen days
a e examined in he p esen s udy. Full de ails o he 5-day p o-
ocol can be ound in McLaughlin, Baese-Be k e al. (2024a);
we epo only he ele an p ocedu es he e.
The speech pe cep ion ask was adminis e ed on a 21.5-
in. iMac (Ve sion 10.15.7, “Ca alina”) and p og ammed wi h
Supe Lab (Ced us, Ve sion 5). Audio was p esen ed ia ci -
cumau al Beye dynamic DT 100 headphones. The measu e
o wo king memo y capaci y was also adminis e ed on his
iMac, bu was p og ammed wi h PsychoPy 2 (Pei ce e al.,
2019), and he measu es o a en ional con ol we e bo h
adminis e ed on pape .
Dual‑ ask pa adigm
The dual- ask design, shown in Fig.1, used in he p esen
s udy was modeled a e S and e al. (2018) and B own
e al. (2020). The p ima y ask in he dual- ask pa adigm
examined ecogni ion accu acy o seman ically anoma-
lous sen ences and he seconda y ask examined nonlin-
guis ic isual ca ego iza ion. Pa icipan s we e old ha
hey would be comple ing bo h asks simul aneously, bu
o p io i ize he p ima y (speech) ask o e he seconda y
ask. Each ial, subjec s we e p esen ed wi h a single sen-
ence o e he headphones. Thei goal was o epea back
he sen ence a he end o he ial as accu a ely as pos-
sible. Simul aneously, wo emp y squa es appea ed on he
sc een (a he onse o he audio). A e an in e s imulus
in e al (ISI) o 600, 700, o 800 ms, a numbe be ween 1
and 8 appea ed in ei he he le o he igh box. Using a
bu on box, pa icipan s we e ins uc ed o make ei he a
le esponse o a igh esponse depending on he ollow-
ing: I an odd numbe appea ed (1, 3, 5, o 7), hey we e
supposed o p ess he bu on on he opposi e side as he
box on he sc een; i , howe e , an e en numbe appea ed
(2, 4, 6, o 8) hey we e supposed o p ess he bu on on
he same side as he box on he sc een1. Fo example,
he co ec esponse o a 1 appea ing in he le box on
Fig. 1 The imings o he p ima y (speech pe cep ion) and seconda y
( isual ca ego iza ion) asks a e shown in pa allel. All ials began
wi h simul aneous p esen a ion o wo emp y boxes onsc een and a
a ge sen ence. A e a pe iod o 600, 700, o 800 ms, he a ge num-
be o he seconda y ask would appea . Sen ences a ied in leng h,
bu his iming app oxima ely p esen s he a ge numbe midway
h ough he sen ence. Pa icipan s a e ins uc ed o espond ia key-
p ess on a bu on box as quickly as possible (i.e., hey a e ins uc ed
no o wai o he sen ence o end). Keyp esses au oma ically clea
he sc een. Once he pa icipan s a e on he clea sc een and ha e
hea d he en i e sen ence, hey can begin o epea he sen ence aloud.
Ve bal esponses a e collec ed ia mic ophone du ing he blank
sc een pe iod. The sc een s ayed blank o a pe iod o 5,000, 5,500,
o 6,000 ms and hen con inued o he nex ial au oma ically
1 Explo a o y analyses con i med ha he numbe ype (odd e sus
e en) and i s in e ac ion wi h ou manipula ion we e nonsigni ican .
We do no examine his e ec u he .

Psychonomic Bulle in & Re iew (2026) 33:71 71 Page 6 o 15
he sc een was p essing he igh bu on, and he co ec
esponse o a 2 appea ing in he le box on he sc een
was p essing he le bu on. Pa icipan s we e ins uc ed
o p io i ize accu acy, bu also espond as quickly as pos-
sible. Only ials wi h accu a e esponses o he second-
a y ask we e e ained in he da ase . This esul ed in he
emo al o 4.6% o ials in he L1 accen condi ion, and
5.7% o ials in he L2 accen condi ion.
Fo he p ima y ask, pa icipan s epea ed he a ge sen-
ence aloud a e i inished playing and hei key p ess was
made o he seconda y ask. Be ween ials, an ISI o 5,000,
5,500, o 6,000 ms occu ed.
No ably, he iming o he p esen a ion o he numbe o
he seconda y ask ensu ed ha he demands o ha ask
occu ed app oxima ely midway h ough he p esen a ion
o he a ge sen ences o he p ima y ask. As such, he
design o his dual- ask pa adigm ensu es ha he cogni i e
esou ces equi ed o comple e each ask a e in compe i ion.
C i ically, his means ha ials in which he demands o he
p ima y ask we e g ea e should esul in longe esponse
imes o he seconda y ask.
The combina ion o i ems o he p ima y and seconda y
asks was andomized ac oss pa icipan s. Fo he seconda y
ask, he p esen a ion o each numbe a each o he wo
loca ions was se o occu a andom. Fo he p ima y ask,
audi o y iles we e p esen ed in a andom o de wi hin a lis
used o p ac ice ials (12 o al) and a lis used o egula
ials (78 o al). An equal numbe o ials o each accen
condi ion and each speake we e included. Fo he egula
ials, his esul ed in 39 ials pe accen , and 13 ials pe
speake . Coun e balancing ensu ed ha a ge sen ences
we e p esen ed in bo h accen condi ions ac oss subjec s.
Du ing he p ac ice ials, a esea che emained in he
oom o obse e he pa icipan and con i m hey we e
making esponses in he co ec o de (i.e., bu on p ess and
hen e bal epe i ion). Feedback on iming o esponses
was included in he p ac ice session only. I subjec s ook
longe han 3,000 ms o espond wi h a bu on p ess a e
p esen a ion o he numbe a ge , “Too slow!” appea ed
onsc een. Da a om p ac ice ials was excluded om
analyses.
Wo king memo y capaci y
The Wo d Audi o y Recogni ion and Recall Measu e
(WARRM) was used o assess wo king memo y capaci y
(Smi h e al., 2016). This ask consis ed o i e blocks, each
con aining up o i e ials. The di icul y (i.e., numbe o
i ems) inc eased each block, beginning wi h se s o wo
wo ds in he i s block up o a se o six wo ds in he las
block. Each ial, pa icipan s lis ened o a wo d and we e
ins uc ed o epea i aloud. Immedia ely a e , hey judged
whe he he i s le e o he wo d belonged o ei he he
i s o second hal o he alphabe (spoken aloud as “ i s ”
o “second”). A e all he wo ds in a gi en ial had been
hea d, he pa icipan had o ecall each wo d in he o de in
which hey hea d hem. I he pa icipan s could accu a ely
ecall e e y wo d in he ial, ega dless o o de o accu acy
o he i s /second judgmen , hey passed he ial. When
hey had passed h ee o he i e ials, hey passed he block
and could mo e on o he nex (mo e di icul ) block. I hey
did no pass a leas h ee o he i e ials, howe e , he
ask ended. A pa icipan ’s sco e (i.e., es ima ed wo king
memo y capaci y) was he highes le el hey passed, plus
any pa ial c edi om a highe le el. Fo example, i a
pa icipan passed le el i e, and success ully comple ed
wo o he h ee ials necessa y o pass le el six, hei sco e
on he es would be 5.66 (5 + 2/3; see Smi h e al., 2016).
Gi en he obus p edic i e capaci y o WARRM (Smi h
e al., 2016) as well as ime cons ain s o he s udy, we did
no op o mul iple measu es o wo king memo y.
A en ional con ol
Two sepa a e asks we e used o assess indi idual di e ences
in a en ional con ol: he T ail Making Tes (Bowie &
Ha ey, 2006) and a S oop es (Jensen & Rohwe , 1966).
We op ed o include wo a en ional con ol asks gi en
he a iable p edic i e capaci ies o a en ion measu es in
p io wo k (D aheim e al., 2022). Fo he T ail Making
Tes , pa icipan s comple ed wo blocks. In Pa A, an 8.5
× 11 p in ed page o numbe s om 1 o 25, dis ibu ed
pseudo andomly, was p esen ed. Pa icipan s we e ins uc ed
o d aw connec ing lines be ween he numbe s in o de .
The ime i ook o comple e he ask was eco ded wi h a
s opwa ch by he esea che . In Pa B o he T ail Making
Tes , pa icipan s had o d aw connec ing lines be ween
pseudo andomly dis ibu ed numbe s and le e s in he
co ec inc easing o de , al e na ing be ween he numbe s
(1 o 13) and he le e s (A o L). The ime i ook o comple e
Pa B se ed as he es ima e o a en ional con ol.
A pape e sion o he S oop es was adminis e ed.
Pa icipan s comple ed h ee sec ions o he es : Wo ds,
Colo s, and Colo -Wo ds. Fo each sec ion o he S oop
es , he pa icipan had 45 s o ead aloud as many i ems
(i.e., wo ds in he Wo ds sec ion, colo s in he Colo s and
Colo -Wo ds sec ions) as possible. A he end o he allo ed
ime, he numbe o wo ds he pa icipan ead was eco ded.
In he Wo ds po ion o he ask, he wo ds we e p in ed
in black and whi e ink. Fo he Colo s po ion o he ask,
he le e s “XXX” we e p in ed in colo ink, p esen ing no
con lic be ween ink colo and ex . In he Colo -Wo ds po -
ion o he ask, he ink colo and he w i en colo we e
misma ched, e.g., “ ed” was p in ed in g een ink, and he
pa icipan was ins uc ed o igno e he p in ed wo d and
ead aloud he colo o he ink. A single S oop in e e ence
Psychonomic Bulle in & Re iew (2026) 33:71 Page 7 o 15 71
sco e pe subjec was calcula ed om pe o mance on hese
h ee sec ions o he es using he o mula:
whe e CW is he Colo -Wo ds sco e, W is he Wo ds
sco e, and C is he Colo s sco e (Sca pina & Tagini, 2017).
This o mula cap u es he numbe o Colo -Wo ds i ems
comple ed while con olling o an indi idual’s speed in he
Wo ds-only and Colo s-only sec ions. Highe sco es om
his me ic indica e be e a en ional con ol.
Analysis
Da a p epa a ion
Th oughou he Analysis and Resul s sec ions, esponse ime
da a e e o hose aken om he seconda y ask, whe eas
ecogni ion accu acy da a e e o hose aken om he
p ima y ask; accu acy o he seconda y ask is no analyzed.
Ou lie s in he esponse- ime da ase we e de ined using
Mean Absolu e De ia ions (MAD). Da a was g ouped by
subjec and condi ion o calcula e MAD alues, and hen
any ials wi h a esponse ime ou side o 3 MAD o he
median we e excluded; a o al o 3.48% o ials we e
emo ed du ing his p ocess. T ials in which an inco ec
esponse o he seconda y ( esponse ime) ask was made
we e also excluded (1.74% o ials). Finally, alues we e
log- ans o med.
The same p ocess was comple ed o he ecogni ion
accu acy da ase , bu no ials we e lagged o emo al.
Assessing independence o memo y anda en ional
con ol measu es
We began by checking o co ela ions among he h ee
measu es o memo y and a en ional con ol (Table1).
Ou a p io i plan was o c ea e an a en ional con ol
composi e om he T ail Making Tes sco es and S oop
sco es. Howe e , pa icipan sco es on he wo asks we e
no signi ican ly co ela ed ( = −.04, = −0.44, d = 118,
p = .66). When compa ing he WARRM sco es agains he
CW
−
[
(W×C)∕(W+C)
]
wo a en ional con ol measu es, a signi ican co ela ion
eme ged wi h he T ail Making Tes sco es ( = −.34, =
−3.87, d = 118, p < .001) bu no he S oop sco es ( = .18,
= 2.02, d = 118, p < .052; no e ha a e Bon e oni-co -
ec ion his end is nonsigni ican ). The signi ican nega i e
co ela ion be ween WARRM sco es and T ail Making Tes
sco es indica es ha lis ene s wi h g ea e wo king memo y
capaci y also had g ea e a en ional con ol.
Nex , we examined he a iance in la ion ac o s (VIFs)
wi h he “ i ” unc ion om he ca package (Fox e al.,
2007) in R (Ve sion 4.0.4; R Co e Team, 2021). VIFs can
p o e in o ma i e o de e mining whe he co ela ions
among a iables in a model may be p oducing issues o
mul icollinea i y. As shown in Table2, he VIFs o he
indi idual di e ence measu es did no exceed 1.17; hese
VIF alues a e well-below ypical h eshold alues ha
would indica e mul icollinea i y issues (e.g., 5 o 10; C aney
& Su les, 2002).
Wi h hese co ela ion alues and VIFs in mind, o
he p esen s udy we op ed o keep he h ee measu es o
indi idual di e ences sepa a e. Al hough he e is some
e idence o sha ed a iance be ween he WARRM and
T ail Making Tes , his ela ionship does no pose an issue o
mul icollinea i y. Addi ionally, by keeping hese p edic o s
independen in he models we we e able o de e mine
whe he hey cap u e unique sou ces o a iance ha may
di e en ially p edic alke and accen swi ching cos s.
C ea ion o swi ching condi ions
By examining he con ex o alke and accen on he
p e ious ial ( ial N − 1), he design c ea es h ee
condi ions: no swi ch ( he same alke on ial N and ial
N − 1), wi hin-accen swi ch (di e en alke s on ial N
and ial N − 1, bu bo h o L1 o bo h o L2 accen ), and
ac oss-accen swi ch (di e en alke s o di e en accen s
Table 1 Co ela ion ma ix o wo king memo y and a en ional con-
ol measu es
No e. A e Bon e oni mul iple-compa isons co ec ion, he alpha
le el o de e mining signi icance is p = .017. WARRM: Wo d
Audi o y Recogni ion and Recall Measu e
T ail Making S oop
WARRM −0.34* 0.18
T ail Making −0.04
Table 2 Va iance in la ion ac o s (VIFs)
No e. WARRM: Wo d Audi o y Recogni ion and Recall Measu e
Fixed E ec VIF
Accen 1.00
Swi ch 1.00
WARRM 1.17
T ail Making 1.13
S oop 1.04
2 Exac p- alues a e epo ed o he second decimal place h oughou
he Analysis and Resul s unless below p < .001. Cases o “p < .05”
deno e aw alues ha would ound up o .05, bu a e in ac below
he signi icance h eshold (such as .049).
Psychonomic Bulle in & Re iew (2026) 33:71 71 Page 8 o 15
on ial N and ial N − 1). No ably, because he six alke s
we e p esen ed in ue andom o de , he p opo ion o ials
co esponding o each o hese condi ions is no equal; no
swi ch ials accoun o 15% o he da a, wi hin-accen
swi ch o 35%, and ac oss-accen swi ch o 50%. The
di e ences in equency o each condi ion is a limi a ion
ha mus be conside ed when in e p e ing he da a.
Modeling p ocedu es
Fo he esponse- ime da a, linea mixed-e ec s eg ession
was implemen ed wi h he “lme ()” unc ion om he lme4
package (Ba es e al.,2015) in R (Ve sion 4.0.4; R Co e
Team, 2021). Random e ec s included andom in e cep s
by subjec and by i em, and andom slopes o accen by sub-
jec ; andom slopes o Swi ch we e a emp ed bu p e en ed
model con e gence and e en ually emo ed. Fo he ecog-
ni ion accu acy da a, he “glme ()” unc ion om he lme4
package was used, speci ying a logi link unc ion. Recogni-
ion accu acy was ea ed as a g ouped binomial, meaning
ha models p edic ed pe o mance using wo columns o
da a (numbe o co ec wo ds, numbe o inco ec /missed
wo ds); his main ains he s uc u e o he da a such ha
each ow o he da ase co esponds o a sen ence wi h ou
key wo ds while also accu a ely ep esen ing he basis o
he da a as binomial. Random e ec s included andom in e -
cep s by subjec and by i em; andom slopes o accen and
swi ch we e a emp ed bu p e en ed model con e gence and
e en ually emo ed. Likelihood a io es s we e conduc ed
o de e mine he signi icance o e ec s o in e es , and p al-
ues o model pa ame e s we e es ima ed using he lme Tes
package (Kuzne so a e al., 2017).
Fixed e ec s in he models included: Accen (dummy-
coded as: L1 Accen [ e e ence le el], L2 Accen ), Swi ch
(dummy-coded as: No Swi ch [ e e ence le el], Wi hin-
Accen Swi ch, Ac oss-Accen Swi ch), WARRM, T ail
Making Tes , S oop Tes , and he in e ac ions o Accen
and Swi ch independen ly and join ly wi h WARRM, T ail
Making Tes , and S oop Tes . The e ec s o WARRM, he
T ail Making Tes , and he S oop Tes we e all ea ed as
nume ic p edic o s; all we e cen e ed and scaled wi h he
scale unc ion in R ( o acili a e model con e gence).
Resul s
Response ime
Log-likelihood model compa isons a e epo ed in Table3,
and a summa y o he maximal model is epo ed in Appen-
dix A. The e ec o Accen indica ed ha lis ene s we e
slowe o espond o he seconda y ask o L2 Accen ials
han he L1 Accen ials, as p edic ed (ß = 0.03; χ2 = 35.93,
d = 1, p < .001). The e ec o Swi ch did no imp o e
model i (χ2 = 0.29, d = 2, p = .86), al hough i did in e ac
signi ican ly wi h he e ec o Accen (χ2 = 48.53, d = 2,
p < .001). As shown in Fig.2, Ac oss-Accen Swi ch i-
als we e mo e challenging han No Swi ch ials in he L1
Accen condi ion, bu no he L2 Accen condi ion (in e -
ac ion es ima e o ß = −0.04, p < .001). Response imes
o Wi hin-Accen Swi ch ials we e simila in bo h accen
condi ions (ß < 0.01, p = .68).
Bo h he measu e o wo king memo y capaci y
(WARRM) and he measu e o a en ional con ol ob ained
om he T ail Making Tes we e ela ed o o e all pe o -
mance on he ask (p = .01 and p = .03, espec i ely). Fig-
u e3 shows he ela ionship be ween each o he h ee cogni-
i e measu es and o e all ask pe o mance. Model es ima es
indica ed he lis ene s wi h la ge wo king memo y capaci y
and be e a en ional con ol (as measu ed by he T ail Mak-
ing Tes ) had as e esponse imes. S oop es sco es, on he
o he hand, did no imp o e model i (χ2 = 1.05, d = 1, p
= .31). O hese measu es, only wo king memo y capaci y
showed a ela ionship wi h he e ec o Swi ch (o he p al-
ues > .05). Al hough he o e all in e ac ion was no signi i-
can (likelihood a io es : χ2 = 3.92, d = 2, p = .14), model
es ima es indica ed an e ec o wo king memo y capaci y on
he di e ence o Wi hin-Accen e sus No Swi ch condi ions
(ß = −0.01, p < .05); hus, pa icipan s wi h la ge wo king
memo y capaci ies had somewha smalle swi ching cos s
in some cases (see Appendix TableA2). The in e ac ion
be ween wo king memo y and Swi ch is isualized in Fig.4;
Table 3 Log-likelihood model compa isons o esponse ime analy-
sis
No e. Colons indica e in e ac ions. Le els o Accen included L1
Accen ( e e ence) and L2 Accen . Le els o Swi ch included No
Swi ch ( e e ence), Wi hin-Accen Swi ch, and Ac oss-Accen
Swi ch. WARRM: Wo d Audi o y Recogni ion and Recall Measu e
E ec χ2d p
Accen 35.93 1 < .001
Swi ch 0.29 2 .86
WARRM 6.09 1 .01
T ail Making 5.07 1 .03
S oop 1.05 1 .31
Accen : Swi ch 48.53 2 <.001
Accen : WARRM 1.17 1 .28
Accen : T ail Making 0.85 1 .36
Accen : S oop <0.01 1 .98
Swi ch : WARRM 3.92 2 .14
Swi ch : T ail Making 0.55 2 .76
Swi ch : S oop 0.09 2 .96
Accen : Swi ch : WARRM 3.46 2 .18
Accen : Swi ch : T ail Making 0.57 2 .75
Accen : Swi ch : S oop 2.18 2 .34
Psychonomic Bulle in & Re iew (2026) 33:71 Page 9 o 15 71
Fig. 2 Response- ime da a o he in e ac ion o Accen and Swi ch is p esen ed wi h ba s and s anda d e o s. Y-axis has been unca ed om
7.2 o 7.4 log-scaled uni s o p esen a ion pu poses. (Colo igu e online)
Fig. 3 Rela ionships o he h ee z-sco ed cogni i e measu es o
wo king memo y capaci y (i.e., he Wo d Audi o y Recogni ion and
Recall Measu e; WARRM) and a en ional con ol (i.e., he T ail
Making and S oop es s) wi h he measu es o pe o mance a e p e-
sen ed wi h sca e plo s and linea model i lines wi h s anda d e o .
Dis ibu ions o indi idual subjec s’ means a e shown a op and
igh . Log-scaled mean esponse ime alues ( op plo s) come om
he seconda y (nonlinguis ic) ask, and mean p opo ion keywo ds
co ec alues (bo om plo s) come om he p ima y speech pe cep-
ion ask. Signi icance was de e mined based on log-likelihood model
compa isons; nonsigni ican ela ionships a e labeled ns