A en ion o Faces Exp essing Nega i e Emo ion a 7 Mon hs P edic s
A achmen Secu i y a 14 Mon hs
Mikko J. Pel ola and Linda Fo ssman
Uni e si y o Tampe e
Kaija Puu a
Uni e si y o Tampe e and Tampe e Uni e si y Hospi al
Ma inus H. an IJzendoo n
Leiden Uni e si y
Jukka M. Lepp€
anen
Uni e si y o Tampe e
To in es iga e po en ial in an - ela ed an eceden s cha ac e izing la e a achmen secu i y, his s udy es ed
whe he a en ion o acial exp essions, assessed wi h an eye- acking pa adigm a 7 mon hs o age (N=73),
p edic ed in an –mo he a achmen in he S ange Si ua ion P ocedu e a 14 mon hs. A en ion o ea ul aces
a 7 mon hs p edic ed a achmen secu i y, wi h a smalle a en ional bias o ea ul exp essions associa ed
wi h insecu e a achmen . A achmen diso ganiza ion in pa icula was linked o an absence o he age- ypi-
cal a en ional bias o ea . These da a p o ide he fi s e idence linking in an s’a en ional bias o nega i e
acial exp essions wi h a achmen o ma ion and sugges educed sensi i i y o acial exp essions o nega i e
emo ion as a es able ai ha could link a achmen diso ganiza ion wi h la e beha io al ou comes.
De eloping du ing he 1s yea , he a achmen
beha io al sys em acili a es di e en esponse pa -
e ns upon h ea ening si ua ions and sepa a ion
om ca egi e s (Bowlby, 1969). Compa ed o
secu ely a ached in an s who eadily seek and
manage o ge com o om he ca egi e when
eeling igh ened, insecu ely a ached in an s a e
cha ac e ized by an inhibi ion o ou wa d signs o
dis ess and appa en indi e ence owa d he ca e-
gi e (a oidance), clinginess and ange owa d he
ca egi e ( esis ance), o a lack o a cohe en beha -
io al s a egy and po en ial signs o ea o he ca e-
gi e (diso ganiza ion; Ainswo h, Bleha , Wa e s,
& Wall, 1978; Main & Solomon, 1990). An ex ensi e
amoun o esea ch has es ablished he impo ance
o sensi i e and suppo i e ca egi ing on he eme -
gence o secu e a achmen (Ainswo h e al., 1978;
Be nie , Ma e-Gagn
e, B
elange , & Whipple, 2014;
De Wol & an IJzendoo n, 1997), and p o ided
e idence o an endu ing impac o a achmen
secu i y on a ange o measu es o social adap a ion
such as ex e nalizing and in e nalizing p oblems,
social compe ence, and emo ion egula ion (Fea on,
Bake mans-K anenbu g, an IJzendoo n, Lapsley, &
Roisman, 2010; G oh, Roisman, an IJzendoo n,
Bake mans-K anenbu g, & Fea on, 2012; G oh
e al., 2014; S ou e, 2005).
Al hough he 1s yea o li e is c i ical o a ach-
men o ma ion and he ole o he ca egi ing en i-
onmen in his p ocess has been in es iga ed in
de ail, he e a e conside ably ewe p ospec i e
s udies in es iga ing how de elopmen al ajec o-
ies du ing his pe iod di e in in an s la e classi-
fied as secu ely o insecu ely a ached. S udies
add essing his ques ion ha e shown ha child
cha ac e is ics such as empe amen (Van IJzen-
doo n & Bake mans-K anenbu g, 2012) and candi-
da e gene ic a ian s (e.g., Luijk e al., 2011) may
no consis en ly p edic a achmen , bu he e is
some e idence o esponses mo e closely ela ed o
in e ac i e con ex s, such as in an s’posi i e a ec
We g a e ully acknowledge he e o s o he amilies who pa -
icipa ed in his p ojec , and Henna Salo aa a, Ma i Fadjuko ,
Mia Huolman, Susanna Kein€
anen, Ii is Ruo salainen, Milla Mie -
inen, Hen iikka Nisula, and Ma ia Ka aja o hei help in da a
collec ion. This esea ch was suppo ed by a g an om he Kone
Founda ion and a pos doc o al ellowship om he Ins i u e o
Ad anced Social Resea ch, Uni e si y o Tampe e, o Mikko J.
Pel ola; an awa d om he he Ne he lands O ganiza ion o
Scien ific Resea ch (SPINOZA p ize) o Ma inus H. an IJzen-
doo n; and a g an om he Eu opean Resea ch Council (G an
283763) o Jukka M. Lepp€
anen.
Co espondence conce ning his a icle should be add essed o
Mikko J. Pel ola, Human In o ma ion P ocessing Labo a o y,
School o Social Sciences and Humani ies, Uni e si y o Tampe e,
33014 Tampe e, Finland. Elec onic mail may be sen o mikko.
pel ola@u a.fi.
©2015 The Au ho s.
Child De elopmen published by Wiley Pe iodicals, Inc. on behal o Socie y
o Resea ch in Child De elopmen .
This is an open access a icle unde he e ms o he C ea i e Commons
A ibu ion-NonComme cial License, which pe mi s use, dis ibu ion and
ep oduc ion in any medium, p o ided he o iginal wo k is p ope ly ci ed
and is no used o comme cial pu poses.
All igh s ese ed. 0009-3920/2015/8605-0002
DOI: 10.1111/cde .12380
Child De elopmen , Sep embe /Oc obe 2015, Volume 86, Numbe 5, Pages 1321–1332
exp essions and beha io al esponses du ing he
S ill Face p ocedu e, o p edic la e a achmen
(e.g., B aunga -Rieke e al., 2014; Mesman, an
IJzendoo n, & Bake mans-K anenbu g, 2009, o a
e iew). In he p esen s udy, we con inued his
line o esea ch by es ing a hypo hesis ha in an s’
a en ion o acial emo ion exp essions, pa icula ly
o emo ions o nega i e alence, is associa ed wi h
la e a achmen secu i y.
The e a e good easons o assume ha a ach-
men o ma ion in in an s is ma ked by changes in
a en ion o acial exp essions. Models o a ach-
men and socioemo ional in o ma ion p ocessing
sugges ha ecu en a achmen - ela ed expe i-
ences shape ep esen a ional, physiological, and
beha io al esponses o emo ional in o ma ion, and
o cues signaling h ea in pa icula (Bowlby, 1969;
Cassidy, Jones, & Sha e , 2013; Dykas & Cassidy,
2011; V i
cka & Vuilleumie , 2012). Cues signaling
h ea can ange om di ec and a achmen -specific
h ea signals (such as he sepa a ion om o dis-
plays o ange by he ca egi e ) o mo e ambiguous
cues o po en ial dange (such as ea ul aces). I is
a gued ha a achmen - ela ed in o ma ion p ocess-
ing biases owa d po en ial h ea cues in he en i-
onmen a e cons an ly ope a ing (c . Bowlby, 1969;
Dykas & Cassidy, 2011), al hough expe iencing
h ea onesel may ampli y he p eexis ing esponse
pa e ns. Al hough he e is e idence om adul s
showing ha a achmen secu i y is associa ed wi h
pa e ns o a ending owa d o away om h ea -
ela ed s imuli such as ang y aces (e.g., Dewi e &
De Houwe , 2008), no s udies ha e in es iga ed he
ea ly o igins o a achmen - ela ed in o ma ion p o-
cessing biases in in ancy. The possibili y ha a ia-
ions in a en ion o acial exp essions o nega i e
emo ion a e associa ed wi h he e ol ing a ach-
men ela ionship in ea ly de elopmen is u he
sugges ed by findings showing ha indi idual a i-
a ions in a en ion o ea ul acial exp essions in 7-
mon h-old in an s a e ela ed o ma e nal s ess and
dep essi e symp oms (Fo ssman e al., 2014) and
ha a en ion o ang y aces is enhanced in school-
age child en wi h his o ies o pa en al mal ea men
(e.g., Pollak & Kis le , 2002).
Resea ch documen ing he ea ly on ogeny o he
pe cep ion o emo ional s imuli has shown ha du -
ing he second hal o he 1s yea , in an s begin o
eliably disc imina e be ween a ange o emo ional
exp essions p esen ed in di e en channels such as
he ace (Lepp€
anen & Nelson, 2009; Nelson & Dol-
gin, 1985), oice (Flom & Bah ick, 2007), and body
mo emen s (Ziebe , Kangas, Hock, & Bha , 2014).
A no able de elopmen al ea u e o emo ion pe -
cep ion a a ound 5–7 mon hs o age is a obus
a en ional bias owa d acial exp essions o ea .
This bias is ypically obse ed in p olonged engage-
men o a en ion o ea ul acial exp essions in
compa ison o happy o neu al s imuli (Nakagawa
& Sukiga a, 2012; Nelson & Dolgin, 1985; Pel ola,
Hie anen, Fo ssman, & Lepp€
anen, 2013; Pel ola,
Lepp€
anen, Palokangas, & Hie anen, 2008), aug-
men ed hea a e decele a ion (Lepp€
anen e al.,
2010; Pel ola, Lepp€
anen, & Hie anen, 2011), and
a en ion- ela ed elec oco ical b ain esponses o
ea ul aces (Lepp€
anen, Moulson, Vogel-Fa ley, &
Nelson, 2007; Nelson & de Haan, 1996; Pel ola,
Lepp€
anen, M€
aki, & Hie anen, 2009). Impo an ly,
he a en ion bias appea s o be ela i ely specific o
ea ul exp essions as simila biases ha e no been
obse ed o sad o ang y exp essions (e.g., G oss-
mann, S iano, & F iede ici, 2007; Soken & Pick,
1999) o o aces displaying no el, non ea ul
exp essions (Pel ola, Lepp€
anen, Vogel-Fa ley, Hie a-
nen, & Nelson, 2009; Pel ola e al., 2008), al hough
he ex en o which in an s assign specific emo ional
meaning (e.g., ea ) o pic u es o acial exp essions
emains deba able (Kagan & He schkowi z, 2005).
The obus ness and ela i e specifici y o in an s’
a en ional bias o ea ul acial exp essions a e ea-
sons o assume ha his phenomenon eflec s an
impo an aspec o ea ly emo ional in o ma ion
p ocessing, and a iabili y in his bias may he e-
o e p esen a pa icula ly sui able s a ing poin o
in es iga ing he ea ly an eceden s o la e a ach-
men . Howe e , as ye , he e ha e been no s udies
examining he po en ial o in e indi idual a ia ion
in a en ional biases o nega i e acial exp essions
as a ma ke o subsequen social and emo ional
de elopmen , o as a p ecu so o mo e efined
unc ional beha io al esponding o emo ional
e en s (c . Walle & Campos, 2012).
In he p esen longi udinal s udy we examined
whe he in an s’a en ional biases o acial exp es-
sions a 7 mon hs o age p edic mo he –in an
a achmen quali y a e 1 yea o age. In an s we e
es ed wi h an eye- acking measu e o a en ion o
acial exp essions (i.e., he O e lap pa adigm) a
7 mon hs o age and mo he –in an a achmen was
assessed a 14 mon hs o age wi h he S ange Si u-
a ion P ocedu e (SSP; Ainswo h e al., 1978). The
O e lap pa adigm (modified om he o iginal pa -
adigm by Aslin & Salapa ek, 1975), examines
in an s’gaze shi s om a cen al s imulus (i.e., a
neu al, happy, o ea ul exp ession, o a ace-
shaped con ol s imulus) o a high-con as s imulus
in he le o igh pe iphe y. Due o he empo ally
o e lapping p esen a ion o he wo s imuli, a gaze
1322 Pel ola, Fo ssman, Puu a, an IJzendoo n, and Lepp€
anen
shi o he pe iphe al s imulus equi es ac i e
disengagemen o a en ion om he cen al ace
s imulus (Colombo, B ez, & Cu indale, 2012),
which makes his ask sui able o p obing di e -
ences in a en ion alloca ion be ween dis inc cen-
ally p esen ed s imuli. A en ional biases can be
assessed mos eliably by calcula ing he p obabili y
o a en ion shi s om he cen ally p esen ed s im-
uli owa d he pe iphe al s imuli, and a obus bias
o ea ul exp essions has been obse ed as a highe
p obabili y o missing a en ion shi s om ea ul
e sus happy/neu al aces in independen in an
samples (Fo ssman e al., 2014; Lepp€
anen e al.,
2011; Nakagawa & Sukiga a, 2012; Pel ola e al.,
2008, 2011).
Ou p ima y aim was o es o di e ences in
a en ion o acial exp essions be ween in an s
g ouped as secu ely o insecu ely a ached (i.e.,
a oidan , esis an , and diso ganized), wi h addi-
ional explo a o y analyses aiming a a u he
cha ac e iza ion o he da a ac oss he a achmen
subg oups and wi h a con inuous measu e o
a achmen diso ganiza ion. As desc ibed abo e,
we we e pa icula ly in e es ed in whe he ea ly
a en ional esponses o nega i e acial exp essions
p edic he ac i a ion o a achmen beha io s in
la e de elopmen . Because p e ious s udies ha e
shown in an s wi hin his age ange o display
obus a en ional biases o ea ul aces, we
selec ed ea ul aces as he nega i e emo ion ca e-
go y in his s udy. The e a e no p io s udies
examining he associa ion be ween p ocessing o
acial exp essions and a achmen secu i y in
in an s, which led us o adop an explo a o y
app oach and e ain om s ong di ec ional
hypo heses. Theo e ical and empi ical wo k wi h
adul s (Dewi e & De Houwe , 2008; Dykas & Cas-
sidy, 2011) sugges ha , possibly se ing a egula-
o y unc ion o supp ess emo ional o e a ousal, a
ela i ely au oma ic endency o di e a en ion
away om h ea - ela ed cues is associa ed wi h
insecu e a achmen . In he p esen pa adigm such
endency would be obse ed as a ela i ely smalle
a en ion bias o ea ul exp essions. Al e na i ely,
he e e se could also be expec ed; ha is, secu e
a achmen could be associa ed wi h a ela i ely
smalle a en ion bias o ea ul exp essions, possi-
bly due o a dec eased need o moni o he en i-
onmen o po en ial signs o h ea in in an s
wi h a mo e es ablished expec a ion o ca egi e
a ailabili y.
Besides in an s’a en ion o acial exp essions,
ma e nal sensi i i y was assessed when he in an s
we e 7 mon hs old. Ma e nal sensi i i y was
assessed o examine whe he a en ional biases o
acial exp essions, and hei po en ial associa ions
wi h la e a achmen , a e associa ed wi h di e -
ences in he mo he s’abili y o de ec and espond
o in an signals. Apa om s udies wi h ma e nal
pe sonali y (de Haan, Belsky, Reid, Volein, & John-
son, 2004) o dep essi e symp oms (Fo ssman e al.,
2014) as a p oxy o he ca egi ing en i onmen ,
he e is a lack o esea ch on he associa ions
be ween pa en al ca egi ing sensi i i y and he p o-
cessing o emo ional s imuli in in ancy. Also, as
pa en al sensi i i y has an es ablished ole in he
de elopmen o a achmen secu i y, i is impo an
o de e mine whe he o he po en ial p edic i e ac-
o s (such as a en ional biases o acial exp essions)
a e independen om o in e ac wi h pa en al sen-
si i i y in p edic ing la e a achmen secu i y. We
expec ed secu e a achmen o be associa ed wi h
highe ma e nal sensi i i y, bu did no ad ance a
specific di ec ional hypo hesis wi h espec o he
influence o ma e nal sensi i i y on a en ion o
acial exp essions.
Me hod
Pa icipan s
The in an s we e ec ui ed om an ongoing lon-
gi udinal coho s udy (N=126; 44% emales;
Fo ssman e al., 2014; Pel ola e al., 2013) in which
eye- acking assessmen s o a en ion o acial
exp essions wi h he O e lap pa adigm ha e been
conduc ed a 5 mon hs (M=152.43 days, SD =3.64)
and 7 mon hs o age (M=213.85 days, SD =4.39).
All in an s we e heal hy, ull- e m (≥37 weeks),
and om u ban, middle-class amilies o Caucasian
e hnici y. A o al o 73 in an s om he o iginal
coho pa icipa ed in he a achmen assessmen a
14 mon hs o age (M=414.85 days, SD =20.24;
44% emales). The no able loss o in an s o he ol-
low-up a achmen assessmen s is pa ly explained
by he ac ha he a achmen assessmen was
in oduced as a supplemen al isi a e he onse o
he longi udinal s udy. Impo an ly, no s a is ical
di e ences we e ound be ween hose who did and
did no pa icipa e in he a achmen assessmen in
ma e nal sensi i i y, a en ion o acial exp essions,
o empe amen (as measu ed wi h he In an
Beha io Ques ionnai e; Ro hba , 1981), ps=.28–.96,
indica ing ha he pa icipan s included in he
cu en analyses a e ep esen a i e o he comple e
longi udinal s udy coho . All in an s wi h ≥3 sco -
able ials in each o he ou s imulus condi ions in
he O e lap pa adigm and a achmen da a we e
A achmen and A en ional Biases in In an s 1323
included in he p esen analyses. Gi en ha a ela-
i ely g ea e numbe o in an s had insu ficien
eye- acking da a in he 5-mon h assessmen
(n=26) as compa ed o he 7-mon h assessmen s
(n=11), he final analyses epo ed he e a e based
on in an s who had alid eye- acking da a om
he 7-mon h-assessmen and had pa icipa ed in he
a achmen assessmen s a 14 mon hs o age (n=62;
39% emales; see Da a S1 in he Online Suppo ing
In o ma ion o he 5-mon h da a analyses). App o al
o he p ojec was ob ained om he E hical Com-
mi ee o Tampe e Uni e si y Hospi al, and in o med
w i en consen was ob ained om he pa en o
each child.
Measu es
A en ion o Facial Exp essions
Eye- acking da a we e collec ed du ing he O e -
lap pa adigm o measu ing a en ion o acial
exp essions and ace-shaped con ol s imuli (Fig-
u e 1). The acial exp ession s imuli we e neu al,
happy, and ea ul exp essions posed by wo emale
models, and he ace-shaped con ol s imuli we e
phase-sc ambled images o bo h o he models’ aces,
p ese ing he ou e con ou o he ace. The ace and
con ol s imuli measu ed 15°and 11° e ically and
ho izon ally, espec i ely. The session s a ed wi h
he placemen o an elec ode ne o eco ding elec-
oencephalog aphy ( o hese da a, see Y iaho,
Fo ssman, Kaa iala, & Lepp€
anen, 2014). Du ing he
expe imen , he in an s sa on hei pa en ’s lap a a
60-cm iewing dis ance om a co neal- eflec ion
eye- acke moni o wi h a 58-cm diame e (Tobii
TX300; Tobii Technology, S ockholm, Sweden),
which eco ded da a on in an s’eye posi ions on he
sc een wi h a 300-Hz sampling a e. Da a collec ion
s a ed wi h he calib a ion o he eye- acke cam-
e as du ing which an audio isual anima ion was
p esen ed sequen ially a e e y co ne and he cen e
o he sc een. I he fi s calib a ion was no success-
ul (i.e., one o mo e calib a ion poin s we e missing
o we e no p ope ly calib a ed), he calib a ion was
epea ed a leas wo imes o a ain sa is ac o y cali-
b a ion o all fi e loca ions. I one o mo e calib a-
ion poin s we e missing a e mo e han 2
ecalib a ion a emp s, he final calib a ion ou come
was accep ed and he expe imen was s a ed. I
none o he loca ions we e success ully calib a ed
a e se e al ecalib a ion a emp s, he expe imen
was hen un wi hou collec ing eye- acking da a. In
he O e lap pa adigm, a fixa ion s imulus (a ed ci -
cle epea edly expanding om 0.4° o 4.3°) p eceded
each ial. Once he in an looked a he fixa ion s im-
ulus, he expe imen e p essed a key o p esen a ace
o a ace-shaped con ol s imulus on he cen e o he
sc een. The cen al s imulus was flanked a e
1,000 ms by a pe iphe al s imulus p esen ed 14°
equip obably on he le o igh o 3,000 ms. The
pe iphe al s imuli we e black-and-whi e e ically
a anged ci cles o a checke boa d pa e n, measu -
ing 15°94° isual angle. Each s imulus condi ion
(i.e., con ol, neu al, happy, ea ul) was epea ed
six imes o bo h models in a andom o de (i.e., a
o al o 48 ials), wi h he cons ain ha he same
s imulus condi ion was p esen ed no mo e han ou
imes in a ow and he pe iphe al s imulus on he
same side no mo e han h ee imes in a ow. S imu-
lus p esen a ion was con olled by E-P ime 2 so -
wa e (Psychology So wa e Tools Inc., www.ps ne .
com).
All da a p ocessing om he aw x–ygaze posi-
ion coo dina es o he pa ame e s eflec ing a en-
ion disengagemen we e comple ed au oma ically
by using gazeAnalysisLib, a lib a y o MATLAB
Figu e 1. The sequence o e en s in he O e lap pa adigm and
examples o he s imuli.
1324 Pel ola, Fo ssman, Puu a, an IJzendoo n, and Lepp€
anen
(Ma hwo ks, Na ick, MA) ou ines o o fline
analysis o aw gaze da a (Lepp€
anen, Fo ssman,
Kaa iala, Y iaho, & Wass, 2014). Fi s , a 37-sam-
ple median fil e was applied o emo ing ab up
spikes in he gaze da a (a ibu able o echnical
a i ac s). Second, da a segmen s wi h a maximum
o 200 ms o missing eye posi ion da a we e in e -
pola ed by con inuing he las eco ded x- and y-
coo dina es un il he acking came back online.
Thi d, bad ials we e excluded om u he analy-
ses. These included ials wi h in an mo emen ,
ina en i eness, o echnical p oblems in eye ack-
ing ( agmen ed acking), esul ing in mo e han
200 ms o missing eye posi ion da a du ing he ial
o he in an looking a he cen al s imulus o less
han 70% o he ime p eceding gaze disengagemen
o he end o he analysis pe iod (on a e age 21.9%
o all ials). Addi ionally, a small numbe o ials
we e excluded due o an icipa o y eye mo emen s
(i.e., a saccade commenced less han 150 ms a e
he pe iphe al s imulus onse ; 0.5% o all ials) o
compu e iming e o s (i.e., he s imuli no p e-
sen ed o he co ec du a ion; 0.15% o all ials).
No significan di e ence was obse ed in he num-
be o excluded ials be ween secu ely and inse-
cu ely a ached in an s, (68) =1.11, p=.27. O he
emaining sco able ials, he p opo ion o missing
a en ion shi s (i.e., no eye mo emen owa d he
pe iphe al s imulus du ing a 150- o 1,000-ms ime
window a e he pe iphe al s imulus onse ) was
calcula ed o each s imulus condi ion (c . Lepp€
anen
e al., 2011). In addi ion, a en ional bias sco es we e
calcula ed o each s imulus ype o pos hoc com-
pa isons. The bias sco es a e de ia ion con as s,
which eflec he ela i e weigh ing o a en ion o
pa icula ace s imuli in he con ex o o he s imuli
(e.g., missing a en ion shi s o ea minus he
mean o missing a en ion shi s o all o he ace
s imuli). To be included in he s a is ical analyses a
minimum o h ee sco able ials was equi ed o
each s imulus condi ion. On a e age, he in an s
included in he analyses had 9.2, 9.4, 9.0, and 9.4
sco able ials in he con ol, neu al, happy, and
ea ul s imulus condi ions, espec i ely, wi h no
di e ences in he numbe o sco able ials be ween
secu ely and insecu ely a ached in an s, F(1, 60) =
.48, p=.49.
Ma e nal Sensi i i y
Mo he –in an dyads we e ideo aped a hei
homes du ing a ee-play in e ac ion o 15 min
wi hin 2 weeks a e he 7-mon h labo a o y ses-
sion. Ma e nal sensi i i y was assessed om ideo-
apes using he Emo ional A ailabili y Scales
(EAS), 4 h ed. (Bi ingen, 2008) by a ce ified code
(au ho 3), unawa e o in an s’a achmen classifi-
ca ions and pe o mance in he eye- acking ask.
Ma e nal sensi i i y was coded on a scale o 1–7,
consis ing o pa en al a ec exp ession, he abili y
o de ec and app op ia ely eac o child signals
and beha io s, and communica ing in espec ul
ways. The EAS da a we e a ailable o 69 (o 73)
mo he –in an dyads. The a e age sensi i i y sco e
wi hin his sample was 5.14 (SD =1.14, skew-
ness =.20), which is simila o na ionally ep esen-
a i e ma e nal sensi i i y da a (e.g., Me as-Salmio
e al., 2013). In e code ag eemen (in aclass co e-
la ion [ICC]) be ween he main code and an inde-
penden ce ified code on a se o 20 cases was
ICC =.95.
In an –Mo he A achmen
In an s and mo he s we e obse ed in he SSP
(Ainswo h e al., 1978). In acco dance wi h s an-
da d p ocedu es, he SSP consis ed o se en 3-min
episodes, including wo sepa a ions om and wo
eunions wi h he mo he , and in e ac ion wi h a
emale s ange . In an s’a achmen beha io s we e
coded om ideo apes acco ding o he o ganized
a achmen scales by Ainswo h e al. (1978) and
he Main and Solomon (1990) coding sys em o
assessing a achmen diso ganiza ion. On he basis
o a ings on 7-poin scales assessing in an s’p ox-
imi y seeking, con ac main enance, esis ance, and
a oidance du ing he wo eunion episodes, in an s
we e fi s classified as secu e (B), insecu e-a oidan
(A), o insecu e- esis an (C). Nex , signs o a ach-
men diso ganiza ion (D) du ing each episode when
he mo he was p esen we e a ed using a scale
om 1 (no signs o diso ganiza ion) o9(s ong signs
o diso ganiza ion), wi h sco es highe han 5 ecei -
ing a diso ganized classifica ion. In an beha io s
con ibu ing o a achmen diso ganiza ion sco ing
include con adic o y beha io s (e.g., apid a oid-
ance ollowing a c y o he mo he ), s e eo ypical
o anomalous beha io s, s illing o eezing, di ec
signs o app ehension ega ding he mo he , and
misdi ec ed o diso ien ed beha io s. In e code
ag eemen was calcula ed om 18% o he sample
by wo code s (au ho s 1 and 4, bo h unawa e o
he ma e nal sensi i i y sco es o in an s’pe o -
mance in he eye- acking ask). Fo he ou -way
ABCD classifica ions, p esence e sus absence o
secu e a achmen (i.e., B s. non-B), and he p es-
ence e sus absence o a achmen diso ganiza ion
(i.e., D s. non-D), in e code ag eemen was 92%
A achmen and A en ional Biases in In an s 1325

(js=.88, .85, and .81, espec i ely). Ag eemen on
he con inuous D sco es was ICC =.91. Fo he
emaining sample, he ABCD classifica ions we e
based on a consensus be ween he fi s au ho and
expe code s om Leiden Uni e si y. In he ull
sample o 73 in an s, 46 in an s (63%) we e classi-
fied as secu ely a ached (B), 10 (14%) as insecu e-
a oidan (A), 6 (8%) as insecu e- esis an (C), and
11 (15%) as diso ganized (D). Wi hin he sample o
62 in an s ha ing su ficien eye- acking da a a
7 mon hs, he dis ibu ion was: B =37 (60%), A =8
(13%), C =6 (10%), and D =11 (17%). These
dis ibu ions we e hus highly conco dan wi h
me a-analy ic da a on a achmen dis ibu ions in
Wes e n samples (e.g., Van IJzendoo n, Schuengel,
& Bake mans-K anenbu g, 1999),
2
<.58, p=.90.
The s a is ical analyses we e conduc ed using he
secu e e sus insecu e g ouping and he con inuous
D sco es as dependen a iables.
Resul s
Fi s , an analysis o a iance o he missing a en-
ion shi da a a 7 mon hs was conduc ed o es
whe he he in an s included in he p esen sample
show a simila pa e n o a en ion o aces as has
been obse ed in p e ious esea ch. Replica ing
p e ious findings, he main e ec o ace s imulus
was significan , F(3, 183) =41.34, p<.0001, pa ial
g
2
=.40, due o he p opo ion o missing a en ion
shi s being significan ly highe o ea ul aces
(M=.52; SD =.25) han o happy (M=.36,
SD =.25) and neu al (M=.36, SD =.25) aces, o
o he con ol s imuli (M=.18, SD =.21), all Bon-
e oni-co ec ed compa isons p<.0001, d>.64.
O he p elimina y analyses showed ha missing
a en ion shi s o aces a 7 mon hs o age we e no
associa ed wi h gende o ma e nal sensi i i y
(ps=.16–.99). A achmen g ouping (secu e s.
insecu e) was no associa ed wi h ma e nal sensi i -
i y, (67) =.45, p=.66, o age (in days) a a ach-
men assessmen , (71) =.49, p=.63. The e we e
mo e males han emales wi h an insecu e classifica-
ion (20 s. 7) as compa ed o he gende dis ibu-
ion wi hin he secu e g oup (21 s. 25),
2
=5.58,
p=.018. The con inuous D sco es we e no associ-
a ed wi h ma e nal sensi i i y, =.05, p=.71; age
a a achmen assessmen , =.03, p=.79; o gen-
de , (71) =.40, p=.69.
In he main analyses, logis ic eg ession analysis
was conduc ed o p edic secu e e sus insecu e
a achmen g ouping and linea eg ession was
used o p edic a achmen diso ganiza ion ( he
con inuous D sco es) on he basis o missing a en-
ion shi sco es o aces and ace-shaped con ol
s imuli a 7 mon hs o age (wi h he p edic o a i-
ables mean cen e ed). The eg ession analyses we e
fi s un wi h gende and ma e nal sensi i i y
included as p edic o a iables, bu as no in e ac-
ions in ol ing hese a iables eme ged (i.e., gende
and ma e nal sensi i i y did no in e ac wi h miss-
ing a en ion shi sco es o p edic a achmen secu-
i y o diso ganiza ion), hey we e excluded om
he final analyses o educe he numbe o p edic-
o s and o imp o e model fi .
The logis ic eg ession model wi h a achmen
g ouping (secu e s. insecu e) as he dependen
a iable and missing a en ion shi sco es as p e-
dic o s was significan ,
2
=10.40, p=.034,
R
2
=.15. A achmen g ouping was p edic ed sig-
nifican ly by he p opo ion o missing a en ion
Figu e 2. Mean p opo ion (p) o missing a en ion shi s in di e en s imulus condi ions in 7-mon h-old in an s, g ouped by a achmen
secu i y. E o ba s ep esen he s anda d e o o mean.
*p<.05.
1326 Pel ola, Fo ssman, Puu a, an IJzendoo n, and Lepp€
anen
shi s o neu al (B=3.90, SE =1.79, p=.03) and
ea ul (B=3.39, SE =1.61, p=.035) aces (Fig-
u e 2). Sepa a e es s wi h he a en ional bias
sco es showed ha secu e a achmen was associ-
a ed wi h a la ge a en ional bias o ea ul aces,
(60) =2.59, p=.012, d=.67, while he bias sco es
o neu al aces we e la ge in insecu e han secu e
in an s, (60) =2.46, p=.017, d=.64. As an explo -
a o y analysis o es ima e how hese e ec s mani-
es in he di e en insecu e subg oups, a
mul inomial logis ic eg ession wi h he ou -way
a achmen classifica ion (A, B, C, D) as he depen-
den a iable and missing a en ion shi sco es as
p edic o s was un. The model was significan ,
2
=23.40, p=.024, R
2
=.31. Inspec ion o he
pa ame e es ima es (wi h secu e a achmen as he
e e ence ca ego y) indica ed ha a achmen a oid-
ance was associa ed wi h a ela i ely la ge bias o
neu al (p=.023) bu no ea ul (p=.79) aces,
whe eas bo h esis an and diso ganized a achmen
we e associa ed wi h a smalle bias o ea ul aces
(ps=.039 and .012, espec i ely), bu did no di e
om secu e a achmen in a en ion o neu al aces
(ps=.44 and .15, espec i ely).
The associa ion wi h a achmen diso ganiza ion
was u he es ed wi h a linea eg ession analysis
wi h he D sco e as he dependen a iable and
missing a en ion shi sco es a 7 mon hs as p edic-
o s. The model was significan , F(4, 61) =2.68,
p=.041, R
2
=.16, wi h a en ion o ea ul aces as
he only significan p edic o o he D sco es,
b=.48, p=.004. Finally, o es whe he he
a en ional bias o ea ul aces is equally s ongly
associa ed wi h diso ganiza ion and esis ance, a
pa ial co ela ion be ween ea bias sco es and
D sco es was calcula ed while con olling o
esis ance sco es (a e aged ac oss he wo eunions
du ing he SSP, wi h highe sco es on a 7-poin
scale indica ing highe esis ance), and ice e sa
(i.e., co ela ing ea bias and esis ance sco es
while con olling o D sco es). This analysis
showed ha he nega i e associa ion be ween ea
bias and a achmen diso ganiza ion ( =.35,
p=.006, i.e., highe D sco es associa ed wi h
a smalle a en ional bias o ea ul aces; see Fig-
u e 3) emained significan e en while con olling
o esis ance sco es ( =.26, p=.046), whe eas
he co ela ion be ween ea bias and esis ance
( =.28, p=.03) became nonsignifican when he
D sco es we e aken in o accoun ( =.14,
p=.28), indica ing ha he dec eased a en ional
bias o ea ul aces was mo e s ongly associa ed
wi h a achmen diso ganiza ion han esis ance. A
one-sample es wi h he subg oup classified as
diso ganized u he indica ed an absence o ea
bias in hese in an s, as he magni ude o hei ea
Figu e 3. The co ela ion ( =.35, p=.006) be ween a achmen diso ganiza ion and a en ional bias o ea ul aces a 7 mon hs o
age.
A achmen and A en ional Biases in In an s 1327
bias (M=.09) did no di e om ze o, (10) =1.59,
p=.14.
Discussion
In he p esen s udy, a en ion o acial exp essions
a 7 mon hs o age p edic ed a achmen secu i y a
14 mon hs o age: As compa ed o secu ely
a ached in an s, insecu ely a ached in an s showed
a smalle a en ional bias o ea ul aces and a la -
ge bias o neu al aces (al hough he la e e ec
appea ed o be es ic ed o he small subg oup o
a oidan ly a ached in an s). Fu he explo a o y
analyses showed ha he smalle a en ional bias o
ea ul aces was mos clea ly associa ed wi h
inc easing signs o a achmen diso ganiza ion.
Toge he , hese findings a e consis en wi h he
hypo hesis ha a ia ions in a en ion o nega i e
emo ions, and possibly o cues signaling h ea in
pa icula , a e associa ed wi h a achmen o ma ion
and may p o ide a use ul ma ke o in an s’a ach-
men s a us p io o he s anda d obse a ional
assessmen o a achmen a e he fi s bi hday.
Rega ding he po en ial mechanisms media ing his
associa ion, he esul s could be seen as consis en
wi h he model o Dykas and Cassidy (2011) p opos-
ing ha an acqui ed endency o di e a en ion
away om h ea - ela ed cues and supp ess emo-
ional o e a ousal is a ma ke o insecu e a ach-
men . I is unclea , howe e , whe he he seemingly
simila pa e n o esul s implies ha he same mech-
anisms (i.e., supp ession o a ousal) a e unc ional in
in an s and adul s, o whe he o he ac o s such as
lowe o e en blun ed sensi i i y o he emo ional
signal con eyed by ea ul aces (and po en ially
o he exp essions o nega i e alence) in insecu ely
a ached in an s a e mo e likely esponsible o he
obse ed e ec s. In e es ingly, ou supplemen a y
analyses showed ha a ia ions in a en ion bias a
5 mon hs we e no associa ed wi h la e a achmen
s a us and ha an inc ease in he a en ional bias o
ea ul exp essions om 5 o 7 mon hs was mo e
p onounced in secu ely a ached in an s han inse-
cu e and diso ganized in an s (see Da a S1).
Al hough en a i e, hese esul s aise he possibili y
ha indi idual a ia ions in he de elopmen al
eme gence o an a en ional bias o aces exp essing
nega i e emo ions (such as ea ul exp essions)
be ween 5 and 7 mon hs o age a e pa icula ly
in o ma i e o unde s anding a achmen o ma ion
(c . Lepp€
anen & Nelson, 2012).
Whe eas a obus a en ional bias o ea ul aces
was ela ed o secu e a achmen and low a ach-
men diso ganiza ion sco es, highe le els o a ach-
men diso ganiza ion we e cha ac e ized by a
diminished a en ional bias o ea ul aces in his
s udy. The absence o he age- ypical (e.g., Fo ss-
man e al., 2014; Pel ola e al., 2008) a en ional bias
o ea ul exp essions in diso ganized in an s is
highly in e es ing in ligh o da a on he associa-
ions be ween a achmen diso ganiza ion in in ancy
and la e emo ional and beha io al ou comes.
Me a-analy ic da a indica e ha while a achmen
diso ganiza ion shows only a negligible associa ion
wi h la e in e nalizing symp oms (e.g., anxie y and
dep ession; G oh e al., 2012), diso ganized a ach-
men is associa ed wi h la e ex e nalizing symp-
oms (e.g., agg essi e and opposi ional beha io ;
see Fea on e al., 2010). Face s o ex e nalizing
symp oma ology such as callous-unemo ional ai s
and an isocial beha io ha e been linked wi h spe-
cific impai men s in sensi i i y o o he indi iduals’
ea ulness a he han a gene alized impai men in
p ocessing emo ion exp essions (Dadds e al., 2006;
Ma sh & Blai , 2008), and con e sely, highe sensi-
i i y o de ec ea ul cues is ela ed o highe p o-
social endencies (Ma sh, Kozak, & Ambady, 2007).
I is in iguing o hypo hesize ha a lowe sensi i -
i y o he dis ess signals o o he s could p o ide a
ema kably ea ly-eme ging endopheno ypic ma ke
ela ing a achmen diso ganiza ion in in ancy o
la e ex e nalizing p oblems.
Wha is no ully e ealed by he cu en esul s
is he causal di ec ion o associa ions, ha is,
whe he a ia ion in a en ional biases o emo ional
s imuli is an inhe en cha ac e is ic o he in an
imposing independen e ec s on he eme gence o
a achmen secu i y, o whe he a en ional biases
a e shaped by he ca egi ing en i onmen a
7 mon hs o age and could be mo e eadily in e -
p e ed as an ea ly co ela e o he a achmen ela-
ionship a he han an independen de e minan .
Pa ly ela ed o his ques ion, Tha ne e al. (2011)
showed ha a achmen diso ganiza ion a
14 mon hs was p edic ed by ana omical a ia ion
in 6-week-old in an s’subco ical s uc u es (i.e.,
he diame e o he ganglio halamic o oid). As his
s uc u e is linked o he limbic a eas subse ing
he p ocessing o emo ional in o ma ion, such find-
ings indica e po en ially inhe en neu obiological
ac o s ha may p edispose in an s o be di e en-
ially sensi i e o emo ional s imuli. Al hough such
findings and he absence o associa ions be ween
a en ional biases and ma e nal sensi i i y in he
p esen s udy could be seen o a o he indepen-
den ole o a en i eness o emo ional signals on
de e mining la e a achmen , we hesi a e o d aw
1328 Pel ola, Fo ssman, Puu a, an IJzendoo n, and Lepp€
anen
such s ong conclusions gi en po en ial limi a ions
in ou assays o ma e nal sensi i i y. Fo example,
he associa ion be ween a achmen secu i y and
sensi i i y was no ound in his s udy al hough i
has been ex ensi ely documen ed me a-analy ically
(De Wol & an IJzendoo n, 1997) and wi h beha -
io gene ic (Fea on e al., 2006) and in e en ion
(Bake mans-K anenbu g, an IJzendoo n, & Ju e ,
2003) designs. Mul iple ac o s ha e likely con ib-
u ed o his null e ec , including modes sample
size, limi ed a iance in ma e nal sensi i i y sco es,
and he ela i ely uncons ained ee-play obse a-
ion, as i has been p oposed ha sensi i i y o
in an dis ess signals in s ess-inducing con ex s
p o ides a be e es ima e o he influence o sensi-
i i y on a achmen han nondis essing assess-
men s (Lee kes, 2011). Fu he mo e, apa om he
global assessmen o ca egi ing sensi i i y used in
he p esen s udy, he influence o mo e ocal in e -
ac i e beha io s on in an s’p ocessing o emo ional
signals has no been in es iga ed. Specific ace s o
in e ac i e beha io such as a ypical pa en al mi -
o ing o in an a ec (DeOli ei a, Bailey, Mo an, &
Pede son, 2004; Ge gely, 2004) as well as igh ened
and igh ening emo ion displays owa d he in an
(Main & Hesse, 1990; Schuengel, Bake mans-K an-
enbu g, & an IJzendoo n, 1999) ha e been ela ed
o he de elopmen o a achmen diso ganiza ion,
and sugges ed o be pa ly media ed by in an s’
abili y o ecognize hei own and o he s’emo ions
(DeOli ei a e al., 2004). A po en ially ui ul
agenda o u u e s udies is o in es iga e in mo e
de ail how such ocal aspec s o pa en ing a e asso-
cia ed wi h in an s’pe cep ual p ocessing o emo-
ion signals.
In a wide con ex , he p esen esul s showing
ha in an s’a en ional bias o ea ul aces is asso-
cia ed wi h an appa en ly posi i e de elopmen al
ou come (secu e a achmen ) a e o in e es as p e i-
ous s udies ha e linked a en ional biases o h ea -
ela ed s imuli (such as ang y and ea ul aces)
wi h exposu e o s ess (Fo ssman e al., 2014) and
p oneness o anxie y (Ba -Haim, Lamy, Pe gamin,
Bake mans-K anenbu g, & an IJzendoo n, 2007).
Fi s , he p esen esul s sugges ha he absence o
age- ypical a en ion biases, no only heigh ened
le els o his bias, may be in o ma i e as a ma ke
o ea ly childhood de elopmen . Second, he p es-
en esul s call o cau ion in in e p e ing in an s’
a en ional biases o h ea - ela ed s imuli as “ isk
ac o s,”e en hough his in e p e a ion appea s
plausible in ligh o se e al s udies in adul s linking
he bias wi h anxie y (Ba -Haim e al., 2007). In pa -
icula , he p esen esul s aise he possibili y ha
he p esence o h ea - ela ed a en ional biases in
in ancy is an impo an and po en ially ansien
(see Pel ola e al., 2013) aspec o ypical socioemo-
ional de elopmen . The esul s also call o u he
wo k o examine in an s’a en ional biases o ea -
ul aces and o he h ea - ela ed s imuli no only in
he con ex o o dina y a ia ions in pa en al
dep ession and s ess symp oms (Fo ssman e al.,
2014) bu also unde mo e se e e condi ions.
Limi a ions
In he u u e, he links be ween a achmen and
social in o ma ion p ocessing need o be in es i-
ga ed in e hnically and socioeconomically mo e
di e se samples o assess he gene alizabili y o he
e ec s ( he p esen s udy being es ic ed o a Cau-
casian, middle-class sample). Likewise, al hough he
associa ion be ween a en ional biases and la e
a achmen secu i y was no influenced by gende
in his s udy, a eplica ion o hese findings in a
la ge sample wi hou boys being o e ep esen ed
in he insecu e ca ego y is needed. A la ge sample
will also be impo an o enable mo e eliable
analyses o he pa e ns o a en ional bias wi hin
each insecu e a achmen subg oup. Finally, i is
impo an o s ess ha he p esen esul s do no
unequi ocally pe mi in e p e ing he obse ed
a en ion e ec s o be specific o ea ul exp essions
ins ead o a mo e gene al bias o h ea - ela ed o
nega i ely alenced emo ion exp essions, and a u -
he in es iga ion o he scope o he obse ed
e ec s will be impo an . Conside ing he idea ha
a achmen ep esen a ions impose a gene alized
influence on he deploymen o p ocessing
esou ces owa d po en ial h ea s in he en i on-
men (e.g., Dewi e & De Houwe , 2008; Dykas &
Cassidy, 2011), i could be expec ed ha he
obse ed e ec s eplica e wi h o he h ea - ela ed
s imuli such as exp essions o ange di ec ed
owa d he in an . We belie e, howe e , ha ea ul
exp essions a e a pa icula ly sui able class o s im-
uli o s a in es iga ing he links be ween a ach-
men and p ocessing o emo ional s imuli in
in ancy as he e is cu en ly no e idence o a simi-
la ly obus a en ional bias o o he nega i ely
alenced acial exp essions, such as ange o
sadness in in an s in he same age ange (e.g.,
G ossmann e al., 2007; Soken & Pick, 1999).
Conclusions
No wi hs anding he limi a ions o he p esen
s udy, he in es iga ion o in an s’a en ion o acial
A achmen and A en ional Biases in In an s 1329