This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
2344BRO
OKER
ETAL.
behaviorproblemsin
preschoolbutthatlevelsoflifestressw
ouldmoderate
thisrelation.Namely,w
eexpected
thatgreaterearlylife
stresswould
exacerbaterisk
forbehavior
problemslinked
toatypicalinfantangerreactivity.
Method
Participants
Thesam
plewasdraw
nfrom
alongitudinaltw
instudy
examin-
inggenetic
andenvironm
entalinfluencesonemotionaldevelop-
mentacross
infancyand
earlychildhood
(Schmidtetal.,2013).
Participantsfor
theparentstudy
were
recruitedfrom
thegreater
Madison,W
isconsin,areathrough
avariety
ofmethods,including
statebirth
records,mothers-of-tw
insclubs,
televisionpublicity,
birthannouncem
entsinnew
spapers,flyersindoctors’offices,the
Internet,andreferrals
fromparticipants.Forthe
parentstudy,989fam
ilieswere
contactedfor
recruitment,
which
was
ongoingthroughout
thestudy.O
nehundred
andforty
onefam
ilieswere
enrolledbutdid
notparticipateinany
phaseofthe
study.Infantscom
pletedlaboratory
assessments
at6months
and12months
ofage.A
tthetimeofthe
6-month
assessment,56
familieschose
nottoparticipate,3
families
hadwithdraw
nfrom
thestudy,and
495fam
ilieswere
notyetenrolled.Thusthe
6-month
sample
included294
families;566
infantsprovided
usabledata
(292girls).A
tthetimeofthe
12-month
assessment,115
families
chosenotto
par-ticipate,a
totalof17fam
ilieshad
withdraw
nfrom
thestudy,and
160fam
ilieswere
notyet
enrolled.Thus,the12-m
onthsam
pleincluded
456fam
ilies;881infantsprovided
usabledata(451girls).
Nearly
500infants
(n!481)
providedobservational
datafrom
laboratoryepisodes
atboth
6and
12months
ofage.
Smaller
numbersofinfantshad
datafrom
onlythe
6-month
(n!85)orthe
12-month
assessment(n
!400).Thesam
pleincludesthoseinfantswith
observationaldata
fromanger
episodesateither
the6-or
12-month
assessment 1
(N!966;51.1%
girls).Consistentw
ithlocaldem
ographics,mostm
othersself-reportedtheir
raceasCaucasian
(95%,African
American
!1.7%
,Asian
American
!0.9%
,American
Indian!0.4%
,Hispanic
!1.4%
,OtherRace
!0.6%
)asdid
fathers(92.8%
,African
American
!2.7%
,Asian
American
!0.9%
,American
Indian!0.3%
,His-
panic!2.1%
,Other
!1.2%
).Thelargest
number
ofmothers
(31.4%)reported
acollege
degreeastheirhighestlevelofeduca-
tion(grade
school!0.1%
,somehigh
school!1.3%
,highschool
graduate!14.5%
,somecollege,29.6%
,somegraduatetraining
!14.1%
,graduatedegree
!9.0%
).Thelargestnum
beroffathers
reportedeithersom
ecollege
(27.5%)ora
collegedegree
(27.6%)
astheir
highestlevel
ofeducation
(gradeschool
!0.3%
,some
highschool
!2.4%
,highschoolgraduate
!19.1%
,somegrad-
uatetraining
!11.5%
,graduate
degree!11.6%
).Ofthose
families
who
chosetoreport
theirannual
income,the
greatestnum
ber(43.1%
)reported
grosshousehold
incomes
ofover
$60,000(2.3%
earned$15,000
orless,
3.5%earned
$16,000–$20,000,8.4%
earned$21,000–$30,000,13.3%
earned$31,000–
40,000,13.8%
earned$41,000–$50,000,
and15.5%
earned$51,000–$60,000).O
nehundred
andone
families
didnotreport
onfam
ilycom
position:Ofthose
who
did,94.9%
reportedthat
childrenlived
withboth
biologicalparents.
ProcedureLaboratory
assessments
occurredwhen
infantswere
6(M
!5.83,SD
!0.60)and
12(M
!11.89,SD
!1.42)m
onthsofage;infants
participatedinanum
berofepisodes
designedtoelicit
discreteemotions.W
efocused
onepisodesdesigned
toelicitanger
ininfants.Follow
ingassessm
ents,mothers
andfathers
completed
apacketofquestionnairesand
mailed
themback
tothe
laboratory.Parents
alsocom
pletedmailed
questionnaireswhen
childrenwere
36months
ofage.Table1provides
anoutline
ofdatacollection,
procedures,assessments,and
measures.
BehavioralAssessm
entGentle
armrestraint.
Atboth
6and
12monthsofage,infant
angerwasassessed
duringthe
gentlearm
restraintepisodefrom
theLaboratory
Temperam
entAssessm
entBattery
(Lab–TAB;
Goldsm
ith&Rothbart,1996).Forthis,the
infantwasseated
ina
highchairand
presentedwithanoveltoy
withwhich
heorshewas
allowedtoplay
for15–30s.The
parent,havingreceived
instruc-tionsfrom
theexperim
enter,thenstood
behindthe
infantandheld
theinfant’s
forearmstohis
orher
sidetoprevent
engagement
withthe
toyfor30
s.Theinfantw
asthenallow
edtoreengage
with
thetoy
for30sbefore
asecond
restrainttrialbegan.Thesecond
trialproceededsim
ilarly,endingwiththe
infantbeingallow
edto
playwiththe
toy.Carseat.
Angerw
asalso
assessedat6
and12months
ofageduring
thecarseatLab–TA
Bepisode.Forthis,the
parentbuckledthe
infantintoastandard
butrestrictivecar
seatandthen
stoodoutside
ofthechild’sview
.Theparentw
asinstructedtoleave
theinfant
inthe
carseat
for30
sand
refrainfrom
speakingtoor
comforting
himorher.Episodes
endedwith
theparent
moving
backinto
theinfant’sdirectline
ofsightandcom
fortinghim
orherasneeded.Behavioral
codingprocedures.
Consistentwith
Lab–TAB
scoringprocedures
(Goldsm
ith&Rothbart,1996),trained
codersassigned
ratingsofanger
andsadness
foreach
infantinthe
armrestraint
andcar
seatepisodes.
Ratingswere
assignedin5-s
epochs,duringwhich
themaxim
umintensity
wasscored
forthe
following:observed
facialanger(0!noanger;3
!angerofthe
highestintensity),bodilyanger(0
!nostruggle
orresistance;4!
continuoushigh-intensitystruggle),and
distressvocalizations(0!
novocal
distress;5
!full
intensitycry/scream
).Latency
(inseconds)to
thefirstangerbehaviorw
asalso
recorded.The
following
sadnessbehaviors
were
similarly
scored:maxi-
mum
intensityoffacialsadness
(0!no
sadness;3!sadness
ofthe
highestintensity),bodilysadness
(0!absent,1
!present),
andlatency
(inseconds)
tothe
firstsadness
behaviorinthe
episode.Anger
andsadness
variableswere
codedseparately
toallow
forthe
possibilitythatinfants
mayhave
shownboth
angerand
sadnessinthe
same5-s
epoch.Allcoders
were
requiredto
achieveaminim
umreliability
("!0.70)
with
amaster
coderbefore
codingindependently.
Roughly10%
ofepisodes
were
double-codedtoestablish
reliabilityand
preventcoding
drift.Mean
intercoderreliabilityforangervariablesw
as"
!0.78.M
eanintercoderreliability
forsadnessvariables
was
"!0.75.
1Resultsare
unchangedwhen
limited
tothose
individualswithdata
atboth
6and
12months.
This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
2345INFANTANGER,LIFE
STRESS,ANDBEH
AVIORPRO
BLEMS
Latencyscores
were
reversedtoparallel
thedirectionality
ofother
behaviors;greaterscores
reflectedamore
rapiddisplay
ofangerorsadness.W
ithin-episodeaverage
compositesw
ereform
edacross
epochsfor
eachofthe
remaining
behaviors.Speedscores
andaverage
composites
were
thenz-scored
andcom
binedinto
singlemeasures 2
ofanger
(mean
r!.59)
andsadness
foreach
episode(mean
r!.45).N
otethatthis
procedureresults
intwo
composites
ofanger:(i.e.,carseatand
armrestraint)ateach
ageand
twocom
positesofsadness
ateach
age.Anger
andsadness
compositesw
eremoderately
correlatedatboth
6months(carseat:
r!.53,arm
restraint:r!.43)and
12months(carseat:r
!.50,
armrestraint:r
!.40).
QuestionnairesEarly
lifestress.
Given
ourinterestinthe
roleofnonsevere,
ormore
“typical”early
lifestressors,the
lifeeventsdom
ainofthe
ParentingStressIndex
(PSI;Loyd&Abidin,1985)w
asselectedas
ameasure
ofearlylife
stress.When
childrenwere
6months
old,the
primary
caregiverreportedwhetherthe
familyhad
experiencedeach
of20possible
stressfullifeevents
thatrangedfrom
regularhassles
(e.g.,encountering
more
stressorsthan
anticipated)to
major
lifeevents
(e.g.,death
ofamember
ofthe
immediate
family).Item
swere
weighted
bythe
severityofthe
lifestressor;
forexam
ple,endorsing
anitem
askingabout
anillness
inthe
immediate
family
(thechild
alongwith
hisorher
parentsand
siblings)receivedascore
of1,andendorsing
anitem
askingabout
divorcereceived
ascore
of7.Item
scoreswere
thensum
medto
createanoverallscale
score.Endorsingallitem
swould
resultina
scoreof80.A
saparentreportm
easure,thePSIisgenerally
more
sensitivetothe
perceptionsandmemoriesofparentsthan
tothose
ofchildren.However,by
focusingonthe
lifeevents
domain,w
ecaptured
objective,discretelife
eventsthataffectallmembersofa
family(e.g.,children
move
when
parentsmove).Thus,this
scaleisafam
ily-levelmeasure
ofcumulative
lifestress.
Behaviorproblem
s.Behaviorproblem
swere
assessedatage
36months
viathe
ChildBehavior
Checklist(Achenbach
&Re-
scorla,2000).Theprim
arycaregiverreported
thedegree
towhich
variousstatem
ents(e.g.,“show
spanic
fornogood
reason”)were
trueoftheirchild’sbehaviorona3-pointscale(0
!notatalltrue;
2!
verytrue
oroften
true).Ascale
scoreofinternalizing
problems(nitem
s!32;
#!.82)and
ascale
scoreofexternal-
izingproblem
s(nitem
s!34;#
!.90)w
asformedforeach
child.Given
thatscoresfor
internalizingand
externalizingscales
were
highlycorrelated
(r!.58)and
thatinternalizingand
externalizingbehaviors
arehighly
correlatedearly
inlife
(Gilliom
&Shaw
,
2004),theTotalProblemsscalescorew
asused(nitem
s!98,#
!.94).The
TotalProblemsscale
includedallitem
sfrom
theinter-
nalizingand
externalizingscales
alongwithitem
sfrom
theother
problemsscale
(nitem
s!32,
#!.79).The
mean
correlationamong
thethree
scaleswasr
!.68.
Missing
data.Because
recruitment
wasongoing
throughoutthe
durationofthe
study(i.e.,
someparticipants
joinedat12
months),a
greaternumberofchildren
participatedinthe
12-month
thaninthe
6-month
laboratoryassessm
ent(Table1).A
nanalysis
ofpatterns
ofmissing
datasuggested
thatdata
were
missing
completely
atrandom
,Little’s
Missing
Completely
atRandom
(MCAR)test,
$2(131)
!148.23,p
%.10.
Profileswere
createdusing
astatistical
procedure(described
below)thatincludesfull-inform
ationmaxim
umlikelihood
(FIML)
estimation.Thatis,allavailable
datawere
usedtoderive
individ-uallikelihoods
ofprofile
groupmembership.Because
ofthis,all
individualswith
lifestress
measures
hadan
availableprofile
assignment,
andacom
plete-casesanalysis
strategywasused.
Additionally,
multilevel
analyseswere
usedtoaccount
forthe
correlatednature
ofthedata
(i.e.,twins
within
families).
Results
Analyses
were
conductedinaccordance
with
studyaim
sand
hypotheses.After
weexam
ineddescriptive
statisticsfor
allvari-ables,profiles
ofinfantangerwere
derivedfrom
6-and12-m
onthobservations
ofanger
andvalidated
bytesting
forprofile-based
differencesincom
positemeasures.Bivariate
associationsbetween
preschoolbehaviorproblem
sand
measures
ofanger
were
exam-
inedforboth
traditional(i.e.,levelsofobservedangerquantified
ascom
posites)and
novel(i.e.,
angerprofiles)
measures
ofinfant
angerpriortomultivariate
analyses.Finally,lifestress
wastested
asamoderatorofthe
relationbetw
eeninfantangerand
preschoolbehaviorproblem
s.
Descriptive
Statisticsand
Age-Based
Change
inVariables
Descriptive
statisticsareshow
ninTable
2.Substantialvariabil-itywasobserved
forallm
easuresacross
assessments,indicating
thepresence
ofbroad
individualdifferences.
T-testscom
paredcotw
insonallstudy
variablestoensure
randomlabeling
oftwins
(i.e.,TwinAvs.Tw
inB)within
family;
allcotw
indifferences
were
nonsignificant.Additionalttestscom
paredboysand
girlsonallvariables.G
irls(M
!&0.02,SD
!0.60)
showedless
angerthan
boys(M
!0.05,SD
!0.53)
atthe
12-month
assessment,
t(879)!
&1.93,p
!.05,d
!0.12.N
oothersex
differenceswere
present.Apaired
samples
ttest
suggestedthat
observationsofanger,
t(480)!0.23,p
%.10,evinced
mean
levelstabilitybetw
een6and
12months
ofage.Additionally,bivariate
correlationssuggested
asmallam
ountofrank-orderstabilityinangerovertim
e(r
!.16,
p'.01).
2Theform
ationofeach
composite
wasalso
supportedbyaprincipal
components
analysis(PCA
).Additionaldetails
ofeachPCA
areavailable
fromauthors.
Table1
Outline
ofData
Collection
Procedures
Age
Assessm
enttypeContexts/m
easuresConstructs
6months
BehavioralvisitArmrestraint
Anger,sadness
CarseatAnger,sadness
Questionnaires
ParentingStress
IndexLife
stress
12months
BehavioralvisitArmrestraint
Anger,sadness
CarseatAnger,sadness
36months
Questionnaires
ChildBehavior
ChecklistChild
behaviorproblem
s
This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
2346BRO
OKER
ETAL.
Heterogeneity
inProfiles
ofAnger
Developm
entWetested
forheterogeneous
profilesofanger
duringinfancy
usinglatentprofile
analysis(LPA
).LPAidentifies
latentprofilesofindividuals
whodiffer
fromthe
overallgroupbyrelaxing
theassum
ptionthatallindividualsare
drawnfrom
asingle
population.Relaxing
assumptions
inthis
wayallow
smean
levelsofangerto
varyaround
separatemeans
andenables
identificationofprofiles
reflectinguniquepatternsofobserved
angeracrossepisodes.Anal-
ysesbegan
with
asingle-profile
solutionand
progresseduntil
addingadditionalprofilesno
longerimproved
fitovertheprevious
model(totalm
odelsrun
!4).Because
theLPA
procedurewas
intendedtodescribe
behavioralvariability
bothacross
contexts(arm
restraintand
carseat)
andacross
time(6-
and12-m
onthassessm
ents),com
positescores
fromindividual
episodeswere
used.Hence,
eachanalysis
includedfour
variables,reflecting
multiple
indicesofangerovertim
e,onwhich
profilegroups
were
based:armrestraintangerat6
months
ofage,carseatangerat6
months
ofage,armrestraintangerat12
months
ofage,andcar
seatangerat12monthsofage.A
multilevelfram
ework
wasused
toaccountforthe
nonindependenceofthe
twindata.
Examination
offitstatistics
(Table3)suggested
thatathree-
profilesolution
fitthedata
best.Thefirstprofile
includedchildren
withlow
scoresinboth
anger-elicitingepisodesat6
and12months
ofage(n
!96).W
elabeled
thisthelow
angerprofile.Thesecond
profileincluded
childrenwithdecreasing
scoresofangerfrom6to
12months
ofagebutw
hosescores
remained
highrelative
tothe
scoresofotherchildren
(n!83).W
elabeled
thisthe
highanger
profile.Themostcom
monprofile
includedchildren
whoshow
edmoderate
levelsofangeracrossallepisodes.However,children
inthis
groupshow
edrelative
increasesindisplays
ofanger
inboth
episodesbetween
the6-and
12-month
assessments(n
!787).W
enam
edthis
classthe
increasingprofile.G
enderwasunrelated
toprofile
membership,
$2(2)
!1.60,p
%.10.Forsubsequentanal -
ysesusing
angerprofiles,
childrenwere
assignedtothe
profile
groupfor
which
theyhad
thehighestprobability
ofmembership
(Mprobability
ofmembership
inassigned
group!0.82).
Internalvalidationofthe
latentprofiles.
Weconducted
atestofinternalvalidation
ofprofilegroup
differencesinobserved
angerineach
ofthelaboratory
episodesateach
assessment.O
uranalysis
was
conductedas
atwo-level
linearmixed
model
(Level1:fam
ily,Level2:twins
within
families)
toaccount
fornonindependence
inthe
twindesign.
Thefirstanalysisshowedthat,assuggested
byFigure1,profiles
differedinlevels
ofanger
observedinthe
carseat,F(2,503)
!25.43,p
'.01,(
p 2!.09,and
armrestraint,F(2,536)
!330.22,
p'.01,(
p 2!.55,episodesat6
months.Follow
-upcontrastsw
ithBonferroni
correctionsuggested
that,inboth
episodes,thehigh
angerprofileshow
edthe
highestlevelsofangerw
hereasthe
lowanger
profileshow
edthe
lowestlevels
ofanger.Sim
ilarly,at12months,profiles
significantlydiffered
inlevels
ofangerobservedinboth
thecarseat,F(2,774)
!15.14,p
'.01,(
p 2!.04,and
armrestraint,F(2,832)
!356.21,p
'.01,(
p 2!.46,episodes.Follow
upcontrasts
with
Bonferronicorrection
suggestedthat,for
bothepisodes,the
lowanger
profileshow
edless
angerthan
boththe
highangerprofile
andthe
increasingangerprofile.The
highand
increasingangerprofiles
were
notsignificantlydifferent.
Anger
and36-M
onthBehavior
Problems
Correlationsincluding
traditional(continuous)measures
ofan-gerrevealed
thatgreaterangerat6(r
!.38,p
'.01)butnot12
Table2
Descriptive
Statisticsfor
Primary
Variables
Variable
Units
nMinim
umMaxim
umM
SDt
6-month
observedanger
zscore
566&1.82
2.070.02
0.62&0.40
12-month
observedanger
zscore
881&1.98
1.450.01
0.570.57
36-month
behaviorproblems
Rawscore
2190.00
0.750.23
0.160.92
Lifestress
Rawscore
4550.00
29.009.66
5.93&0.13
Note.t
!productofa
t-testcomparison
oftwins.
Table3
FitIndicesfor
LatentProfilesofAnger
Number
ofprofilesinmodel
AIC
BIC$2
Entropy
Average
classassignm
entprobability
25,705.17
5,770.5290.80
!!
0.720.80
35,646.57
5,734.2970.60
!!
0.580.81
45,633.77
5,754.8622.80
0.550.86
Note.$2isderived
fromthe
Vuong-Lo-M
endell-RubinLikelihood
RatioTest.The
three-profilemodelw
asdeterm
inedtohave
thebestfit.A
IC!
Akaike
information
criterion;BIC!Bayesian
information
criterion.!!p
'.01.
-1.5 -1
-0.5 0
0.5 1
1.5
6 mo. A
rm Restraint
6 mo. Car Seat
12 mo. A
rm Restraint
12 mo. Car Seat
Observed Anger Composite (z-score)
Low AngerIncreasing AngerHigh AngerFigure
1.Cross-situational,cross-age
latentprofiles
ofinfant
anger.The
errorbars
reflectthestandard
errorofthe
mean.
This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
2347INFANTANGER,LIFE
STRESS,ANDBEH
AVIORPRO
BLEMS
(r!.05,p
%.10)m
onthswaslinked
tomore
behaviorproblems
at36months
ofage.To
ensurethat
associationstruly
reflectedlinks
toobserved
angerand
notmore
generallevelsofnegative
emotionality,w
ecreated
adirectionality
composite
thataccountedfor
levelsofsadness
observedduring
thetwoanger
episodes.Sadness
andangerw
eresignificantly
correlatedatboth
6months
(r!.39,
p'.01)
and12months
ofage
(r!.35,
p'.01).
Modeled
aftertheproceduresofEssex
etal.(2011),directionalitycom
positeswere
formedbyhalving
thedifference
between
thestandardized
angerand
sadnesscom
posites.Given
thatcoding
proceduresdid
notrequire
sadnessand
angertobe
mutually
exclusive,childrenwhoexpressed
highlevelsofnegative
emotion
overall(i.e.,bothangerand
sadness)ratherthanonly
angerwould
receiveadirectionality
scorenearzero.In
contrast,childrenwho
displayedgreaterrelativeangerw
ouldreceivepositivescores,w
ithmore
positivevalues
indicatinggreater
anger;childrenwhodis-
playedgreaterrelative
sadnesswould
receivenegative
scores,with
more
negativevalues
indicatinggreater
sadness.Thus,direction-ality
compositesreflectpropensitiesforanger(relative
tosadness)
thatare
independentfrom
overalllevels
ofnegative
emotion
(sadnessplus
anger).Mostim
portant,theyallow
atestof
addi-tional
confidencethat
findingsare
linkedspecifically
toinfant
anger.Substitutingthe
directionalitycom
positeinto
bivariateanal-
ysesrevealedthataccounting
forlevelsofsadnessleftassociationsbetw
eenangerand
behaviorproblemsunchanged
(6months:r
!.24,p
'.01;12
months:r
!–.00,p
%.10).Thatis,itrem
ainedthe
casethatonly
greaterlevelsof6-month
angerwere
associatedwithmore
36-month
behaviorproblems.
Angerprofiles
were
alsosignificantly
associatedwithbehavior
problemsat36
monthsofage,F(2,218)!
3.23,p'.05,(
p 2!.03.
Childreninthe
normative,increasing
angergroup
showedfew
erbehaviorproblem
satage
36months
thandid
childreninthe
highanger
group(t
!&8.38,
p'.05,
d!0.60).
Thisdifference
remained
significantupon
correctionfor
multiple
comparisons
(Bonferroni).
LifeStress
asaModerator
oftheLink
Between
EarlyAnger
andProblem
BehaviorsInour
finalset
ofanalyses,
wetested
earlylife
stressasa
moderator
ofthe
associationbetw
eenobserved
angerininfancy
andbehaviorproblem
sduringthe
preschoolyears.Traditionalandprofile
measures
ofanger
were
testedinseparate
analyses.A
two-levelm
ixedmodelw
asused
toaccountforthe
nestednature
ofthedata
(twins
within
families).
First,wetested
themodelthatincluded
traditionalmeasures
ofobserved
angerat6
and12months
ofage.A
t6months
ofage,
greateranger
wasassociated
withmore
behaviorproblem
sat36
months
ofage,B!0.10,SE(B)
!0.03,p
'.01.Life
stressdid
notpredict
preschoolbehavior
problems,B
!&0.01,
SE(B)!
0.00,p%.05,and
therewasnosignificant
interactionbetw
eenanger
andlife
stress,B!–0.01,SE(B)
!0.01,p
%.05.A
t12months
ofage,neitheranger,B!–0.01,SE(B)
!0.03,p
%.10,
norlife
stress,B
!&0.00,
SE(B)!0.00,
p%.10,
predictedbehaviorproblem
sduring
preschoolandnosignificantinteraction
waspresent,B
!&0.01,SE(B)
!0.01,p
%.10.
Asecond
two-levellinearm
ixedmodeltested
whetherearly
lifestress
moderated
linksbetw
eenprofiles
ofinfantanger
andpre-
schoolbehavior
problems. 3
Consistentwith
thesuggestions
of
Aiken
andWest(1991),anger
profileswere
dummycoded
suchthat
theincreasing
group,the
normative
developmental
profilebased
onitsfrequency
inthe
currentsample
andconsistency
with
expectationsbased
onthe
extantliterature,servedasthe
referentgroup.Continuousvariablesw
erecentered
priortothe
creationof
interactionterm
s.Asshow
ninTable4,therew
asasignificantinteractionbetw
eenlifestressand
bothofthedum
myvariables()R
2!.10). 4To
probethe
natureofthis
moderation,
weexam
inedthe
associationbe-
tween
earlyangerprofiles
andlaterbehaviorproblem
swhen
lifestress
wasrecentered
atlow
(–1SD)and
high(*1SD)levels
(Aiken
&West,1991;Cohen
&Cohen,1983).Recentering
thelifestress
variableallow
edustoexam
ineitseffecton
theassociation
between
angerprofiles
andbehavior
problemsasacontinuous
variablewhileelim
inatingtheneed
tocreatearbitrary
“highstress”
and“low
stress”groups.Results
showedthatatlow
levelsoflife
stress,childreninboth
thelowanger,B
!0.34,SE(B)
!0.09,p
'.01,and
thehigh
angerprofiles,B!0.20,SE(B)
!0.06,p
'.01,
showedgreaterbehaviorproblem
sthandid
childreninthe
increas-ing
angerprofile(Figure2).Incontrast,athigh
levelsoflifestress,neitherthe
low,B
!&0.18,SE(B)
!0.10,p
%.05,northe
high,B
!0.04,
SE(B)!0.06,
p%.05,
angergroups
significantlydiffered
fromthe
increasinggroup
inlevelsofbehaviorproblem
s. 5
Discussion
Ouroverarching
goalwastorefine
existingknow
ledgeabout
earlyangerdevelopm
entandthe
ways
thatangerininfancy
may
serveasarisk
factorfor
preschoolbehaviorproblem
s.Weiden-
tifiedthree
uniqueprofiles
ofanger
duringinfancy:
highanger,
lowanger,
andincreasing
anger.Wefound
thatdifferences
inpreschool
behaviorproblem
scould
bepredicted
bynovel
angerprofiles
andearly
lifestress;
preschoolbehavior
problemswere
inconsistentlylinked
totraditionalm
easuresofanger.Remarkably
forsuchalarge
sample,observationsofm
oreextrem
e,highlevels
ofinfant
angerwere
notconsistent
predictorsoflater
behaviorproblem
s.However,findings
showedthatearly
lifestress
moder-
3Forcom
pleteness,and
toensure
thatanalyses
were
notbiased
bygloballevelsofnegative
affect,wealso
testedamodelthatsubstituted
thedirectionality
compositesfortraditionalm
easuresofanger.Similarto
what
wasfound
inanalysesusing
traditionalmeasures,greaterangerat6
months
ofagewasassociated
withmore
behaviorproblemsat36
months
ofage,B
!0.18,SE(B)
!0.08,p
'.05.Life
stressdid
notpredictpreschoolbehaviorproblem
s,B!
&0.00,SE(B)
!0.00,p
%.10,and
therewasno
significantinteractionbetw
eenthe
directionalitycom
positeand
lifestress,
B!
&0.02,SE(B)
!0.01,p
%.10.A
t12months
ofage,neither
thedirectionality
composite,
B!
&0.09,
SE(B)!0.07,
p%.10,
norlife
stress,B!
&0.00,SE(B)
!0.00,p
%.10,predicted
behaviorproblem
sduring
preschool,andnosignificantinteraction
waspresent,B
!&0.01,
SE(B)!0.01,p
%.10.
4AnR2calculation
isnota
partoftheestim
ationprocedure
formixed
models;therefore,w
ecalculated
R2asthe
decreaseinthe
proportionof
errorvariance
between
models
thatdid
anddid
notinclude
theLife
Stress+AngerClassinteractions.A
lternatively,rerunningthe
analysesasaStepw
iseregression,w
hichdoes
notaccount
fortwinrelatedness
butallow
sforthe
estimation
ofR2,returns
anidenticalestim
ateof
)R2and
aprobability
levelofp'.01.
5Recodingthe
dummyvariable
totestdifferencesbetw
eenlow
andhigh
angerprofilesrevealednodifferencesin
behaviorproblemsatlow
levelsofstress,
B!
&0.14,
SE(B)!0.10,
p%.10,
andatrend
toward
more
behaviorproblemsin
thehigh
angerprofilethan
inthe
lowangerprofile
athigh
levelsoflife
stress,B!0.22,SE(B)
!0.11,p
!.05.
This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
2348BRO
OKER
ETAL.
atedpreschoolbehaviorproblem
swhen
angermeasures
reflectedcross-contextand
cross-timevariability.W
henearly
lifestressw
aslow
,infantswithatypicalprofilesofangerin
infancyshow
edmore
preschoolbehaviorproblemsthan
didtypically
developinginfants.
Athigh
levelsofearly
lifestress,infants
withatypicalprofiles
ofanger
showedlevels
ofpreschool
behaviorproblem
ssim
ilarto
typicallydeveloping
infants.Notably,
thesefindings
were
notpresentin
analysesusing
traditionalmeasures
ofanger,which
didnotdifferentiate
typicalandatypicaldevelopm
entalprofiles.
IndividualDifferences
inContinuous
andCategorical
Measures
ofAnger
Consistentwithhypotheses,w
efound
evidencefor
threepro-
filesduring
infancy.Profiles
were
formed
frommultim
easurecom
positesofangerintwocontextsat6
and12monthsofage.A
substantialproportionofinfants(19%
)showedatypicalprofilesof
angerduringthis
period.Atypicalprofiles
ofangerdevelopment,
which
donotadhereto
thetypicaldevelopmentalcourseasdefined
inthe
literature,have
largelygone
unnoticed,and
highlightingthese
atypicalprofilesisamajorcontribution
ofthiswork.
Characterizingindividualdifferences
inprofiles
ofearly
angeralso
helpedclarify
theimportance
ofearlyangerforlaterbehavior
problems.Wefound
thatoverall
levelsofearly
anger,when
quantifiedusing
traditionalcom
positemeasures,did
notconsis-
tentlypredictparent-reported
behaviorproblemsatage
36months.
Incontrast,
more
novelprofiles
ofanger
thatdescribed
cross-contextand
cross-timevariability
didshow
groupdifferences
inbehaviorproblem
s.Childrenwithtypicalprofilesofangershow
edfew
erbehavior
problemsthan
didchildren
with
atypicalanger
profilesbetw
een6and
12months
ofage,
althoughsubsequent
analysesshow
edthat
thisassociation
wasfurther
moderated
byearly
lifestress.
Weemphasize
thatexpressionsofanger,especially
duringin-
fancy,donot
necessarilyreflect
maladaptive
tendencies.Anger
typicallyincreases
duringthe
6–12-month
period(Braungart-
Riekeretal.,2010),suggestingthatincreases
inangerare
norma-
tiveratherthan
problematic
earlyinlife.O
urfindingsare
consis-tentw
iththesestudies.W
efoundthatthem
ajorityofinfants(81%
)show
edincreasesin
displaysofangerbetween
6and
12monthsof
ageintwolaboratory
contextsand,atthe
12-month
assessment,
theseinfants
were
indistinguishablefrom
otherinfantswhowere
classifiedasstably
highinanger.This
overlapbetw
een“typical”
andputatively-at-risk
infantswasprecisely
thetype
ofunderlyingprocess
thatwesuggested
might
beobscuring
therelationship
between
levelsofanger
at12months
andbehavior
problems2
yearslater.Stated
differently,highlevels
ofanger
at12months
may
notbe
predictiveoflater
outcomesbecause,
atthis
age,putatively
at-riskand
typicallydeveloping
childrendisplay
similar
levelsofanger.
ProfilesofInfantA
nger,EarlyLife
Stress,and
BehaviorProblem
sConsistentw
ithexpectations,w
efound
thatlevelsofstress
inthe
earlyenvironm
entmoderated
theassociation
between
angerininfancy
andpreschool
behaviorproblem
s.Given
lowlevels
ofearly
stress,themostim
portantriskfactorforpreschoolbehavior
problemswasnotlevels
ofangerduringinfancy
butwhetherthe
developmentalprofile
ofanger
wasnorm
ativeoratypical.Chil-
drenwithnonnorm
ativeprofiles
ofanger
ininfancy
hadgreater
behaviorproblem
sat36
months
compared
withthose
withmore
typicalprofiles
ofinfant
anger.This
risk-relateddistinction
be-tween
normative
andatypicalprofiles
ofearlyangerw
asnotseen
athigh
levelsofearly
family
stress.Sim
ilarly,this
moderated
effectwasnotevidentin
analysesthatused
continuousmeasures
ofanger,underscoringthe
importance
ofdifferentiatingindividual
differencesinearly
emotion,particularly
asitrelatestonorm
ativeand
atypicaldevelopment.
Itisasom
ewhatsurprising
findingthatearly
lifestress
didnot
directlypredictincreased
behaviorproblemsinyoung
children.Itispossible
that,inatypically
developingsam
pleofyoung
chil-dren,there
isdim
inishedoverallvariability
inbehaviorproblem
s,which
makes
significanteffectsmore
difficulttodetect.Itis
alsopossible
thatourmeasure
oflifestressisone
thatisdistalfromthe
truemechanism
linkinglife
stresswithbehaviorproblem
s,making
main
effectsmore
difficulttoobserve.
However,
thislack
ofa
significantmain
effectmaybebetterexplained
bythe
presenceof
thesignificantinteraction
between
earlylife
stressand
behaviorproblem
s.Thatis,theinteraction
suggeststhatthedevelopm
entofbehaviorproblem
sisdependentnotonly
onlife
stressbutalso
onearly
anger.Statistically,
thepresence
ofmoderation
hasbeen
Table4
TestofLifeStress
asaModerator
oftheRelation
BetweenAnger
Profileand
BehaviorProblem
s
Variable
BSE(B)
,t
)R2
Step1
0.12!!
Lowanger a
0.100.05
0.192.00
!
High
anger b0.12
0.040.28
2.91!!
Lifestress
&0.01
0.00&0.17
&1.73
†
Step2
0.10!!
Lowanger a
0.080.05
0.151.57
High
anger b0.12
0.040.27
2.98!!
Lifestress
&0.00
0.00&0.03
&0.28
LowAnger a
+Life
stress&0.04
0.01&0.28
&3.02
!!
High
Anger b
+Life
stress&0.01
0.01&0.20
&2.04
!
aDummyvariable
testslow
angerprofile
relativetoincreasing
angerprofile
referencegroup.
bDummyvariable
testshigh
angerprofilerela -
tivetoincreasing
angerprofilereference
group.†p
'.10.one-tailed.
!p'.05.tw
o-tailed.!!p
'.01.tw
o-tailed.
Figure2.
Atypicalangerprofilesshow
heightened36-m
onthbehaviorprob-
lemsatlow
levelsoflife
stressrelative
totypicalangerprofiles.
!!p
'.01.
This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
2349INFANTANGER,LIFE
STRESS,ANDBEH
AVIORPRO
BLEMS
knowntoobscure
thepresence
ofmain
effectsand
make
themmore
difficulttointerpret
(Aiken
&West,
1991).This
maybe
particularlytrue
incasesin
which
effectsarenonlinear,suggesting
thatmoderate
levelsofstress
areadaptive
whereas
abnormally
loworabnorm
allyhigh
exposuretostress
may
inhibitthe
development
ofregulatory
mechanism
sthat
helptokeep
be-havior
problemsatbay.
Ourresults
suggestthat
riskfactors
fordeveloping
behaviorproblem
sare
distinctatdifferent
levelsoflife
stress.Previous
work
hasshow
nthat
thepresence
ofrisk
factorssuch
aslow
socioeconomicstatus,
parentmental
illness,ormarital
discordincrease
riskfor
behaviorproblem
s(Sam
eroff,Seifer,
Barocas,Zax,&
Greenspan,1987).Each
oftheserisk
factorsiscapturedto
someextentin
ourmeasure
ofstressfullifeevents.W
epointout
thatthe
PSI(Loyd
&Abidin,
1985)taps
nonsevere,relatively
commonlife
stressorsratherthan
extremeorsevere
conditionsof
stress(e.g.,maltreatm
ent).Thus,acontribution
ofthiswork
isthefinding
thatatypicalangerprofilesmayalso
serveasrisk
factorsfor—
andalternate
pathways
toward—
negativeoutcom
es.When
levelsoflife
stresswere
low,the
riskassociated
with
atypicalprofiles
ofdevelopm
entwasmore
tightlylinked
tochildhood
outcomes.H
owever,atypicalprofilesdid
notappeartoexacerbate
theeffects
ofhighstress.
Themechanism
sby
which
earlylife
stressorsmight
affectchildren’s
functioningduring
infancyare
thesubject
ofawide
rangeofinvestigations
acrossnum
erousfields
ofpsychology.
Anim
aland
humanmodels
positthat
stressmay“get
underthe
skin”inearly
development
bydisrupting
thedevelopm
entof
physiologicalsystems,contributing
tooverallw
earandtearon
thebody,altering
thedevelopm
entofneuralcircuitry,and
program-
ming
individual-leveltendencies
forhypo-
orhyperactive
stressresponses
(McEw
en,2012;Shirtcliff&Ruttle,2010).A
ddition-ally,an
environmentofearly
stressmayindirectly
affectchildrenby
impeding
parents’abilities
tointeract
with
theirchildren
inwaysthatcontribute
tothe
developmentofindependentregulation.
Namely,greaterpoverty
inthe
earlyenvironm
enthasbeen
linkedtogreaterstress
andless
maternalsensitivity
duringparent–child
interactions(NationalInstitute
ofChildHealth
andHumanDevel-
opmentEarly
ChildCare
ResearchNetwork,2005);less
sensitiveparentsm
aybelesslikely
toprovide
thescaffold
thatisnecessaryfor
childrentolearn
toappropriately
regulatenegative
emotions
suchasanger(K
opp,1982).While
thesemechanism
sarelikely
toplay
importantroles
inthe
linkbetw
eenangerand
earlybehavior
problems,directtests
ofthese
possibilitieswillbe
importantfor
futureresearch.Sim
ilarly,theidentification
ofmore
refinedbio-
logicalandsocialm
echanismsofearly
riskrem
ainsacriticalaim
fordevelopmentalresearch.
Tothe
extentthat
stressand
atypicaldevelopm
entrepresent
independentrisk
factors,alack
ofassociation
between
angerprofiles
andbehaviorproblem
sathigh
levelsofstress
wassom
e-whatsurprising.Conceivably,facets
ofearly
lifethatw
edid
notstudy,
suchasparent
sensitivityand
responsiveness(Cam
pbell,Shaw
,&Gilliom
,2000;Zeanah,Boris,&Larrieu,1997)
might
furthermoderate
theimpactof
earlystressors
onchildren’s
out-com
es.Forexample,despite
continuedadversity
andstress
inthe
family,youthsw
ithhighly
responsiveparentsduring
infancyare
atdecreased
riskfordisruptive
behaviordisordersrelative
toyouths
with
unresponsivemothers
(Wakschlag
&Hans,
1999).Thus,
underconditions
ofhigh
stress,thedegree
towhich
parentscan
maintain
sensitive,responsiveorientation
toward
childrenmaybe
anadditional
predictorofchild
outcomesthan
children’sanger
profilesorlife
stress.Examinations
ofangerprofileswithin
vari-ous
parentingcontexts
willbeanimportant
avenuefor
futureresearch.Theories
abouthowthe
socialenvironmentinteracts
withtem
-peram
entalcharacteristics,suchasanger,to
predictmentalhealth
outcomeslargely
centeronthe
ideathatgenetic
and/orbiologicalfactors
canshape
individualsensitivity
toenvironm
entalinflu-
ences(Reiss,Leve,&
Neiderhiser,2013).W
emightask
whether
ourresults
areconsistentw
iththese
theories.Ourresults
donot
implicate
asingle
profileofearly
angerthatishighly“susceptible”
accordingtoexisting
definitions.Additionally,although
associa-tions
between
profilesand
behaviorproblemsdiffered
undercon-ditions
ofhighand
lowstress,these
were
notclearconditionsof
bothdisrupted
andenriched
development,w
hichisacentraltenet
ofbiological
susceptibility.That
is,the
identificationofhighly
sensitiveprofilesdependson
measuresthatinclude
afullrange
ofpositive
andnegative
environmental
conditions.Although
ourmeasure
ofenvironm
entalstress
enabledparents
toreport
veryhigh
levelsofearlystress,they
didnottap
enrichedenvironm
entsonthe
samespectrum
.Rather,our“lowstress”
environmentbetter
reflectedalack
ofenvironmentalstressors.
Limitations
Thecurrent
studyisnot
without
limitations.
First,unequal
numbers
ofinfantsineach
oftheprofiles
ledtodifferentdegrees
ofpow
ertodetectsignificantdifferences
acrosspairings
oftra-
jectoryclasses.A
dditionally,clinicalassessments
ofparents
andwithchildren
atfollow-up
areunavailable
forthecurrentsam
ple.Thus,although
wecan
testassociationsbetw
eenearly
angerand
parent-reportedpreschoolbehaviorproblem
s,thedegree
towhich
anytrajectory
isassociatedwithsubsequentdiagnosesisunknow
n.Finally,although
wedonotview
thetwindesign
asalimitation,
werecognize
thattheremaybeconcerns
thattwinsam
plesmay
notrepresent
thegeneral
populationonone
ormore
aspectsof
behaviorthat
maybias
resultsinsom
esystem
aticway.In
fact,there
aredifferences
between
twins
andnontw
insearly
inlife
infactors
suchasgestationalage
atbirth,birthweight,and
rateof
languageacquisition
(Plomin,D
eFries,McClearn,&
McGuffin,
2008).However,m
anyofthese
differencesdecline
anddisappear
bythe
earlyschoolyears.Results
fromtwinstudies
largelygen-
eralizetonontw
ins(Andrew
etal.,2001),includingonmeasures
relevanttothe
currentwork,such
asearly
temperam
ent(Gold-
smith&Cam
pos,1990),personality(Johnson,M
cGue,K
rueger,&Bouchard,
2004)and
psychopathology(Christensen,
Vaupel,
Holm
,&Yashin,1995).
ConclusionIndividual
differencesinprofiles
ofanger
arepresent
duringinfancy
andare
associatedwithvariability
inbehaviorproblem
satage
36months.In
addition,atlow—butnothigh—
levelsoflife
stress,nonnorm
ativepatterns
ofinfant
angerpredicted
greaterparent-reported
preschoolbehaviorproblems.O
urresultsprovide
additionalevidence
thatstress
inthe
earlyfam
ilycontext,even
when
notsevere,
isrelated
todevelopm
entaloutcom
es.Impor-
tantly,thiswork
underscoresthe
needtoconsider
individualdif-
This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.