Biological Sensitivity to Context

(1)

Delft University of Technology

Biological Sensitivity to Context

Cortisol Awakening Response Moderates the Effects of Neighbourhood Density on the

Development of Adolescent Externalizing Problem Behaviours

Yu, Rongqin; Nieuwenhuis, Jaap; Meeus, Wim; Hooimeijer, Pieter; Koot, Hans M.; Branje, Susan

DOI

10.1016/j.biopsycho.2016.08.004

Publication date

2016

Document Version

Final published version

Published in

Biological Psychology

Citation (APA)

Yu, R., Nieuwenhuis, J., Meeus, W., Hooimeijer, P., Koot, H. M., & Branje, S. (2016). Biological Sensitivity

to Context: Cortisol Awakening Response Moderates the Effects of Neighbourhood Density on the

Development of Adolescent Externalizing Problem Behaviours. Biological Psychology, 120, 96-107.

https://doi.org/10.1016/j.biopsycho.2016.08.004

Important note

To cite this publication, please use the final published version (if applicable).

Please check the document version above.

Copyright

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons. Takedown policy

Please contact us and provide details if you believe this document breaches copyrights. We will remove access to the work immediately and investigate your claim.

This work is downloaded from Delft University of Technology.

(2)

Contents lists available atScienceDirect

Biological

Psychology

j o u r n a l h o m e p a g e :w w w . e l s e v i e r . c o m / l o c a t e / b i o p s y c h o

Biological

sensitivity

to

context:

Cortisol

awakening

response

moderates

the

effects

of

neighbourhood

density

on

the

development

of

adolescent

externalizing

problem

behaviours

Rongqin

Yu

a,∗

_,

_Jaap

_Nieuwenhuis

b,c

_,

_Wim

_Meeus

a,d

_,

_Pieter

_Hooimeijer

c

_,

_Hans

_M.

_Koot

e

_,

Susan

Branje

a

a_Department_of_Youth_and_Family,_Utrecht_University,_Utrecht,_The_Netherlands

b_Department_OTB_–_Research_for_the_Built_Environment,_Delft_University_of_Technology,_Delft,_The_Netherlands c_Department_of_Human_Geography_and_Planning,_Utrecht_University,_Utrecht,_The_Netherlands

d_Department_of_{Developmental}_Psychology,_Tilburg_University,_Tilburg,_The_Netherlands

e_Department_of_Clinical_{Developmental}_Psychology_and_EMGO_Institute_for_Health_and_Care_Research,_VU_University_Amsterdam,_Amsterdam,_The

Netherlands

a

r

t

i

c

l

e

i

n

f

o

Articlehistory: Received30July2015

Receivedinrevisedform6July2016 Accepted13August2016 Availableonline16August2016 Keywords:

Biologicalsensitivity Externalizingbehaviours Cortisolawakeningresponse(CAR) Neighbourhooddensity

a

b

s

t

r

a

c

t

Thisfour-yearlongitudinalstudyattemptedtotestperson-environmentinteractiontheoryandbiological sensitivitytheorybyassessingwhetherindividuals’biologicalstressactivityCARAUCg(CortisolAwakening

ResponseAreaUndertheCurvewithrespecttoground)moderatestheeffectsofneighbourhooddensity onthedevelopmentofadolescentexternalizingproblembehaviours.Participantswere358Dutch adoles-centswithameanageof15yearsattheﬁrstmeasurement.OuranalysesshowedthatCARAUCgmoderated

theeffectsofneighbourhooddensityonthelevelofparent-reporteddelinquencyandaggressionand adolescentself-reporteddelinquency.Morespeciﬁcally,foradolescentswithhighCARAUCg,higher

neigh-bourhooddensitysignificantlypredictedhigherlevelsofparent-reportedandadolescentself-reported delinquencyandaggression,whereastheassociationwasreversedornon-significantforadolescents withlowCARAUCg.OurfindingssuggestthatadolescentswithdifferentlevelsofCARAUCgrespond

differ-entiallytothedensityoftheneighbourhoodtheylivein,supportingforperson-environmentinteraction perspectivesandbiologicalsensitivitytheory.

1. Introduction

Highneighbourhooddensityhasbeencommonlyconsidered

anenvironmentalstressor.It couldcreate variousstressessuch

aslimitedenvironmentalresources,restrictedfreedomofchoice,

andtheexperienceofacrowdedandnoisyenvironment,which

could potentially stimulate individuals’ externalizing problem

behaviours(Regoeczi,2008).Theextenttowhichhigh

neighbour-hood density contributes toadolescents’ externalizing problem

behavioursmight bedetermined bywithin-individualvariables

thataffectthewayneighbourhooddensityisprocessed.Theeffects

of neighbourhooddensity on thedevelopment of externalizing

problembehaviours mightbe dependentonadolescents’stress

sensitivityasindicatedbywell-knownbiologicalmarkerslikethe

cortisolactivity(Hellhammer,Wüst,&Kudielka,2009).Cortisolis

∗ Correspondingauthor.Currentaddress:DepartmentofPsychiatry,Universityof Oxford,WarnefordHospital,WarnefordLn,OxfordOX37JX,UnitedKingdom.

E-mailaddress:Rongqin.yu@psych.ox.ac.uk(R.Yu).

oneofthestresshormonessecretedbyamajorhuman

physiolog-icalstresssensitivitysystem,thehypothalamic-pituitary-adrenal

(HPA)axis.Cortisolactivity isanimportant biological indicator

ofself-regulationandit playsacrucial roleintheregulationof

individuals’emotionalandbehaviouralresponsetoenvironmental

stressors(Fries,Dettenborn,&Kirschbaum,2009).Thus,cortisol

activitymightmoderatetheeffectsofneighbourhooddensityon

adolescents’externalizingproblembehaviours.

Several theories support the assumption that effects of

neighbourhooddensitydependonindividualcharacteristics.An

interactionistperspectivepositsthatindividuals’developmental

outcomes depend on the interplay between individual

charac-teristicsandenvironmentalfactors(Magnusson&Stattin,2006).

Morespeciﬁcally,hormones-contextinteractiontheoryassertsthat

individuals’hormone secretioncaninteractwithenvironmental

factorstopredictantisocialbehaviours(Susman,1997).

Further-more,accordingtothenotionofbiologicalsensitivity-to-context

(Ellis,Essex,&Boyce,2005),certainbiologicalmarkersthatmake

individualsmorevulnerabletoenvironmentaladversityalsomake

them morelikely to beneﬁtfrom positive environmental

inﬂu-http://dx.doi.org/10.1016/j.biopsycho.2016.08.004

(3)

ences.Takentogether,individualswithdifferentstressresponse

setsasindicatedbytheircortisolactivitywouldbeexpectedto

differentially respondtohighdensity neighbourhoods in terms

ofelevatedlevelsofexternalizingproblembehaviours.

Individu-alshighincortisolactivitymay,duetotheirhighersensitivityto

theenvironment,bemorelikelyinﬂuencedbybothpositiveand

negativeaspectsoftheneighbourhoodinwhichtheylive.The

cur-rentstudyaimedtoexaminewhetherbiologicalcharacteristics(i.e.,

morningcortisolactivity)moderatedtheeffectsof

environmen-talcontext(i.e.,neighbourhooddensity)onthedevelopmentof

adolescents’externalizingproblembehaviours.

1.1. Neighbourhooddensityandexternalizingproblem

behaviours

Environmentalstressorssuchasahighdensityneighbourhood

might have a negative impact on youthexternalizing problem

behaviours(Lepore,Evans,&Palsane,1991).Increasesin

neigh-bourhood density may be accompanied by thepresence of an

increasingnumberofpeople,whichcouldcauseacrowding

experi-ence,andbyadecreaseinopenspace,whichmayposeconstraints

suchaslimitedaccesstopublicresources.Thesepotential

nega-tiveenvironmentalstressorsmightrestrictandinterferewiththe

attainmentofone’sgoals,whichmaythusgeneratefrustrationthat

couldstimulateaggressiveanddelinquentbehaviours(Regoeczi,

2003).Inaddition,high-densityenvironmentsmightcreateamore

impersonalatmosphereandprovideagreaternumberoftargets

forexternalizingbehaviours(Cohen&Felson,1979;Stark,1987).

Theaccompanyinganonymityof highpopulationdensitymight

alsoformastructurallimittowhatcanbeachievedthroughsocial

control(Sampson&Raudenbush,1999).

Empirical studies investigating the associations between

neighbourhood density and adolescents’ externalizing problem

behaviourshaveyieldedmixedresults.First,studieswhich

mea-suredthephysical density ofa neighbourhood(e.g.,number of

residentialaddressespersquarekilometre)reportedinconsistent

ﬁndings.Weenink(2011)revealedthatcomparedtoadolescents

livinginDutchurbanareas(>1000addressespersquarekilometre),

adolescentslivinginruralareasareslightlylesslikelytoengage

indelinquentbehavioursincludingpropertyoffensesandviolence.

However,anotherDutchstudyreportedthaturbanization(i.e.,high

residentialaddressesdensity)hadnoeffectsonyouthparent-and

self-reportedexternalizingproblembehaviours(Reijneveldetal.,

2010).Otherstudieswhich used social density (i.e.,population

density:numberofinhabitantsinacertainarea)toassess

neigh-bourhooddensityalsorevealedcontradictoryresults.Somestudies

foundpopulationdensitytobelinkedtoadolescentdelinquentand

criminalbehaviours(Schmitt, 1957,1966),whereas other

stud-iessuggestthatpopulationdensityhaslittleeffectondelinquency

orcrime(Gillis&Hagan,1981;Gillis,1974).Furthermore,Harden

etal.(2009)reportedinconsistentresultsregardingthe

neighbour-hooddensity-externalizingproblembehavioursassociationacross

informants.Inparticular,theyfoundthatpopulationdensitywas

positively associated withyouth self-reportedbut not

mother-reporteddelinquentbehaviours.Hence,existingempiricalstudies

didnotalwaysﬁndasigniﬁcantassociationbetween

neighbour-hooddensityandadolescentexternalizingproblembehaviours.

1.2. Cortisolawakeningresponseandexternalizingproblem

behaviours

TheCortisolAwakeningResponse(CAR)isconsideredanindex

oftheeverydayhumanadrenalcortisolsecretoryactivityandis

deﬁnedasthechangeincortisolconcentrationthatoccurs

dur-ingtheﬁrsthourafterawakingfromsleep(Clow,Thorn,Evans,&

Hucklebridge,2004).Itistypicallyfeaturedbyabriskincreaseof

cortisollevelswithin20-30minafterawakeninginthemorning

andadeclinethereafter(Clowetal.,2004).Theoverallvolumeof

cortisolreleasedoverthewakingperiod(AreaUndertheCurvewith

respecttoground[AUCg];furtherreferredtoascortisolactivity),

whichwasthemainindicatorofCARinthecurrentstudy,presentsa

usefulandreliableindexofadrenocorticalactivity(Pruessneretal.,

1997;Schmidt-Reinwaldetal.,1999).

Different theories have hypothesized different directions of

the link between cortisol activity and externalizing problem

behaviours. Two major theories have postulated an

associa-tionbetweenexternalizingproblembehavioursandphysiological

hypo-arousal.Thefearlessnesstheorysuggeststhatalowtendency

tobecomearousedinreactiontofearfulstimuliwouldresultina

higherlikelihoodtobecomedisruptiveandantisocial(Raine,1993).

Inaddition,thesensation-seekingtheory(Raine,1993;Zuckerman

&Neeb,1979)hypothesizesthatlowarousalisanunpleasant

phys-iologicalstateandinordertogetridofthisstate,individualswith

lowarousallevelswouldseekstimulationbyinitiatingantisocial

behavioursthat increasephysicaltension. Followingthis lineof

reasoning,lowercortisollevelsshouldberelatedtohigherlevelsof

externalizingproblembehaviours(Brennan&Raine,1997).Onthe

otherhand,however,researchersrecentlyhavehypothesizedthat

hyper-arousalisariskfactorinstressfulcontexts.Hyper-arousal

maybeevidentasheightenedphysiologicalresponsestostressor

threats,suchasirritability,hypervigilance,frustration,andan

exag-geratedstartleresponse,whichcouldinturntriggerexternalizing

problembehaviourssuchasaggressiveanddelinquentbehaviours

(Kerig,Vanderzee,Becker,&Ward,2012).Hence,followingthis

lat-terlineofreasoning,highercortisollevelsindicatinghyper-arousal

shouldbeassociatedwithhigherlevelsofexternalizingproblem

behaviours.

Evidenceregardingtheassociationsbetweenmorningcortisol

levelandadolescentexternalizingproblembehavioursismixed.

Several studies examined cross-sectional associations between

activityoftheHPA-axisanddisruptivebehavioursinadolescents.

Some researchers found an association between low morning

basalcortisollevelsandadolescentdisruptivebehaviours(Pajer,

Gardner,Rubin,Perel,&Neal,2001),whereasothersdidnotfound

suchassociation(Ruttleetal.,2011;Scerbo&Kolko,1994).

Longi-tudinalstudieslookingatwhetherlowmorningcortisolactivity

is a risk factor for future externalizing problembehaviour also

revealedinconsistentresults(Shirtcliff,Granger,Booth,&Johnson,

2005; Shoal, Giancola, & Kirillova, 2003; Van Bokhoven et al.,

2005).Speciﬁcally,lowmorningcortisollevelspredicted

adoles-centaggressivebehaviours5yearslater(Shoaletal.,2003),risky

behaviours oneyearlater(Shirtcliffetal.,2005), andpersistent

aggressivebehaviouroverthreeyears(Platjeetal.,2013).Onthe

contrary, Van Bokhoven et al. (2005) found that higher

morn-ingcortisollevelsatage13predictedhigherconductdisorderin

boysatage14–16.Furthermore,studiesalsoreportedalackof

associationsbetweenmorningbasalcortisolleveland

externaliz-ingproblembehavioursinadolescents(Dabbs,Jurkovic,&Frady,

1991;Klimes-Dougan,Hastings,Granger,Usher,&Zahn-Waxler,

2001;Sondeijkeretal.,2008).Therefore,evidenceforthe

associa-tionbetweenmorningcortisolactivityandexternalizingproblem

behavioursinadolescenceislargelyinconsistent,asnegative,

pos-itive,aswellasnon-signiﬁcantrelationshavebeenreported.

1.3. Moderationofneighbourhooddensityeffectson

externalizingproblembehavioursbyCARAUCg

According to person-environment interaction perspectives,

externalizing problem behaviours are likely the result of both

environmentalinﬂuencesandindividuals’biological

(4)

individualswithhighbiologicalsensitivitymightespeciallybeneﬁt

fromapositiveenvironmentbutmightbevulnerabletothe

neg-ativeinﬂuenceofanadverseenvironment(Ellis&Boyce,2011).

Thepresenceofinteractioneffectsbetweenneighbourhood

den-sityandstresssensitivityasindicatedbymorningcortisolactivity

mightprovidesomeexplanationsfortheinconsistentﬁndingson

theassociationbetweenneighbourhooddensityandexternalizing

problems.Responses toa high-densityneighbourhood

environ-mentmaydifferforadolescentswhovaryintheirmorningcortisol

levels(Hellhammeretal.,2009).Forinstance,althoughsome

ado-lescentsmayviewahighdensityneighbourhoodanopportunity

forsocialinteractionsandnetworks,otheradolescents,whohave

difﬁcultyinsocialandemotionalregulation,mayseeitascrowded

andnoisyorfeelthreatedintheiraccessibilitytoneighbourhood

recourses.

Apattern of hypo-arousalhasbeentermed a low biological

sensitivitytocontext (Elliset al.,2005).A lowbiological

sensi-tivitytocontextis deﬁnedasindividuals’lowabilitytoprocess

environmentalinﬂuences. Low biological sensitivity individuals

mightdisplayalackofbiologicalsensitivitytoenvironmental

chal-lenges,theymightbebufferedagainststressors(Ellisetal.,2005).

Ontheotherhand,however,individualswithhyper-arousalhave

a higher biological sensitivity to context, which means greater

responsivenesstobothpositiveandnegativeenvironmental

inﬂu-ences.Therefore,adolescentswithhighCARAUCgmightbemore

likelytoshowincreasesinexternalizing problembehaviours in

responsetoahigh-densityneighbourhoodandalsobemorelikely

toshowdecreasesinexternalizingproblembehavioursinresponse

tolow-densityneighbourhood.Incontrast,adolescentswithlow

CARAUCgmaybenotaffectedbytheirenvironmentandtheirlevels

ofexternalizingbehavioursmightnotvarywiththedensityoftheir

neighbourhood.Empiricalresearchontheinterplaybetweenbasal

cortisollevelandneighbourhoodadversityhasbeenlimited.This

studyattemptedtoﬁlltheknowledgegapbytestingthe

moderat-ingroleofCARAUCgontheeffectsofneighbourhooddensityonthe

developmentofadolescents’externalizingproblembehaviours.

1.4. Thecurrentstudy

The aims of the current study were to examine

person-environmentinteractioneffectsinpredictingadolescents’

devel-opmentofexternalizingproblembehaviours.Weﬁrstreplicated

previousresearchbyestimatingthemaineffectsofneighbourhood

density and CARAUCg on adolescents’ development of

adoles-centexternalizingproblembehavioursincludingaggressionand

delinquencyacross a period of four years frommiddle to late

adolescence.Asresultsfrompreviousstudieswereinconsistent,

noconcrete hypothesescouldbe formulated.Subsequently, we

investigatedinteractioneffectsbetweenCARAUCgand

neighbour-hooddensityonthedevelopmentof adolescentaggression and

delinquency. We expectedthat adolescents withhighCARAUCg,

comparedtothosewithlowCARAUCg,wouldscorehigherin

exter-nalizingproblembehavioursinahigh-densityneighbourhoodbut

scorelowerinexternalizingproblembehaviourinalow-density

neighbourhood.

2. Method

2.1. Sample

Participantswere358adolescents(205[57.3%]boys)who

par-ticipatedinalongitudinalstudyonadolescentdevelopmentand

whotookpartincortisolawakingmeasurementsatWave3,with

ameanageof15.03years(SD=0.45).TheywerepartoftheRADAR

study(N=497,ResearchonAdolescentDevelopmentAnd

Relation-ships;Meeusetal.,2010).RADARisanongoinglongitudinalstudy

focusingonvariousadolescentdevelopmentaloutcomesincluding

externalizingproblembehaviours.Thecurrentstudywasbasedon

datafromthethirdtothesixthwaveofRADAR.Allparticipants

inthecurrentstudyidentiﬁedthemselvesasDutch.Inthis

sam-ple,10.5%ofadolescentswerefromlowSESfamilyinwhichfather

and mother were unemployedor held an elementary job (e.g.,

constructionworker,janitor,truckdriver;Statistics-Netherlands,

1993).FamilySESwasmediumorhighfortheother89.5%ofthe

adolescents,implyingthat atleastoneof theparents’ jobswas

classiﬁedasmediumlevel (e.g.,policeofﬁcer, physician’s

assis-tant) or high level (e.g., doctor, scientist,high school teacher).

BetweenthetotalRADARsampleof497adolescentsandthe358

adolescentswhoparticipatedinthecortisolmeasurementatWave

3,theonlysigniﬁcantdifferencewasinself-reportedaggression

(F[1,373]=5.24, p=0.02), withadolescentswho participated in

thecortisolmeasurementreportingahigherlevelofaggression.

Therewerenosigniﬁcantdifferencesbetweenthesetwogroups

inparent-reporteddelinquency(F[N=1,384]=0.25,p=0.66)and

aggression(F[N=1,384]=0.19,p=0.66),self-reporteddelinquency

(F[N=1, 387]=2.52, p=0.11), gender distribution(␹2 [N=497,

1]=0.05, p=0.82),age(F[3,494]=0.05, p=0.77),and

neighbour-hooddensity(F[N=469,1]=1.92,p=0.17).

ThenumberofparticipantsacrossWave3toWave6ﬂuctuated

peryear,with358participantsatWave3,347atWave4,338at

Wave5,and323atWave6.Attritionwas9.8%overtheperiodof

threeyears.Toestimatethepatternofmissingdata,Little’sMissing

AtRandom(MCAR)testwasconducted(Little,1988)onallvariables

usedinthisstudy.Little’sMCARtestrevealedanormed␹2(␹2/df)

of1.31which,accordingtoguidelinebyBollen(1989),indicates

thatthepatternofthemissingdatawasnotmeaningfully

differ-entfromamissingcompletelyatrandompattern.Therefore,we

appliedFullInformationMaximumLikelihood(FIML)inMplusfor

themodelestimations(Schafer&Graham,2002).

2.2. Procedure

Participants were recruited from various Dutch elementary

schools.Aninvitationletterandadescriptionofthestudywere

senttoadolescents’homeaddresses.Bothparentsandadolescents

providedinformedconsenttoparticipate.Adolescentsandparents

ﬁlledoutvarious questionnairesduringtheannual homevisits,

supervisedbytrainedresearchassistants.Inadditionto

administra-tionofthebehaviouralmeasurements,trainedresearchassistants

gavedetailedverbalandwritteninstructionforcortisol

measure-mentsduringthehomevisit.Conﬁdentialitywasassuredexplicitly

beforeparticipationinthestudy.ParticipantsreceivedD15asa

rewardfortheirparticipationineachwave.TheRADARstudyhas

beenapprovedbytheresponsiblemedicalethicscommittee,and

wasconductedinaccordancewiththeDeclarationofHelsinki.

2.3. Measure

2.3.1. Adolescentexternalizingproblembehaviours

Externalizingproblembehavioursweremeasuredthrough

par-ents’reportoftheirchild’sbehavioursand throughadolescents’

self-report.ParentsﬁlledouttheDutchversionoftheAchenbach

ChildBehaviourChecklist(CBCL;Achenbach,1991),reportingtheir

child’s delinquent and aggressive behaviours. Delinquency was

measuredusingtheDelinquent Behaviourscale (13items; e.g.,

“Setsﬁres”)andaggressionwasmeasuredusingtheAggressive

Behaviourscale(20items;e.g.,“Fightsalot”).Parentsresponded

tothequestionsona3-pointscale(i.e.,0=never;1=sometimes;

2=often). The validity and reliability of this measure hasbeen

showntobeadequate(Achenbach,1991).Inthecurrentsample,

(5)

to0.80fortheDelinquentBehaviourscale,andfrom0.89to0.90

fortheAggressiveBehaviourscale.

AdolescentsalsoﬁlledouttheYouthSelfReport(YSR;Verhulst,

VanderEnde,&Koot,1997),whichistheself-reportversionofthe

CBCL.Speciﬁcally,11itemswereemployedtoassessdelinquency

(e.g.,“Isetﬁres”)and19itemswereusedtoindicateaggressive

behaviours(e.g.,“Iﬁghtalot”).Adolescentsratedtheirfrequency

ofdelinquencyandaggressiononathree-pointLikertscale(i.e.,

0=never;1=sometimes;2=often). Acrossfourwaves,Cronbach’s

alphasrangedfrom0.70to0.75fordelinquency,andfrom0.86to

0.88foraggression.

2.3.2. Neighbourhooddensity

Neighbourhooddensitywasassessedbytheaveragenumberof

addressespersquarekilometre.Welinkedsix-digitpostcodedata

inoursampletoStatisticsNetherlands(2011)whichprovidesdata

ontheaddressdensityofthesurroundingareaforeachpost-code

in2010(i.e.,aroundtheﬁfthwaveofdatacollection).As

neigh-bourhoodswithahighdensitymightbepoorerthanthosewith

lowdensity,wecontrolledfortheeffectofneighbourhoodwealth

onneighbourhooddensityinthecurrentstudy.Weobtainedthe

averagepropertyvalueofdwellingsinneighbourhoodsmeasured

in 2004 from Statistics Netherlands (2006). The average

prop-ertyvaluecaptured thequalityof dwellingsand thesocial and

physicalattributesoftheneighbourhoodandthereforewas

con-sideredagoodproxyforneighbourhoodwealth(Visser,VanDam,

&Hooimeijer,2008).

2.3.3. CARAUCg

CARAUCgwasmeasuredinsalivawhichwascollectedby

pas-sivedrooling, immediately afterawakening (Cort0),30minutes

(Cort30),and60minutes(Cort60)later.Thesalivasamplingwas

scheduledonatypicalweekday,followingdetailedverbaland

writ-teninstruction.Participantswereinstructedtorinsetheirmouths

withwaterbeforesampling,andnottoeat,drink,smokeorbrush

theirteethbeforecompletingCort60.Theywererequestedto

col-lecttheirsalivathroughasmallstrawintoapolypropylenetube,

andlabelthesetubeswiththetimeanddateofsampling.After

collection,participants were askedto storethe samplesin the

refrigeratorandsendthembymailtotheresearchcenterthesame

day.

At the research center, the cortisol collections were stored

uncentrifugedat-20◦Cuntilanalysis.Salivarycortisollevelswere

analyzed using electrochemiluminescence immunoassay ECLIA

(E170 Roche, Switzerland). The lower detection limit was 0.5

nmol/l,andmeanintra-assayandinter-assaycoefﬁcientsof

vari-ationwere,respectively,3%and12%.Allsampleswerechecked

forcorrectnessofsampling.Caseswereexcludedfromanalysesif

thecortisoldatawereofincorrectsamplingtime,unclearhowit

wassampled(i.e.,notregistered),contaminated(e.g.,bysmoking

or brushingteeth), orof extreme values (i.e.,>3 SDfrom

aver-age).Inthecurrentstudy,358participantsprovidedqualiﬁeddata.

CARAUCgisconsideredasummaryparameteroftherepeated

mea-surementsofCAR(i.e.,0,30,and60minutesafterawakening).Thus,

itisanestimationoftotaladrenalcortisolsecretionduringtheﬁrst

hourafterawakening.WecalculatedtheCARAUCgwiththeformula

providedbyPruessner,Kirschbaum,MeinlschmidandHellhammer

(2003).

2.3.4. Controlvariables

Asgender,physicaldevelopment,substanceuse,andstressful

lifeexperiencesmightaffectCARAUCg(Clowetal.,2004;Platjeetal.,

2013),wetestedwhetherthesevariableshadasigniﬁcanteffect

onCARAUCgandcontrolledforthesigniﬁcantpredictorsinourﬁnal

statisticalmodels.Thesevariableswereassessedwhenadolescents

were15yearsold,thesameagewhentheirCARAUCgwasmeasured.

Physicaldevelopmentwasindicatedwithadolescents’bodymass

index(BMI)whichwascalculatedasweightinkg/(lengthinm)2_.

Substanceusewasindicatedasnicotineandalcoholuse.Nicotine

usewasassessedbyanine-optionquestionrangingfrom“Ihave

neversmoked”to“Ismokeeveryday”.Thedataweredichotomized:

responsesfrom“Ihaveneversmoked”to“Ismokelessthanoncea

month”weredeﬁnedas“notusenicotine”andthosefrom“Idonot

smokeweekly,butatleastonceamonth”to“Ismokeeveryday”were

deﬁnedas“usingnicotine”.Alcoholuseoverthelastfourweekswas

assessedwithaquestionwithsixresponseoptions,rangingfrom

“none”to“daily”.Thesedataweredichotomizedtoo:“none”was

deﬁnedas“notusingalcohol”andresponsesfrom“1-3daysinthe

lastfourweeks”to“everyday”weredeﬁnedas“usingalcohol”.

Stress-fulexperiencesincludingsexualassault,physicalassault,andbeing

threatenedwithviolence,weremeasuredwiththeInternational

CrimeVictimsSurvey(ICVS;Nieuwbeerta,2002).Ourregression

model withCARAUCg as a dependentvariable andwith gender,

smoking,alcoholuse,BMI,sexualabuse,physicalassault,andbeing

threatenedwithviolenceaspredictorsindicatedasigniﬁcanteffect

ofgender(␤=0.14,p=0.02).Alltheothervariablesinthemodeldid

nothaveasigniﬁcanteffectonCARAUCg(␤srangedfrom-0.01to

0.11,andps>0.05).Therefore,wecontrolledforgendereffectson

CARAUCginourfollowingstatisticalmodels.

2.4. Statisticalanalysis

Multipleregressionmodelsincorporatinglatentgrowthmodels

wereconductedtotesttheinteractioneffectsbetweenCARAUCgand

neighbourhooddensityonthedevelopmentofexternalizing

prob-lembehaviours.Beforetestingmultipleregressionmodels,weﬁrst

determinedtheshapeofgrowthinadolescentexternalizing

prob-lembehaviours(i.e.,delinquencyandaggression)fromage15to18.

Tothatend,wecomparedthechi-squarevaluesofmodels

includ-ingalinearandquadraticgrowthtocapturechangesindelinquency

andaggression(Satorra&Bentler,2001).

The modelsfor parent-reporteddelinquencyand aggression

indicated that adding quadratic slopes signiﬁcantly improved

modelﬁt(i.e.,asigniﬁcantlylowerchi-squarevalue;␹2_[N₌_358,

4]=56.38,p=0.00 and ␹2 _[N₌_358, _4]₌_21.11,_p₌_0.01,

respec-tively). For self-reported delinquency and aggression, adding

quadratic slopes did not signiﬁcantly improve model ﬁt (␹2

[N=358, 4]=9.49, p=0.26 and ␹2 _[N₌_358, _4]₌_9.54, _p₌_0.13,

respectively). However, tofacilitate the comparability between

modelsacrossparent-and self-reportedexternalizing problem

behaviours,wechosemodelswithbothlinearandquadraticslopes.

Toavoidconvergenceproblems,thevariancesofquadraticslopes

wereﬁxedatzero.

Afterdeterminingtheshapeofthegrowthofexternalizing

prob-lembehaviours,weaddedpredictorstothegrowthmodel.Weﬁrst

assessedthemaineffectsofneighbourhooddensityandCARAUCg

onthedevelopment(i.e.,interceptandlinearslope)ofadolescent

delinquencyandaggression.Next,wetestedtheinteractioneffects

byincluding theCARAUCg byneighbourhooddensity interactive

term.Inallmodels,wecontrolledforeffectsofSocial Economic

Status(SES)andgender.Furthermore,asneighbourhoodswitha

highdensitymightbepoorerthanthosewithlowdensity,

neigh-bourhooddensitywascontrolledfortheeffectsofmeanlevelsof

propertyvalues (␤s=-0.26, ps<0.00acrossmodels).AsCARAUCg

differbetweengirlsandboys,CARAUCgwascontrolledforgender

effects(␤srangedfrom0.26to0.27,psrangedfrom0.01to0.02

acrossmodels).Moreover,whentheinteractioneffectsbetween

CARAUCgandneighbourhooddensityweresigniﬁcant,weassessed

SESdifferencesintheinteractioneffectsbyconductingthree-way

interactionanalyses.Inaddition,wetestedwhetherthe

moderat-ingeffectsofCARAUCgcanbeexplainedbythepotentialoverlap

(6)

interac-Table1

MeansandStandardDeviationsoftheObservedValuesofCARAUCg,Neighbourhood

Density,Parent-andSelf-ReportedExternalizingProblemBehavioursfrom15to18 Years.

Measure M(SD)

15years 16years 17years 18years

1.CARAUCg 1030(365) – – – 2.NeighbourhoodDensity 2030(1911) – – – 3.Delinquency(CBCL) 1.85(2.15) 2.12(2.57) 2.22(2.57) 2.09(2.26) 4.Aggression(CBCL) 7.00(5.42) 6.55(5.42) 5.79(5.45) 5.05(5.04) 5.Delinquency(YSR) 3.45(2.95) 3.77(2.97) 3.83(2.87) 3.75(2.87) 6.Aggression(YSR) 7.11(5,75) 6.87(5.68) 6.43(5.49) 5.70(5.09) Note:M(SD)=Mean(StandardDeviation).CARAUCg=CortisolAwakeningResponse AreaUndertheCurvewithrespecttoground.CBCL=ChildBehaviourChecklist. YSR=YouthSelfReport.

tioneffects betweenSES and CARAUCg in predicting adolescent

externalizingproblembehaviours.Finally,wealsoexaminedifthe

moderatingeffectsofCARwerespeciﬁctoAUCg(i.e.,theoverall

cortisolsecretionpost-awakening),oralsooccurredforthe

abso-lutechangeincortisollevelspost-awakening:AUCi(Areaunder

thecurvewithrespecttoincrease;seePruessneretal.,2003for

detailedcomputationalprocedure).Tothisend,weassessedthe

interactioneffectsbetweenCARAUCiandneighbourhooddensityon

adolescents’externalizingproblembehaviours.Allanalyseswere

conductedwithinMplus7.0(Muthén&Muthén,1998–2012).In

addition,sincethedependentvariablesdelinquencyand

aggres-sionwerenotnormallydistributed,weadoptedarobustmaximum

likelihoodestimator(MLR; Satorra&Bentler, 2001),totakethe

non-normaldistributionofthedataintoaccount.

Whentheinteractioneffectsweresigniﬁcant,weexaminedthe

shapeoftheinteractionbyprobingtheinteractioneffects.To

exam-inethis,simpleslopesforeachinteractionwerepresented(at1SD

aboveandbelowmeanofthemoderators).Moreover,weapplied

theJohnson-Neymantechnique,usingthecomputationaltoolof

Preacher,Curran,andBauer,(2006),toidentifyforwhichregions

intherangeofthemoderatorvariableeffectsofthefocalpredictor

ontheoutcomeweresigniﬁcant(p<0.05)(Bauer&Curran,2005;

Hayes&Matthes,2009).

3. Results

Table1presentsdescriptivestatistics,includingthemeansand

standarddeviationsofneighbourhooddensityandCARAUCgwhen

adolescentswereonaverage15yearsofageandparent-and

self-reporteddelinquentandaggressivebehavioursacrossfourannual

waves(i.e.,fromage15to18years).Table2presentsbivariate

cor-relationsamongCARAUCgandneighbourhooddensitymeasuredat

age15years,andparent-andself-reportedexternalizing

prob-lembehavioursassessedfromage15to18years.Tables3and4

presenttheresultsofourﬁnalstructuralequationmodels

examin-ingthemaineffectsofandtheinteractioneffectsbetweenCARAUCg

andneighbourhooddensityonthedevelopmentofexternalizing

problembehavioursacrossmiddletolateadolescence.

3.1. MaineffectsofneighbourhooddensityandCARAUCgonthe

developmentofparent-reportedadolescentexternalizingproblem

behaviours

Weexaminedthemaineffectsofneighbourhooddensityand

CARAUCg on the development of parent-reported externalizing

problem behaviours, controlling for effects of SES and gender.

In the two models predicting development of parent-reported

delinquencyand aggression,the onlysigniﬁcanteffect wasthe

effect of SES (i.e., low vs. high) on the intercept of aggression

(N=358,B=2.47,␤=0.15,p=0.04).LowerSESwaslinkedtohigher Table

2 Bivariate Intercorrelations among CAR AUCg , Neighbourhood Density, Parent-and Self-Reported Externalizing Problem Behaviours from 15 to 18 Years. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 1 CAR AUCg T3 2 NB Density T3 0.05 3 Delinquency CBCL T3 0.05 0.02 4 Delinquency CBCL T4 0.04 0.03 0.74** 5 Delinquency CBCL T5 0.03 0.02 0.67** 0.77** 6 Delinquency CBCL T6 -0.04 0.05 0.60** 0.66** 0.79** 7 Aggression CBCL T3 0.00 -0.03 0.70** 0.60** 0.58** 0.55** 8 Aggression CBCL T4 0.02 -0.00 0.59** 0.72** 0.69** 0.60** 0.81** 9 Aggression CBCL T5 0.06 -0.02 0.52** 0.54** 0.73** 0.61** 0.75** 0.85** 10 Aggression CBCL T6 0.04 0.02 0.43** 0.42** 0.57** 0.66** 0.68** 0.76** 0 .83 ∗ ∗ 11 Delinquency YSR T3 0.11 0.05 0.43** 0.43** 0.41** 0.40** 0.35** 0.33** 0 .31 ∗ ∗ 0 .26 ∗ ∗ 12 Delinquency YSR T4 -0.01 0.08 0.31** 0.37** 0.37** 0.40** 0.29** 0.31** 0 .28 ∗ ∗ 0 .25 ∗ ∗ 0 .70 ∗ ∗ 13 Delinquency YSR T5 0.08 0.07 0.29** 0.33** 0.44** 0.42** 0.25** 0.29** 0 .32 ∗ ∗ 0 .26 ∗ ∗ 0 .59 ∗ ∗ 0 .66 ∗ ∗ 14 Delinquency YSR T6 -0.05 0.12* 0.17** 0.22** 0.27** 0.32** 0.17** 0.16** 0 .18 ∗ ∗ 0 .14 ∗ 0 .46 ∗ ∗ 0 .58 ∗ ∗ 0 .68 ∗ ∗ 15 Aggression YSR T3 0.12* -0.02 0.28** 0.25** 0.29** 0.25** 0.38** 0.34** 0 .35 ∗ ∗ 0 .29 ∗ ∗ 0 .68 ∗ ∗ 0 .54 ∗ ∗ 0 .50 ∗ ∗ 0 .40 ∗ ∗ 16 Aggression YSR T4 0.08 -0.01 0.27** 0.28** 0.31** 0.26** 0.39** 0.42** 0 .40 ∗ ∗ 0 .31 ∗ ∗ 0 .60 ∗ ∗ 0 .68 ∗ ∗ 0 .59 ∗ ∗ 0 .49 ∗ ∗ 0 .76 ∗ ∗ 17 Aggression YSR T5 0.14* -0.03 0.24** 0.24** 0.34** 0.24** 0.33** 0.35** 0 .38 ∗ ∗ 0 .29 ∗ ∗ 0 .47 ∗ ∗ 0 .51 ∗ ∗ 0 .68 ∗ ∗ 0 .52 ∗ ∗ 0 .68 ∗ ∗ 0 .73 ∗ ∗ 18 Aggression YSR T6 0.00 0.03 0.12* 0.10 0.17** 0.13* 0.21** 0.22** 0 .29 ∗ ∗ 0 .25 ∗ ∗ 0 .41 ∗ ∗ 0 .45 ∗ ∗ 0 .52 ∗ ∗ 0 .64 ∗ ∗ 0 .60 ∗ ∗ 0 .65 ∗ ∗ 0 .73 ∗ ∗ Note. CAR AUCg = Cortisol Awakening Response Area Under the Curve with respect to ground. NB = Neighbourhood. CBCL = Child Behaviour Checklist. YSR = Youth Self Report. *p < 0.05; ** p < 0.01.

(7)

aggression.Neighbourhooddensitydidnotpredictthe

develop-ment (i.e.,intercept and slope)of parent-reported delinquency

(N=358, B=0.04, ␤=0.02, p=0.64 andN=358, B=0.02,␤=0.03,

p=0.67,respectively)andaggression(N=358,B=-0.17,␤=-0.03,

p=0.49andN=358,B=0.06,␤=0.06,p=0.46,respectively).

More-over,CARAUCgdidnothavemaineffectsontheinterceptandslope

ofparent-reporteddelinquency(N=358,B=0.17,␤=0.08,p=0.19

andN=358,B=-0.06,␤=-0.10,p=0.16,respectively),neitherdid

ithavemaineffectsontheinterceptandslopeofparent-reported

aggression(N=358,B=0.13,␤=0.03,p=0.69andN=358,B=0.05,

␤=0.05,p=0.57,respectively).

3.2. MaineffectsofneighbourhooddensityandCARAUCgonthe

developmentofadolescentself-reportedexternalizingproblem

behaviours

Weexaminedthemaineffectsofneighbourhooddensityand

CARAUCgonthedevelopmentofself-reportedexternalizing

prob-lembehaviours,controllingforeffectsofSESandgender.Inthetwo

modelspredictingdevelopmentofself-reporteddelinquencyand

aggression,signiﬁcanteffectsofSESandCARAUCgwereobserved.

Specially,lowerSESwasrelatedtolowerslope(i.e.,changerate)

ofself-reporteddelinquency(N=358,B=-0.46,␤=-0.18,p=0.01).

In addition,lowerSES wasrelated tohigher intercept(N=358,

B=3.48,␤=0.20,p=0.00)andlowerslope(N=358,B=-0.85,␤=

-0.22,p=0.00)ofself-reportedaggression.Furthermore,CARAUCg

waspositivelyassociatedwiththeinterceptofadolescent

aggres-sion(N=358,B=0.84,␤=0.16,p=0.02).

No other main effects of CARAUCg on the intercepts and

slopesofadolescentexternalizingproblembehaviourswerefound.

CARAUCgdidnotpredicttheintercept(N=358, B=0.33,␤=0.13,

p=0.06) nor the slope of self-reported delinquency (N=358,

B=-0.12, ␤=-0.15, p=0.09).Moreover,CARAUCg wasnot

signiﬁ-cantly related totheslope of self-reportedaggression (N=358,

B=-0.20, _␤=-0.16, p=0.08). Inaddition, neighbourhooddensity

did not predict the intercept and slope of self-reported

delin-quency(N=358, B=0.13, ␤=0.05, p=0.35 and N=358, B=0.04,

␤=0.05, p=0.35, respectively)and aggression(N=358,B=-0.16,

␤=-0.03, p=0.48 andN=358, B=0.05,␤=0.04, p=0.50,

respec-tively).

3.3. InteractioneffectsbetweenCARAUCgandneighbourhood

densityonthedevelopmentofparent-reportedexternalizing

problembehaviours

Table3presentstheresultsoftheinteractioneffectsbetween

CARAUCgand neighbourhooddensityonparent-reported

adoles-centdelinquencyandaggression,whilecontrollingfortheeffects

ofSESandgender.Theresultsfromtheregressionanalysesshowed

signiﬁcantinteraction effects between CARAUCg and

neighbour-hood density on the intercept of parent-reported delinquency

(N=358,B=0.31,␤=0.15,p=0.01)andparent-reportedaggression

(N=358, B=0.98, ␤=0.19, p=0.01).These signiﬁcantinteraction

effects indicated that the associations between neighbourhood

densityandparent-reporteddelinquentandaggressivebehaviours

differedforadolescentswithvaryinglevelsofCARAUCg.Regionof

signiﬁcancetestsrevealedthatforadolescentswithCARAUCg

lev-els≥0.62SDand≥1.90SDabovethemean,higherneighbourhood

densitysigniﬁcantly predictedhigherlevelsof delinquencyand

aggression,respectively.Remarkably,thereversewastrueforthe

relationbetweenneighbourhooddensityandexternalizing

prob-lembehavioursforadolescentswithCARAUCglowerthancertain

levels.Speciﬁcally,foradolescentswithCARAUCglevels≤-0.97SD

and_≤-0.09SDbelowthemean,lowerneighbourhooddensity

sig-niﬁcantlypredictedhigherlevelsofdelinquencyandaggression.

Todeterminewhethertheinteractionwasinsupportofthe

biolog-icalsensitivitytheory,wealsoprobedthisinteractionforvarying

levelsofneighbourhooddensity.Wefoundthatforadolescents

liv-inginaneighbourhoodwithdensitylevels≥0.27SDand≥0.85SD

abovethemean,higherCARAUCgsigniﬁcantlypredictedhigher

lev-elsofdelinquencyandaggression,respectively.Thereversewas

truefortherelationshipbetweenCARAUCgandexternalizing

prob-lembehaviours, foradolescentsliving ina neighbourhoodwith

density levels≤-2.57SDand≤-0.96 SDbelow themean,lower

CARAUCgsigniﬁcantlypredictedhigherlevelsofdelinquencyand

aggression,respectively.Theinteractiveeffectsarevisualizedby

showingsimple slopes for adolescentshigh(at1 SDabove the

mean)andlow (at1SDbelowthemean)inCARAUCginFigs.1a

and2a,andforadolescentswholiveinhigh(at1SDabovethe

mean)andlow(at1SDbelowthemean)densityneighbourhood

inFigs.1band2b.Furtherthree-wayinteractionanalysisdidnot

indicatethatthesesigniﬁcantinteractioneffectsweredifferentby

SESlevels.

3.4. InteractioneffectsbetweenCARAUCgandneighbourhood

densityonthedevelopmentofself-reportedexternalizingproblem

behaviours

Table4presentstheresultsoftheinteractioneffectsbetween

CARAUCgandneighbourhooddensityonself-reportedadolescents’

delinquencyand aggression,whilecontrolling fortheeffects of

SESandgender.Thereweresigniﬁcantinteractioneffectsbetween

CARAUCg and neighbourhooddensity inpredictingtheintercept

(N=358, B=0.30, ␤=0.11, p=0.04) but not the slope (N=358,

B=0.02,␤=0.03,p=0.68)ofself-reporteddelinquency.Therewere

nosigniﬁcantinteractioneffectsbetweenCARAUCgand

neighbour-hooddensityinpredictingtheintercept(N=358,B=0.21,␤=0.04,

p=0.46)neithertheslope (N=358, B=0.05,␤=0.04,p=0.63)of

adolescents’self-reportedaggression.Probing of theinteraction

fordifferentlevelsofCARAUCgrevealedthatforadolescentswith

CARAUCglevels≥0.02SDabovethemean,higherneighbourhood

density signiﬁcantly predicted higher levels of delinquency. To

determinewhethertheinteractionwasinsupportofthebiological

sensitivitytheory,wealsoprobedthisinteractionforvarying

lev-elsofneighbourhooddensity.Wefoundthatforadolescentsliving

inaneighbourhoodwithdensitylevels_≥1.07SDabovethemean,

higherCARAUCglevelspredictedhigherlevelsofself-reported

delin-quentbehaviours.Theinteractiveeffectsarevisualizedbyshowing

simpleslopesforadolescentshigh(at1SDabovethemean)andlow

(at1SDbelowthemean)inCARAUCginFig.3a,andforadolescents

wholiveinhigh(at1SDabovethemean)andlow(at1SDbelowthe

mean)densityneighbourhoodinFig.3b.Furtherthree-way

inter-actionanalysesdidnotindicateSESdifferencesinthesesigniﬁcant

interactioneffects.

3.5. InteractioneffectsbetweenSESandneighbourhooddensity

onthedevelopmentofexternalizingproblembehaviours

To test whether the moderating effects of CARAUCg can be

explainedbythepotentialoverlapbetweenneighbourhooddensity

andSES,weexaminedinteractioneffectsbetweenSESandCARAUCg

inpredictingadolescentexternalizingproblembehaviours.Infour

statisticalmodelspredictingeightoutcomes(i.e.,fourintercepts

andfourslopesofexternalizingproblembehaviours),theonly

sig-niﬁcantCARAUCgby SESinteractioneffectwasin predictingthe

slopeofself-reporteddelinquency(i.e.,B[SE]=0.34[0.16],␤=0.14,

p=0.03).ThisinteractionshowedthatforadolescentswithCARAUCg

levels≤0.20SDbelowthemean,lowerSESsigniﬁcantlypredicted

higherslope(i.e.,fastergrowth)ofdelinquentbehaviours.Noother

(8)

Table3

InteractionEffectsbetweenAdolescentCARAUCgandNeighbourhoodDensityonParent-ReportedExternalizingProblemBehavioursfrom15to18Years.

(N=358) Delinquency(CBCL) Aggression(CBCL)

Intercept Slope Intercept Slope

B(SE) ␤ B(SE) ␤ B(SE) ␤ B(SE) ␤

Model1

SES(Lvs.H&M) 0.73(0.40) 0.11 0.25(0.15) 0.14 2.47(1.20) 0.15* 0.43(0.31) 0.14 Gender(Girlvs.Boy) -0.28(0.23) -0.07 -0.12(0.07) -0.12 -0.41(0.58) -0.04 0.24(0.15) 0.13

CARAUCg 0.17(0.13) 0.08 -0.06(0.04) -0.11 0.13(0.33) 0.03 0.05(0.09) 0.05

NBDensity 0.04(0.04) 0.02 0.02(0.04) 0.03 -0.17(0.24) -0.03 0.06(0.08) 0.06

Modelﬁt CFI=0.94,RMSEA=0.07,SRMR=0.05 CFI=0.98,RMSEA=0.06,SRMR=0.05 Model2

SES 0.77(0.40) 0.12 0.25(0.15) 0.14 2.61(1.17) 0.16* 0.42(0.31) 0.14

Gender -0.26(0.22) -0.06 -0.13(0.07) -0.12 -0.33(0.57) -0.03 0.23(0.15) 0.12

CARAUCg 0.18(0.12) 0.09 -0.06(0.04) -0.11 0.15(0.31) 0.03 0.05(0.09) 0.05

NBDensity -0.02(0.08) -0.01 0.02(0.04) 0.04 -0.35(0.22) -0.07 0.08(0.07) 0.08

CARAUCg*NBDensity 0.31(0.12) 0.15** -0.02(0.04) -0.03 0.98(0.37) 0.19** -0.10(0.10) -0.11 ModelFit CFI=0.94,RMSEA=0.06,SRMR=0.05 CFI=0.97,RMSEA=0.05,SRMR=0.05

Note:Scoreofdelinquencyandaggressionwasbasedonthemeanscoreofmotherandfatherreports.CBCL=ChildBehaviourChecklist.SES=SocialEconomicStatus.Lvs. H&M=Lowvs.High&Medium.CARAUCg=CortisolAwakeningResponseAreaUndertheCurvewithrespecttoground.NB=Neighbourhood.B=unstandardizedregression coefﬁcient.␤=standardizedregressioncoefﬁcient.*p<0.05;**p<0.01.

0 1 2 3 4

Low NB Density High NB Density

De lin q ue n cy C B C L Low CARaucg High CARaucg 0 1 2 3 4

Low CARaucg High CARaucg

De lin q ue n cy C B C L Low NB Density High NB Density

a

b

Fig.1.InteractionEffectsbetweenAdolescentCARAUCgandNeighbourhoodDensityonDevelopmentofParent-ReportedDelinquency.

3 4 5 6 7 8 9 10 11 12

Aggression CBCL Low CARaucg High CARaucg 3 4 5 6 7 8 9 10 11 12

Aggression CBCL

a

_b

Fig.2. InteractionEffectsbetweenAdolescentCARAUCgandNeighbourhoodDensityonDevelopmentofParent-ReportedAggression.

3.6. InteractioneffectsbetweenCARAUCiandneighbourhood

densityonthedevelopmentofexternalizingproblembehaviours

WeadditionallytestedwhetherthemoderatingeffectsofCAR

werespeciﬁctocortisolactivity(i.e.,AUCg),oralsooccurredfor

cortisol reactivity (i.e., AUCi). Our analyses showed signiﬁcant

CARAUCi byneighbourhooddensityinteractioneffectsin

predict-ingadolescentself-reporteddelinquency(i.e.,B[SE]=-0.21[0.11],

␤=-0.08, p=0.06inpredictingtheintercept;B[SE]=0.08[0.04],

␤=0.11, p=0.03 in predicting the slope) and adolescent

self-reported aggression (i.e.,B [SE]=0.41 [0.19], ␤=- 0.08, p=0.03

(9)

Table4

InteractionEffectsbetweenAdolescentCARAUCgandNeighbourhoodDensityonSelf-ReportedExternalizingProblemBehavioursfrom15to18Years.

(N=358) Delinquency(YSR) Aggression(YSR)

Intercept Slope Intercept Slope

B(SE) ␤ B(SE) ␤ B(SE) ␤ B(SE) ␤

Model1

SES(Lvs.H&M) 1.07(0.60) 0.12 -0.46(0.18) -0.18** 3.48(1.12) 0.20** -0.85(0.29) -0.22** Gender(Girlvs.Boy) -0.52(0.32) -0.10 -0.04(0.11) -0.02 -0.42(0.60) -0.04 0.09(0.17) 0.04

CARAUCg 0.33(0.18) 0.13 -0.12(0.07) -0.15 0.84(0.35) 0.16* -0.20(0.11) -0.16

NBDensity 0.13(0.14) 0.05 0.04(0.05) 0.05 -0.16(0.24) -0.03 0.05(0.07) 0.04

Modelﬁt CFI=0.97,RMSEA=0.05,SRMR=0.04 CFI=0.98,RMSEA=0.04,SRMR=0.03 Model2

SES 1.11(0.59) 0.13 -0.46(0.18) -0.18* 3.50(1.12) 0.20** -0.85(0.29) -0.22**

Gender -0.50(0.32) -0.09 -0.04(0.11) -0.02 -0.41(0.60) -0.04 0.10(0.18) 0.04

CARAUCg 0.34(0.18) 0.13 -0.12(0.07) -0.15 0.84(0.35) 0.16* -0.19(0.11) -0.16

NBDensity 0.08(0.13) 0.03 0.04(0.04) 0.05 -0.20(0.25) -0.04 0.04(0.07) 0.03

CARAUCg*NBDensity 0.30(0.15) 0.11* 0.02(0.05) 0.03 0.21(0.29) 0.04 0.05(0.10) 0.04 ModelFit CFI=0.97,RMSEA=0.04,SRMR=0.04 CFI=0.97,RMSEA=0.04,SRMR=0.04

Note.YSR=YouthSelfReport.SES=SocialEconomicStatus.Lvs.H&M=Lowvs.Medium&High.CARAUCg=CortisolAwakeningResponseAreaUndertheCurvewithrespect toground.NB=Neighbourhood.B=unstandardizedregressioncoefﬁcient.␤=standardizedregressioncoefﬁcient.*p<0.05;**p<0.01.

1 2 3 4 5 6 7

Delinquency YSR Low CARaucg High CARaucg 1 2 3 4 5 6 7

Delinquency YSR

a

_b

Fig.3. InteractionEffectsbetweenAdolescentCARAUCgandNeighbourhoodDensityonDevelopmentofSelf-ReportedDelinquency.

in predicting the slope).The interaction pattern indicated that

foradolescentswithahighCARAUCi,delinquencyandaggression

were overall higher for those from a high density

neighbour-hood.TherewerenosigniﬁcantCARAUCibyneighbourhooddensity

interactioneffectsonparent-reportedexternalizingproblem

beha-viours.

4. Discussion

Thislongitudinalstudyexaminedtheeffectsofneighbourhood

densityandadolescents’morningcortisolactivityonthe

devel-opmentofexternalizingproblembehavioursfrommiddletolate

adolescence. Our results indicated that neither neighbourhood

densitynoradolescents’cortisolactivityingeneraldidhavemain

effects ontheinitiallevelsand developmental changesof

ado-lescent externalizing problembehaviours.However, therewere

signiﬁcantinteraction effects between CARAUCg and

neighbour-hooddensityinpredictingadolescents’aggressionanddelinquency

over a period of four years. Level of CARAUCg moderated the

effectsofneighbourhooddensityonthedevelopmentof

adoles-cents’externalizing problembehaviours. Theseﬁndingsprovide

supportforperson-environmentperspectivesandalsobiological

sensitivitytheoryonthedevelopmentofadolescent

externaliz-ingproblembehaviours(Ellis&Boyce,2011;Magnusson&Stattin,

2006;Susman&Ponirakis,1997).

4.1. Maineffectsofneighbourhooddensityonadolescent

externalizingproblembehaviours

Thelackofmaineffectsofneighbourhooddensityonaggressive

and delinquent behaviours is consistentwith someprior

stud-ies (Gillis, 1974; Reijneveld et al.,2010; Wichstrom, Skogen, &

Osia,1996).However,ﬁndingswerecontradictorytoresultsfrom

otherstudiesreportingpositiveassociations(Hardenetal.,2009;

Schmitt,1957),negativerelations(Bao,Haas,&Pi,2004),and

non-linearlinks(Browningetal.,2010).Theinconsistentﬁndingscould

bedue toneighbourhooddifferencesacrossstudies.Speciﬁcally,

thesestudiesarefromdifferentcountriesordifferentareaswithin

acountry,anditmightbethatindifferentcountriesorareas

liv-inginahighdensityenvironmentimpliesdifferencesintermsof

distributionsofeducationalresources(Boyle,Georgiades,Racine,&

Mustard,2007),incomeequality(Kawachi,2000),orethnic

compo-sition(Semyonov,Gorodzeisky,&Glikman,2012)whicharerelated

toadolescentexternalizingproblembehaviours(MacDonald,Hipp,

&Gill,2013;Schindleretal.,2015;Wilkinson&Pickett,2009).

Theinconsistencycouldalsoreﬂectotherdifferencesin

meth-ods amongstudies, suchasthecontrolling variables and study

design.Forinstance,someresearcherscontrolledfor

unemploy-ment,meanincome,andeducationallevel(Reijneveldetal.,2010),

others controlledfor parental divorce(Weenink,2011).

Differ-encein these measuredcovariates as statistical controlsmight

(10)

stud-ies adopted a cross-sectional design which does not provide

informationonthedirectionofeffects.Amoreidealapproachis

toadoptalongitudinaldesignthattakesintoaccountthemobility

ofadolescents(i.e.,movefromareasofhightolowdensity

neigh-bourhoodenvironmentand viceversa).It hasbeenshown that

somesigniﬁcant relations between neighbourhooddensity and

externalizingproblembehavioursdemonstratedincross-sectional

analysesdisappearin longitudinalinvestigations(Hardenet al.,

2009).Furthermore,individualcharacteristicsmightmoderatethe

associationbetweenneighbourhoodenvironmentsand

external-izing problembehaviours (Jaffee et al., 2015; Obradovi ´c, Bush,

Stamperdahl, Adler, & Boyce, 2010). In sum, more research is

neededonfactorsthatqualifytheeffectofneighbourhooddensity

onadolescentexternalizingproblembehaviours.

4.2. MaineffectsofCARAUCgonadolescentexternalizingproblem

behaviours

Thecurrentstudyindicatesthatingeneraltherewereno

sig-niﬁcanteffectsofCARAUCgonthedevelopment(i.e.,initiallevels

anddevelopmentalchanges)ofexternalizingproblembehaviours

acrossmiddletolate adolescence.The onlyexception wasthat

CARAUCgwaspositivelylinkedtotheinterceptofadolescent

self-reportedaggression.Thenon-signiﬁcantlinksbetweenCARAUCg

andexternalizingproblembehavioursfoundinthecurrent

com-munitysampleare inaccordance withﬁndingsfrom studiesof

earlytomiddlenormallydevelopingadolescents(Klimes-Dougan

et al., 2001; Sondeijker et al., 2008). In addition, the lack of

CARAUCg-externalizingproblembehavioursassociationsarein

gen-eralconsistentwitha priormeta-analyticreviewconcludingno

associationsbetween(morning)basalcortisollevelsand

external-izingbehavioursinadolescentsampleswithagesrangedfrom12

to19years(Alinketal.,2008).

Findings from the current study are, however, inconsistent

withstudiesshowingnegativeassociationsbetweenmorningbasal

cortisollevelsandadolescents’externalizingproblembehaviours

(Pajeretal.,2001;Shirtcliffetal.,2005;Shoaletal.,2003).The

divergentﬁndingsontherelationbetweenCARAUCgand

external-izingbehaviourproblemscouldbeduetostudiesnotaccounting

for issues of sample characteristics such as severity of

exter-nalizing behaviours and comorbidity of traits or internalizing

problembehaviours (Marsmanet al., 2008; Ruttleet al., 2011;

Shirtcliff&Essex,2008).Somestudieshavesuggestedthatthe

neg-ativecortisol-externalizingproblembehavioursassociationmay

becharacteristic for a particularsevere subgrouponly, suchas

adolescents withconduct disorders, oppositional deﬁant

disor-der,oradolescentswithpersistentaggressivebehaviours(Platje

et al.,2013; Shoal et al.,2003). Thisassociation might alsobe

dependentonthecomorbidityof individualtraits(e.g.,

callous-unemotional traits; Hayes & Matthes, 2009) and internalizing

problembehaviours(Shirtcliff&Essex,2008).Hence,further

stud-iesarenecessarytounderstandtherelationshipbetweenCARAUCg

andexternalizingproblembehaviours.

4.3. ModeratingeffectsofCARAUCgontheassociationbetween

neighbourhooddensityanddevelopmentofadolescent

We found signiﬁcant interaction effects between CARAUCg

and neighbourhood density in predicting adolescent

parent-reported aggression and delinquency and self-reported

delin-quency.CARAUCgappearedtobeamoderatorof theassociation

betweenneighbourhood density and adolescents’ externalizing

problem behaviours. In predicting parent-reported aggression

and delinquency, the interaction effects showed a contrastive

pattern—meaningthat the valence of theassociations between

neighbourhood density and adolescent externalizing problem

behaviours runs in the opposite direction at relatively high

versuslowlevelsofthemoderatorCARAUCg(Belsky,

Bakermans-Kranenburg, & Van IJzendoorn, 2007). The contrastive pattern

suggeststhathighCARAUCgcouldbeaprotectivefactorinalow

den-sityenvironmentbutariskfactorinahighdensityenvironment.In

contrast,lowCARAUCgcouldbeaprotectivefactorinahighdensity

environmentbutavulnerablefactorinalowdensityenvironment.

TheseﬁndingssuggestthatneitherhighnorlowCARAUCgisa

pro-tectiveorvulnerablefactorassuch.Instead,itismoreofthematch

andcombinationofthelevelofanindividual’sCARAUCgandthelevel

ofdensityofaneighbourhoodenvironmentthatpredictsthe

devel-opmentofparent-reportedexternalizingproblembehaviours.We

alsoobservedinteractiveeffectsbetweenCARAUCgand

neighbour-hood density in predicting adolescent self-reported delinquent

behaviours.ForadolescentswithlowCARAUCg,therewasno

signif-icantassociationbetweenneighbourhooddensityanddelinquent

behaviours,whereasforthosewithhighCARAUCg,therewere

posi-tiveassociationsbetweenneighbourhooddensityanddelinquency.

Theseﬁndingstogetherprovidesupportforgeneralinteractive

per-spectivesandhormones-contextinteractionstheory,claimingthat

individuals’developmentaloutcomesaredependenton

interac-tiveprocessesbetweenindividuals’characteristicsandcontextual

factors(Magnusson&Stattin,2006;Susman&Ponirakis,1997).

In general, the interactivepatterns between neighbourhood

densityandCARAUCginpredictingparentalreporteddelinquency

andaggression andself-reporteddelinquencysuggestthat high

CARAUCgisasensitivityfactorwhichincreasesthenegativeeffects

of adverseenvironments (i.e.,highdensity neighbourhood)but

enhancesthebeneﬁcialeffectsofpositiveenvironments(i.e.,low

density neighbourhood). Hence, these results support the

bio-logicalsensitivitytheory(Ellis&Boyce,2011).Morespeciﬁcally,

theseﬁndingsindicatethathyper-arousedadolescentsmighthave

highersensitivitytowardsnotonlythestressesthatcouldbe

gen-eratedbyahighdensityenvironmentsuchasovercrowding,noise,

impersonalspace,orlimitedaccessibilitytoresourcesbutalsothe

beneﬁtsthatmightbeprovidedinalowdensityneighbourhood

suchasphysicalspace,quietness,oravailabilityofresources.For

instance,ithasbeensuspectedthataseriousconsequenceof

high-densitylivingmaybearealorperceivedinabilitytocontrolthe

environmentandtoregulateone’ssocialinteractions(Rodin,1976).

Itislikelythathyper-arousedadolescentssensehigherstressof

inabilitytocontroloverwhatgoesoninaneighbourhoodandover

interpersonalchange,thanless-arousedadolescents,thustheyact

moredelinquentandaggressiveinhigh-densityenvironment.In

sum,ﬁndingscorroborateaconceptualizationofstressactivityas

biologicalsensitivitytocontextbyshowingthathighCARAUCgcan

bothhinderandpromoteadaptivefunctioning.

In predicting parent-reported externalizing problem

behaviours, we observed detrimental effects of a

combina-tionofahypo-arousedbiologicalresponse,namelyalowcortisol

awakingresponse,andalowdensityneighbourhoodinpredicting

adolescents’ externalizing problem behaviours. This ﬁnding is

inconsistentwiththetheoryofbiologicalsensitivitytocontextin

whichadolescentswithalowCARAUCgwouldbeexpectedtobe

notinﬂuencedbytheirenvironment.Thisﬁndingmighthowever

providesomesupportforthesensationseekingtheoryofantisocial

behaviours(Raine,1993;Zuckerman&Neeb,1979).Alowdensity

neighbourhood environment is probably not only quieter and

calmer but also provides fewer opportunities for involvement

in interpersonal contacts or social events than a high-density

neighbourhoodenvironment.Hence,suchanenvironmentmight

be boring for hypo-aroused adolescents as they are less likely

tobearousedby theirenvironment. Aslow arousal represents

anaversivephysiologicalcondition,toattainahigherandmore

(11)

stimulation through delinquent and aggressivebehaviours in a

low-densityneighbourhood.

There was no signiﬁcant association between CARAUCg and

neighbourhooddensityinthisstudy,thusitisratherunlikelythat

therewouldbeabidirectionalprocesswherebycertainadolescents

develophigherCARAUCgasaresultoflivinginhighdensity

neigh-bourhoods,orviceversa.Hence,theinteractioneffects between

CARAUCg andneighbourhood densitywerenot likely tobe

con-foundedbyanybidirectionaleffectsbetweentheindependentand

moderatingvariables.

Ourstudy showedsigniﬁcantmain effectsof SESon

aggres-sivebehaviours,withadolescentsfromlowSESfamiliesreporting

higherlevelsofparent-andself-reportedaggression.Theseresults

areconsistentwithﬁndingsfromalargebodyofprevious

stud-ies(seeBradley&Corwyn,2002forareview).However,despite

thesemaineffects,theonlysigniﬁcantSESbyCARAUCginteraction

effectwasontheslopeofself-reporteddelinquency.These

ﬁnd-ingssuggest thatthemoderatingeffects ofCARAUCgonthelink

betweenneighbourhooddensityandexternalizingproblems

can-notbeexplainedbythepotentialoverlapbetweenneighbourhood

densityandSES.Theseresultsfurtherhighlighttheimportanceof

investigatingtheinteractioneffectsbetweenneighbourhood

den-sityandCARAUCg.

4.4. ModeratingroleofCARAUCiontheassociationbetween

neighbourhooddensityanddevelopmentofadolescent

Wefoundsigniﬁcantinteractioneffectsbetweencortisol

reac-tivityCARAUCi neighbourhooddensity inpredictingadolescents’

self-reporteddelinquentandaggressivebehaviours.The

interac-tionpatternsuggeststhatadolescentswithahigherCARAUCi(i.e.,a

biggerchangeincortisollevelspost-awakening)mightbemore

likely tobenegatively inﬂuencedby a highdensity

neighbour-hood and therefore they develop higher externalizing problem

behaviours.Theseﬁndingsingeneralseemtobeconsistentwiththe

interactionpatternsbetweenneighbourhooddensityandcortisol

activityCARAUCginpredictingexternalizingproblembehaviours,

providingfurthersupportforthebiologicalsensitivitytheory(Ellis

&Boyce,2011).

4.5. Limitationsandstrengths

Severallimitationsofthecurrentstudyhavetobenoted.First,

ﬁndingsmightbelimited toyouthsfrommiddletolate

adoles-cence,theageperiodcoveredbythis study.Theeffectsofbasal

cortisollevelsonexternalizing problembehaviourswere found

tobeage dependent(Alink etal., 2008).It ispossible thatthe

interactioneffectsbetweenCARAUCgandneighbourhooddensityin

predictingexternalizingproblembehavioursarealsodependenton

individuals’developmentalstages.Second,thisstudyfocusesona

generalpopulationsample.Thus,resultsreﬂectnormativelevelsof

aggressiveanddelinquentbehaviours.Therefore,theresultscannot

begeneralizedtoclinicallyreferredyouthswithconductdisorders

oroppositionaldisruptivedisorders.Third,inourstudyCARwas

measuredonasingledayinsteadofovermultipledays.Althougha

recentreviewhasshownthatthecortisolawakingresponsehas

relatively highbetween-visit reliabilities (r=0.33–0.67;Golden,

Wand,Malhotra,Kamel,&Horton,2011),factorssuchassubstance

use,timeofawakening,andstressfulexperiencescouldpotentially

affectCAR(e.g.,Williams,Magid,&Steptoe,2005).Inourstudy,

wepartiallyaddressthelimitationofonesinglemeasurementby

testingtheeffectsofsomepotentialcovariatesandincludingthe

signiﬁcantpredictorsinouranalyses.However,tofurtherincrease

thereliabilitiesofCARmeasurement,futurestudiesareencouraged

touserepeatedmeasuresovermultipledays(Hellhammer,Fries,

Schweisthal,Schlotz,Stone,&Hagemann,2007).

This studyhas alsoseveral strengths. It is the ﬁrst attempt

investigatinginteractioneffectsbetweenneighbourhooddensity

andindividuals’CARonthedevelopmentofadolescent

external-izingproblembehaviours.Themultipleinformantapproach(i.e.,

both self- and parent-report) adopted in this study provided a

morecomprehensivepictureoftheinteractionprocessesbetween

neighbourhooddensityandCAR.Moreover,thelongitudinaldesign

allowedustoinvestigatewhethertheinteractionbetween

neigh-bourhood density and CAR predicts changes in externalizing

problembehaviourswithinindividuals.Furthermore,thepresent

studycontrolledforthepropertyvalueoftheneighbourhoodas

wellasfamily SES.Asitispossiblethatneighbourhoods witha

highdensityareonaveragepoorerthanthosewithlowdensity,

controllingforbothfamilySESandneighbourhoodwealthvalue

couldprecludethisselectionbiaswhichcouldobscurethe

indepen-denceoftheeffectsofneighbourhooddensity(Hipp,Butts,Acton,

Nagle,&Boessen,2013).Finally,thefocusonneighbourhood

den-sityratherthangeneralneighbourhooddeprivationprovidesmore

speciﬁcinformationonthecontributionofneighbourhood

charac-teristicsonadolescents’externalizingproblembehaviours,which

iscalledforinarecentreview(VanVuuren,Reijneveld,Vander

Wal,&Verhoeff,2014).

5. Conclusion

The present study revealed signiﬁcant interaction effects

betweenneighbourhooddensityandCARinpredictingadolescent

aggressiveanddelinquentbehaviours.Ingeneralitindicatesthat

individuals’biologicalsensitivitytoenvironmentalcontextplaysan

importantroleinmoderatingtheeffectsofneighbourhooddensity

onadolescents’adjustment. Findingshighlightthecomplexities

thatcontributetoadolescents’externalizingproblembehaviours

andsupportthetheoreticalideaofinteractiveprocessesbetween

environmentalcontextandindividualscharacteristicscontributing

toadolescents’developmentaloutcomes.

Acknowledgements

Data of the RADAR (Research on Adolescent Development

AndRelationships)studywereused.RADARhasbeenﬁnancially

supported by main grants from the Netherlands Organisation

forScientiﬁc Research(GB-MAGW480-03-005,GB-MAGW

480-08-006, CID; 024.001.003) and various other grants from the

NetherlandsOrganisationforScientiﬁcResearch,theVU

Univer-sityAmsterdam,andUtrechtUniversity.Thisresearchisfundedby

theSeedMoneyDOY(DynamicsofYouth),UtrechtUniversity,The

NetherlandsandbytheEuropeanResearchCouncilunderthe

Euro-peanUnion’sSeventhFrameworkProgramme(FP/2007-2013)/ERC

GrantAgreementn.615159(ERCConsolidatorGrant

DEPRIVED-HOODS, Socio-spatial inequality,deprived neighbourhoods, and

neighbourhoodeffects).Thefundersofthisresearchhadnorole

inthestudydesign,analysis,andinterpretationofdata,writingthe

report,orinthedecisiontosubmitthepaperforpublication.

References

Achenbach,T.M.(1991).Manualforthechildbehaviorchecklist/4-18and1991 proﬁle.BurlingtonVT:UniversityofVermont,DepartmentofPsychiatry. Alink,L.R.,VanIJzendoorn,M.H.,Bakermans-Kranenburg,M.J.,Mesman,J.,Juffer,

F.,&Koot,H.M.(2008).Cortisolandexternalizingbehaviorinchildrenand adolescents:Mixedmeta-analyticevidencefortheinverserelationofbasal cortisolandcortisolreactivitywithexternalizingbehavior.Developmental Psychobiology,50,427–450.http://dx.doi.org/10.1002/dev.20300 Bao,W.N.,Haas,A.,&Pi,Y.(2004).Lifestrain,negativeemotions,and