Short and LongTerm Effects of a Novel on Connectivity in the Brain Gregory S - PDF document

Todetermineatimescaleoverwhichconnectivitychangespersist,wemeasuredchangesinresting-stateconnectivityasaresultofreadinganovel.Wechoseanoveloverashortstorybecausethelengthanddepthofthenovelwouldaffordasetofrepeatedengagementswithassociated,uniquestimuli(sectionsofthenovel)setinabroader,con-trolledstimuluscontextthatcouldbeconsumedbetweenseveralscanningperiods.Awithin-subjectsdesignwasse-lectedforthispilotstudybecauseofitssubstantivecontrolofindividualvariability,statisticalpower,andeconomicad-vantagesinthistypeofstudy(Anderson,2001;Shadishetal.,2002).MaterialsandMethodsAtotalof21participantswerestudied.TwowereexcludedfromthefMRIanalyses:oneforinsufcientattendance,andtheotherforimageabnormalities.Beforetheexperiment,par-ticipantswerescreenedforthepresenceofmedicalandpsy-chiatricdiagnoses,andnoneweretakingmedications.Therewere12femaleand9maleparticipantsbetweentheagesof19and27(mean21.5).EmoryUniversity’sInstitutionalReviewBoardapprovedallprocedures,andallparticipantsgavewritteninformedconsent. FIG.1.Designofexperiment(above).Participantsunderwentresting-statefunctionalmagneticresonanceimagingscanson19consecutivedays(blackarrows).Ontheeveningbeforethemiddle9daysofscanning,participantsreadaportionofthe.Themeanarousalratingacrossparticipants(below)showedarisingtrendtowardtheclimaxofthenovel(errorbars1standarderror).EFFECTSOFNOVELONBRAINCONNECTIVITY591 ReadingmaterialEachparticipantwassubjectto19consecutivedays(July18,2011–August5,2011)ofresting-statescansthatconsistedofatotalappointmenttimeoflessthan30minatthesametimeeachday.Therst5daysandlast5dayswere‘‘wash-in’’and‘‘washout’’sessions,respectively.Eachofthemiddle9scanswasprecededbyreadingapproximately1/9thofthenovel(Pompeii:ANovel,byRobertHarris,Fawcett,2003).Thisnovelwaschosenbecauseitwasbasedontrueeventsbutwrittenashistoricalctionandconveyedinaclassicnarra-tivearc(Freytag,1900).Duringthe‘‘washin’’and‘‘wash-out’’sessions,theparticipantsdidnotperformanyothertasksexceptfortheresting-statescan(Fig.1).Foreachoftheother9days,thestorydays,theparticipantsperformedtheresting-statescanaftertakingaquizandself-reportabouttheeffectofthematerialpresentedintheportionofthenovelthatwasassignedforthepreviousnightandin-cludedave-pointratingscaleofhowarousingthereadingwas(seeSupplementaryDataforquizzes;SupplementaryDataareavailableonlineatwww.liebertpub.com/brain).Throughrepeatedscans,eachparticipantservedashisorherowncontroltomeasurechangesinresting-stateconnec-tivityaftertheconsumptionofthenovel.ThescanningwasperformedonaSiemens3TTrio.EachparticipantreceivedonlyoneT1-weightedstructuralimage2600ms,TE3.93ms,ipangle,224256matrix,176sagittalslices,and1mmcubicvoxelsize)throughoutthedurationoftheexperiment.Onefunctionalresting-statescanwasacquiredeachday(223volumes,TR2000ms,TE30ms,ipangle,FOV192mm,6464matrix,33axialslices,and33.5mmresolutionwithanadded10%gapin-direction,resultinginaresolutionof33.85mm).Participantswereinstructedtorestquietlywitheyesclosed.Allofthepreprocessingwasperformedusingthe1000FunctionalConnectomesScriptsavailablefromNITRC(www.nitrc.org).Theonlymodicationtothesescriptswastheadditionofaniterativelooptocyclethroughthe19daysofdata.ThescriptsperformedthefollowingpreprocessingproceduresusingFSL(AnalysisGroup,FMRIB)andAFNI(NIMH).First,theanatomicalimagewasdeobliquedandreorientatedtothecoordinatespacethatiscompatiblewithFSL.Next,theimagewasskullstripped. WashoutVersusWashinConnectionsNodeMNILabelNodeMNILabelwashin(15Lcerebellum13129Lcerebellum130124Cerebellarvermis13129Lcerebellum21Lcerebellum1403330Rcerebellum21Lcerebellum15033Lcerebellum15Lcerebellum15033Lcerebellum15Lcerebellum15133Lcerebellum130124Cerebellarvermis15133Lcerebellum29Lcerebellum1551833Rcerebellumwashin(61734Lmidcingulate430152SMA23027ACC1034711Lmiddletemporal61734Lmidcingulate1034711Lmiddletemporal126Lthalamus1034711Lmiddletemporal5811126Rthalamus1034711Lmiddletemporal141Lputamen1034711Lmiddletemporal430152SMA10921Lcerebellum311Linsula10921Lcerebellum1559Lpre/postcentralg11037Lcerebellum42311Linsula11324Lcerebellum2760Lpre/postcentralg11324Lcerebellum311Linsula12015Lcerebellum1559Lpre/postcentralg12015Lcerebellum2760Lpre/postcentralg12015Lcerebellum311Linsula12134Lcerebellum3101545SMA1282122Rcerebellum430152SMA1282122Rcerebellum311Linsula130124Cerebellarvermis2760Lpre/postcentralg130124Cerebellarvermis311Linsula13129Lcerebellum4430Lsupramarginalg1367241Lprecuneus4711Lmiddletemporal1367241LprecuneusNodeisthenodenumberbasedonthesortingintableS6byDosenbachetal.(2010).Networksignicanceisbasedon5000permutations,correctingforFWER.MNIarecoordinates.LabelisbasedonAFNI‘‘whereami’’functionandCA_ML_18_MNIAatlas.FWER,familywiseerrorrate;MNI,montrealneurologicinstitute;AFNI,analysisoffunctionalneuroimage.BERNSETAL. Theresting-statefunctionalimageswerepreprocessedthroughamulti-stepprocedure.Theimagesweredeobli-quedandreorientedsimilarlytotheanatomicalimages.Ameanfunctionalimagewascomputedtoserveasatargetformotioncorrection.Using3dvolreg,thefunctionalimageswerethenalignedtothemeanimageusingtwo-passFourierinterpolation.TodecreaseedgeartifactsfromFourierinter-polation,azeropadoffourvoxelswasaddedaroundtheedgesandstrippedoffaftermotioncorrection.Theimageswerethenskullstrippedtocreateamaskthatwasthenap-pliedtothemotion-correcteddata.Toallowforfullmagne-tizationandsettlingonanystartleresponsesfromtheonsetofthescanning,theeighthvolumewasusedforregistrationtotheanatomicalimage.Spatialsmoothingwasperformedusinga6-mmGaussiankernel.Grandmeanscalingwasper-formedwithanintensitynormalizationto10,000.Alow-passlterof0.1Hzandahigh-passlterof0.005Hzwereappliedfortemporalltering.TheimagesweredetrendedbycalculatingthemeanofthetemporallylteredimageanddetrendingwiththeadditionofLegendrepolynomialsofanorderuptoandincludingtwo.Animagethatwastheadditionofboththemeananddetrendedimagewascreated.Threeseparateregistrationalignmentswereperformed.Thefunctional(usingtheeighthimageacquisitionasatem-plate)toanatomicalalignmentwasproducedusingatrilin-earinterpolation(sixdegreesoffreedom).Theanatomicaltostandardbrain(MNI152_T1)wascreatedagainusingatri-linearinterpolation(12degreesoffreedom).Thetransfor-mationmatricesofbothofthesestepsweresaved.Athirdmatrix,fortheabilitytotransformbetweenfunctionaltostandard,wascreatedbyconcatenatingthematricesoftheprevioustwosteps.Theinverseofeachofthesematriceswasalsoproduced,providingtheoptionofregisteringfreelybetweenanycombinationoffunctional,anatomical,orstandardimages.Segmentationwasperformedtocreateindividualimagesforeachtissuetypeandindividualprobabilitymaps.Thetissuetypesrecognizedascerebrospinaluid(CSF)andwhitematter(WM)weremasked.Thesemaskswereusedtocontrolfornuisancesignals.Weutilizedtheglobalsig-nal,WMandCSFsegmentationmasks,andthesixmotionparameterstoadjustthefunctionalsignalsfortheeffectsofphysiologicalnoiseandmotion(Yanetal.,2013).Althoughadjustmentforglobalsignalsiscontroversial,weoptedtotakeaconservativeapproachandcontrolforphysiolog-icalnoise(Foxetal.,2009).Thisapproachmayprotectagainstfalsepositivesbutmayintroducespuriousnegativecorrelations(Murphyetal.,2013;Weissenbacheretal.,2009),soouranalysisfocusedonlyonchangesinpositivecorrelations.Twooftheparticipantswerenotpresentontherstdayofscanning,andathirdwasabsentonthelastdayofscanning.Therefore,the19participantswereanalyzedover17consec-utivedaysofscanning(days2–18).Usingapredenednetworkof160regionsofinterest(ROI)(Dosenbachetal.,2010),weextractedthetimeseriesofeachROIforspheresof6mmradiusforeachpersononeachday.WechosethissetofROIsbecausethenumberofROIsstrikesabalancebetweenareasonablenumberandgoodcorticalcoverage.Becausetheoriginalpaperwasadevelopmentalneurosciencestudy,ourresultsonreadingareparticularlyrelevanttothestudyofcognitivedevelopment.Thisyielded FIG.2.Networksassociatedwithincreasedconnectivityafterthenovel[0.022correctedforfamilywiseerrorrate(FWER)]anddecreasedconnectivity(0.003correctedforFWER).Bothnetworksshowedgenerallymonotonicchangesincorrelationstrengthwithtime,suggestingthatthesechangesmaynotberelateddirectlytothenovelitself.Itisnoteworthythatbothnet-workshavestronghubsinthecerebellum.EFFECTSOFNOVELONBRAINCONNECTIVITY593 afour-dimensionalmatrixofROIperson17).Next,wecomputedthepairwisecross-correlationbetweenROIsforeachscansession(Mackeyetal.,2013),yieldingamatrixwithdimensions16017.WethenappliedtheFisher-transformationtonor-malizethecorrelationcoefcients,whichareboundedStatisticswereperformedusingtheNetwork-BasedStatis-tics(NBS)Connectomev1.2(Zaleskyetal.,2010).Allconnec-tionsinthe-transformedcorrelationmatricesweresubmittedtoaone-sided-testtoseewhichindividualcon-nectionsweresignicantlydifferentbasedonthespeciedcontrast.Thedesignmatrixforthesetestsincludedacolumnforeachdayanddummyvariablesforeachsubjecttocontrolforsubjectwisedifferencesinmeancorrelations.Thus,therewere36columns(17days19subjects).Contrastswerespeciedasvectorsofdifferencesacrossthe17day-columns(seebelow).Duetothelargenumberofelementsinthecor-relationmatrix(12,720uniqueelements),connectionsthat0.001(3.32)insignicancewerethensubmit-tedtoapermutationtesttocontrolforfamilywiseerrorrate.NBSismoresensitivetodetectingnetworksoftopologicallyconnectednodes,whiletherelatedapproach—falsediscov-eryrate—ismoresensitivetostrong,focalconnections.Sinceweassumethatanovelengagesmanyregions,itsef-fectsarelikelytobeextendedoveranetworkofconnectionsratherthanasmallnumberofconnections.Forthisreason,weusedtheNBSintensitystatistic.Foreachcontrast,5000permutationswereperformed.Permutationswererestrictedtobedonewithinsubjectsonly.Threesetsofcontrastswerespecied.First,todeterminewhethertherewereanysignicantchangesinconnectivitybetweenthebeginningandtheendofthestory,weexaminedthecontrasts:[washoutwashin]and[washinSecond,todetermineshort-termchangesinconnectivityasaresultofthestoryandpotentialreactivationduetothedailyquiz,wecontrasted:[storywashout],withappropriateweightingsforthedifferentnumberofdays.Finally,todeterminelong-termchangesthatwererelatedtothestorybutpersistedbeyondthereadingdays,wecon-trasted:[storywashin],againwithappropriateweightingsforthedifferentnumberofdays. StoryVersusNonstoryConnectionsNodeMNILabelNodeMNILabelNetwork1(413Llingualg/hippocampus974430Lsupramarginalg114517Lsupmedialg1045039Langularg138425RACC1045039Langularg17233347Rsupfrontalg1045039Langularg17233347Rsupfrontalg107445247Rangularg4430Lsupramarginalg1085217Lprecuneus5039Langularg1085217Lprecuneus4430Lsupramarginalg111105517Rprecuneus4430Lsupramarginalg1155817Lprecuneus5039Langularg1346940Langularg4430Lsupramarginalg1367241Lprecuneus5039Langularg1367241Lprecuneus4430Lsupramarginalg1417532Lcuneus4430Lsupramarginalg1551833RcerebellumNetwork2(430152SMA823716Lsuptempg70422417Rsuptempg86343965Rpostcentralg3716Lsuptempg129191Rlingualg1850Lpostcentralg1457633Lsupoccipitalg/cuneus3716Lsuptempg1457633Lsupoccipitalg/cuneus89584120Rsuptemp/supramarginalg1457633Lsupoccipitalg/cuneus5146824Rpostcentralg/insula148157732Rcuneus70422417Rsuptempg/insula148157732Rcuneus89584120Rsuptemp/supramarginalg148157732Rcuneus3716Lsuptempg1562LinfoccipitalgNetwork3(6059138Rsuptempg68229Lsuptempg229Lsuptempg773064Lpostcentralg649Lpostcentralg12346625Rmiddletempg1559Lprecentralg12346625Rmiddletempg229Lsuptempg12346625Rmiddletempg69412355Rpre/postcentralg12346625Rmiddletempg2760Lpre/postcentralg12346625Rmiddletempg2760Lpre/postcentralg135397113Rmiddletemp/occipitalgNodeisthenodenumberbasedonthesortingintableS6byDosenbachetal.(2010).Networksignicanceisbasedon5000permutations,correctingforFWER.MNIarecoordinates.LabelisbasedonAFNI‘‘whereami’’functionandCA_ML_18_MNIAatlas.BERNSETAL. Networkswerevisualizedbydisplayingnodesandconnec-tionsinBrainNetViewer(www.nitrc.org/projects/bnv/).Timeseriesforeachnetworkwerecomputedbyaveragingthecorrelationcoefcientsineachconnectionofthenetworkforeachsubjectoneachdayandthencomputingthemeanandstandarderroracrosssubjectsforeachday.Consistentwiththeoriesofplotstructure,themeanarousalratingsofthestoryroseconsistentlythroughoutthestoryandculminatedwiththeclimax—theeruptionofthevolcanoandthedestructionofPompeii(Fig.1).Fortherstsetofcontrasts,[washoutwashin]showedpositivecorrelationsthatchangedsignicantlybetweenthebeginningandtheendofthestory.Thiswasasmallnetworkofeightnodesandeightconnections,allinthecerebellum(Table1andFig.2).Thetimeseriesofthisnetworkshowedbothamonotonictrendthroughoutmostdays,butinterest-ingly,thelargestincreaseincorrelationwasaftertherstnight’sreading.Duringthestorydays,thecorrelationuctu-ated,butnotbelowpre-storylevels,androseonthelaststorydaywithacontinuedriseafterthestory.Theoppositecon-trast,[washinwashout]revealedaslowdeclineincorrela-tionswithinanetworkbetweentheleftcerebellumandleftpre/postcentralgyrus.Thesecorrelationsweregenerallylow(0.1)anddecreasedto0.05bytheendoftheexperi-ment.Giventhesemonotonictrendsandpreponderanceofconnectivitychangeswithincerebellarregions,wedonotconsiderthesechangesrelatedtothestory.Toisolatetheshort-termchangesassociatedwithreadingthestory,wecombinedthewashinandwashoutperiodsandcontrastedthemwiththestorydays.Thiscontrastiden-tiedthreeindependentnetworksthathadsignicantin-creasesinconnectivityduringthestorydays(Table2).Network1hadaprominenthubaroundtheleftangularandsupramarginalgyriwithconnectionstoboththeprecu-neusandmedialfrontallobe(Fig.3).Therewasalsoasignif-icantconnectiontotheleftlingualgyrusinthevicinityofthehippocampus.Thetimecourseofcorrelationswithinthisnet-workshowedastrikingpatternofasharpriseontherststoryday,reachingitspeakonthelaststoryday,followedbyanonlineardecay.Network2wasabilaterallydistributednetworkwithoutprominenthubs(Fig.4).Signicantconnec-tionsweremostlyposteriorandlocatedinthesuperiortem-poralgyriandcuneus(Table2).Thetimecoursewasnotasclearlyrelatedtothestory,withcorrelationspeakingonthesecondstorydayandthendeclining.ThecorrelationswerealsolowerthaninNetwork1.Finally,Network3hadsignif-icantconnectionsbetweenahubintherightmiddletemporalgyrusandtheleftpre/postcentralgyrusandleftsuperiortemporalgyrus(Fig.5).Thetimecourseofcorrelationsof FIG.3.Network1(correctedforFWER)ofnodesandconnectionswithsigni-cantlyincreasedcorrelationduringstoryversusnonstorydays.Thisnetworkwascon-centratedinahubaroundtheleftangularandsupramargi-nalgyri,withconnectionstomedialprefrontalcortex.Thetimecourseofcorrelationsacrossdaysshowedasharprisebeginningontherstpost-storydayandadecayaftertheendofthenovel.EFFECTSOFNOVELONBRAINCONNECTIVITY595 thisnetworkshowedthesamestrikingincreasewiththeonsetofthestory.UnlikeNetwork1,thisnetworkdidnothaveanonlineardecay.ThemagnitudeofcorrelationsweresimilartothatofNetwork1.Toidentifypotentiallong-termchangesinconnectivity,wecontrasted[storywashin].Thiscontrastidenti-edincreasesincorrelationduringstorydaysandpersistedduringthewashoutperiod.Onenetworkwasidentiedwithbilateralconnectionsbetweenpre/postcentralgyri,middleandsuperiortemporalgyri,andinsula(Table3andFig.6).Thetimecourseofcorrelationsshowedtheincreaseoc-curringwiththeonsetofthestory,peakingonthesixthoreighthstoryday,anddecliningslightlyafterward.Allthecor-relationsduringthewashoutdayswerehigherthanthewashinperiod(Fig.6).BeforeinterpretingthechangesinRSNs,itisworthexam-iningtherepeatabilityofresting-statescans.Previousworkhasshownthatthreeresting-statescans—twowithinanhourandone5–16monthslater—demonstratedamodesttohighdegreeofrepeatabilityinthespatialcomponentsidentiedthroughtemporalconcatenationindependentcom-ponentanalysis(ICA)aswellastargetedROIs(Damoiseauxetal.,2006;Shehzadetal.,2009;Zuoetal.,2010),especiallyifsignalsfromtheCSFwereregressedoutasnuisancevariables(ChangandGlover,2009;Lietal.,2012).Similartest–retestre-liabilityfromtemporalconcatenationICAwasobtainedinolderadultsscannedtwice,1yearapart(Guoetal.,2012).Measuresoftheoreticalgraphconnectivityshowedmoderatetest–retestreliability,dependingontemporallterparame-ters(Braunetal.,2012).Therefore,RSNsareaviablemeasureofbrainnetworkreorganizationduetoasalientexperience,asthesenetworksappearrelativelystableandreliableacrosstimeintheabsenceofsignicantevents.Thisraisestheques-tionofwhetherreadinganovelissufcientlypowerfultocauseadetectablereorganizationofcorticalnetworks.Thetimescaleoftheeffectofanovelmaybebothshortandlongterm.Short-termeffectsmightbeobservedimmediatelyafterreading.Forexample,RSNsareknowntobealteredbyrecentlanguagecomprehensiontasks(Hassonetal.,2009)aswellasvisualcategorizationtasks(Stevensetal.,2010).Althoughthechapterreadingswereperformedduringtheeveningsbeforescans,thequizzesoccurredjustbeforethescan.Thequizzes,therefore,mightberesponsibleforsuchimmediatechangesinrestingstate,thoughthetasksdifferintheirorientation.Theprimary(evening)taskinvolvedac-tiveconsumptionofthestory,whilethenextmorningthequiztaskinvolvedreectionofthestory,wherethelatter FIG.4.Network2(correctedforFWER)ofnodesandconnectionswithsigni-cantlyincreasedcorrelationduringstoryversusnonstorydays.Thissparsenetworkwaslocatedinposteriortem-poralgyriwithconnectionstothecuneus.Thetimecourseofcorrelationsacrossdaysshowedarisebeginningontherstpost-storyandpeak-ingonthesecondstoryday,followedbyadeclineBERNSETAL. taskwouldlikelyengageinareasassociatedwithautobio-graphicalrecall.Themagnitudeofthearousalscorewasnotsignicantlycorrelatedwiththechangesinconnectivity.Evenso,recentevidencesuggeststhatresting-statechangespersistforadayafteracognitiveintervention,suchasneuro-feedbackwitha‘‘focusing’’effectonlociofactivity(Harme-lechetal.,2013).Thus,althoughthereadingquizmaybepartlyresponsibleforshort-termchangesinRSNs,thereisnowevidencethatthesechangesmayalsobeduetocarry-overfromthepreviousevening.Consideringboththeeveningcarryoverandthequizreac-tivationasshort-termeffects,threeindependentcorticalnet-worksdemonstratedincreasesinconnectivityasaresultofthenovel.Network1(Fig.3)displayedthestrongestlocaliza-tiontoahubcenteredaroundtheleftangularandsupramar-ginalgyruswithincreasesinconnectivitytoboththemedialprefrontalcortex(MPFC)andcuneus.Thetimecourseofcor-relationwithinthisnetworkdisplayedanonlineardecaydur-ingthewashoutperiodthatisconsistentwithalingering,butdecayingeffect.Thisdecaysuggeststhatthechangeswerenotsolelyduetothequiz,whichwouldhavehadmoreofanon–offeffectasseeninNetwork3(Fig.5).ThenodalcenteraroundtheleftangulargyrusinNetwork1isconsistentwiththisregion’swell-knownroleinlanguagecomprehension.Arecentmeta-analysisoftheoryofmindstudiesidentiedtheleftangulargyrusasthethirdmostlikelyregiontobeacti-vatedifthetaskwasstorybased(behindrightangulargyrusandMPFC—bothofwhichalsoappearinNetwork1)(Mar,2011).Thus,theimplicationisthattheactivationoftheseregionsduringtheeveningreadingcarriedovertothenextmorningaschangesinconnectivity.Oneexplanationexpandsontheconceptofrestingstateasadynamicorganizationalconstruct(Decoetal.,2011).These‘‘resting-state’’networksmarshalthebrain’searlierfunc-tionalengagementofthenovel,particularlythemultimodalassociativeregionsaroundthetemporoparietaljunction(TPJ)(angularandsupramarginalgyri,andmiddleandsupe-riortemporalgyri).Thisinterpretationrestsontheprinciplethatthebrainisapredictionengine.Thatis,therestingstateofthebrainisbestviewedasbeingina‘‘constantinnerstateofexploration,inwhichthebraingeneratespre-dictionsaboutthelikelynetworkcongurationthatwouldbeoptimalforagivenimpendinginput’’(Decoetal.,2011).Accordingly,sinceearliercognitiveexperiencesmaymodu-latetheresting-stateconnectivitymaps,thetaskofreadingconditionallyalteredtherestingstateofourpartici-pantswithabiastowardahybridmentalizing-narrativenet-workcongurationeventhoughtheywerenotactivelyengagedinatask.Couldthesespecicneuraleffectsbemereconsequencesnotrelatedtostoryconsumption?Varianceattributedtothestoryconsumption‘‘treatment’’certainlycontainsanerror FIG.5.Network3(correctedforFWER)ofnodesandconnectionswithsigni-cantlyincreasedcorrelationduringstoryversusnonstorydays.Thissparsenetworkwaslocatedinposteriortem-poralgyriwithconnectionstothecentralsulcus.Thetime-courseofcorrelationsacrossdaysshowedasharprisebe-ginningontherstpost-storythatwassustainedatarela-tivelyconstantlevelthrough-outthestory,followedbyasharpdeclinepost-story.EFFECTSOFNOVELONBRAINCONNECTIVITY597 term,buttheresultantnetworkcomponentsidentiedaredistinctlytaskspecicandrevealedacrossparticipantsrisingabovetheheterogeneousinuencesoftheir‘‘off-task’’(read-ingdays)aswellas‘‘pre-task’’(washinperiod)experientialmilieu.Itis,however,possible,thattheobservedchangesinconnectivitycouldbeduetotheoverallexperimentalcontextofbeingscannedafterreadingchaptersfromanovel—theex-perimentitselftriggersanactiveprocessofrememberingthepreviousnight’sreading,whichwasfurtherprimedbyaquiz.Evenso,thefactthatreadinganovelcausedchangesincorticalconnectivityplacesaboundonthestabilityofRSNs.Whilelargelystable,therestingstateshouldproperlybeconceivedofasquasi-staticandsubjecttobothshort-andlong-termdynamicrecongurations.Longer-termchangesinconnectivitywereidentiedbythecontrastof[storywashin].Thiscontrastidenti-edconnectionsthatincreasedinstrengthduringthestorydaysandremainedelevatedafterthenovel(Fig.6).Thisnet-workwasheavilyconcentratedaroundthecentralsulcusbi-laterallywithconnectionstobilateralposteriortemporalgyriandinsula.ThisnetworkcorrespondscloselytoapreviouslyidentiedRSNcomprisingsomatosensoryandmotorregions(DeLucaetal.,2006).Correlateductuationsinmotorcortexarewellknownandmaynotbeduetoaspeciccognitiveprocess(Biswaletal.,1995;Xiongetal.,1999).Thus,weareleftwiththequestionastowhythesecorrelationsincreasedwiththeonsetofthenovel.Onepossibilityforincreasesinsomatosensorycortexcon-nectivityisthatreadinganovelinvokesneuralactivitythatisassociatedwithbodilysensations.Thisiscalledthetheoryof‘‘embodiedsemantics’’(Aziz-ZadehandDamasio,2008).Somatosensorycortexactivationhasbeenpreviouslydemon-stratedbythereadingofmetaphors,especiallyiftheyaretac-tilemetaphors(Laceyetal.,2012).Itisplausiblethattheactofreadinganovelplacesthereaderinthebodyoftheprotago-nist,whichmayaltersomatosensoryandmotorcortexcon-nectivity.Itisinterestingtonote,however,thattheregionspreviouslycalledthe‘‘protagonistnetwork’’—dorsomedialPFCandrightTPJ(MasonandJust,2009)—constituteaverydifferentnetworkthanthesomatosensoryregionsandbearmoresimilaritytothenetworksweidentiedashavingshort-termchanges.Thisnetworkcanbecomparedwiththepredominatelycerebellarnetworkidentiedbythe[wash-washin]contrast(Fig.2).Althoughthelatterrepresentsthemostobviouschangefrombeforeandafterthenovel,thetimecourseislargelymonotonicandnotclearlyrelatedtotheperse.However,itispossiblethatboththecerebellarchangesandthesomatosensorychangesreectchangesinmotorcontrolrelatedtotheactofreading.Suchprocessesmightrelatetooculomotorcoordinationandattention,forex-ample,andhavenothingtodowiththecontentofthenovel.Insummary,wehavedemonstratedthatacrossthelikelyarrayofdiverseexperiencesencounteredbyourparticipants,therewasadetectableandsignicantcommonalterationof WashoutVersusWashinConnectionsNodeMNILabelNodeMNILabelwashin(1834327Rinffrontalg/insula4243112Rinsula53441138Rpre/postcentralg621559Lprecentralg4243112Rinsula69412355Rpre/postcentralg430152SMA69412355Rpre/postcentralg649Lpostcentralg70422417Rsuptempg923Lpre/postcentralg74182762Rpre/postcentralg5146824Rpre/postcentralg752760Lpre/postcentralg70422417Rsuptempg752760Lpre/postcentralg1236Lpre/postcentralg76289Linsula74182762Rpre/postcentralg76289Linsula1559Lprecentralg773064Lpostcentralg1850Lpostcentralg773064Lpostcentralg69412355Rpre/postcentralg773064Lpostcentralg1559Lprecentralg7851305Rsup/midtempg70422417Rsuptempg7851305Rsup/midtempg74182762Rpre/postcentralg7851305Rsup/midtempg289Linsula823716Lsuptempg1850Lpostcentralg833713Lsup/midtempg4243112Rinsula9543438Rsup/midtempg3716Lsuptempg1185Linfoccipitalg649Lpre/postcentralg12346625Rmiddletempg923Lpre/postcentralg12346625Rmiddletempg65462045Rpre/postcentralg12346625Rmiddletempg69412355Rpre/postcentralg12346625Rmiddletempg1850Lpostcentralg1457633Lsupoccipitalg/cuneus65462045Rpre/postcentralg1457633Lsupoccipitalg/cuneus9543438Rsup/midtempg1457633Lsupoccipitalg/cuneus3716Lsuptempg1562LinfoccipitalgNodeisthenodenumberbasedonthesortingintableS6byDosenbachetal.(2010).Networksignicanceisbasedon5000permutations,correctingforFWER.MNIarecoordinates.LabelisbasedonAFNI‘‘whereami’’functionandCA_ML_18_MNIAatlas.BERNSETAL. theirRSNassociatedwithreadingsectionsofanovelthepre-viousevening.Moreover,thesechangescouldbesegregatedintonetworksassociatedwithshort-termchangesoriginatingneartheleftangulargyrusandlong-termchangesdispersedbilaterallyinsomatosensorycortex.Itremainsanopenques-tionforfurtherstudyastohowlastingtheseeffectsare,butourresultssuggestapotentialmechanismbywhichreadingstoriesnotonlystrengthenlanguageprocessingregionsbutalsoaffecttheindividualthroughembodiedsemanticsinsen-sorimotorregions.ThisresearchwassupportedbyagrantfromDARPA(D11AP00289).Theviews,opinions,and/orndingscon-tainedinthisarticlearethoseoftheauthorsandshouldnotbeinterpretedasrepresentingtheofcialviewsorpolicies,eitherexpressedorimplied,oftheDefenseAdvancedResearchProjectsAgencyortheDepartmentofDefense.ThisworkwasapprovedforPublicRelease,DistributionUnlimited.TheauthorsaregratefulfortheinputofAndrewBrooks,LisaLaViers,W.GavinEkins,andTingLi.AuthorDisclosureStatementNocompetingnancialinterestsexist.AbbottHP.2008.TheCambridgeIntroductiontoNarrative.2ndEd.,Cambridge:CambridgeUniversityPress.AndersonNH.2001.EmpiricalDirectioninDesignandAnalysisMahwah,NJ:LawrenceErlbaumAssociates.Aziz-ZadehL,DamasioAR.2008.Embodiedsemanticsforac-tions:ndingsfromfunctionalbrainimaging.JPhysiolParis102:35–39.BiswalB,YetkinFZ,HaughtonVM,HydeJS.1995.Functionalconnectivityinthemotorcortexofrestinghumanbrainusingecho-planarMRI.MagnResonMed34:537–541.BiswalBB,etal.2010.Towarddiscoveryscienceofhumanbrainfunction.ProcNatlAcadSciUSA107:4734–4739.BraunU,PlichtaMM,EsslingerC,SauerC,HaddadL,GrimmO,MierD,MohnkeS,HeinzA,ErkS,WalterH,SeiferthN,KirschP,Meyer-LindenbergA.2012.Test–retestreliabilityofresting-stateconnectivitynetworkcharacteristicsusingfMRIandgraphtheoreticalmeasures.Neuroimage59:1404–ChangC,GloverGH.2009.Effectsofmodel-basedphysiologicalnoisecorrectionondefaultmodenetworkanti-correlationsandcorrelations.Neuroimage47:1448–1459.DamoiseauxJS,RomboutsSARB,BarkhofF,ScheltensP,StamCJ,SmithSM,BeckmannCF.2006.Consistentresting-statenetworksacrosshealthysubjects.ProcNatlAcadSciUSA FIG.6.Network(correctedforFWER)ofnodesandconnectionswithsigni-cantlyincreasedcorrelationduringstorydaysandthatpersistedbeyondthestory.Thisnetworkwaslocatedbi-laterallyaroundthecentralsulcuswithsparseconnec-tionstotheinsulaandoccipi-talregions.Thetimecourseofcorrelationsacrossdaysshowedagradualrisebegin-ningontherstpost-storydaythatwassustainedbe-yondtheendofthenovel.EFFECTSOFNOVELONBRAINCONNECTIVITY599 DeLucaM,BeckmannCF,DeStefanoN,MatthewsPM,SmithSM.2006.fMRIrestingstatenetworksdenedistinctmodesoflong-distanceinteractionsinthehumanbrain.NeuroimageDecoG,JirsaVK,McintoshAR.2011.Emergingconceptsforthedynamicalorganizationofresting-stateactivityinthebrain.NatRevNeurosci12:43–56.DosenbachNUF,NardosB,CohenAL,FairDA,PowerJD,ChurchJA,NelsonSM,WigGS,VogelAC,Lessov-SchlaggarCN,BarnesKA,DubisJW,FeczkoE,CoalsonRS,PruettJR,BarchDM,PetersenSE,SchlaggarBL.2010.PredictionofindividualbrainmaturityusingfMRI.Science329:1358–1361.FoxMD,ZhangD,SnyderAZ,RaichleME.2009.Theglobalsig-nalandobservedanticorrelatedrestingstatebrainnetworks.JNeurophysiol101:3270–3283.FreytagG.1900.Freytag’sTechniqueoftheDrama.AnExpositionofDramaticCompositionandArt.Chicago:Scott,ForesmanandGuoCC,KurthF,ZhouJ,MayerE,EickhoffSB,KramerJH,See-leyWW.2012.One-yeartest–retestreliabilityofintrinsicconnectivitynetworkfMRIinolderadults.NeuroimageHarmelechT,WertmanE,MalachR.2013.Theday-aftereffect:longterm,Hebbian-likerestructuringofresting-statefMRIpatternsinducedbyasingleepochofcorticalactivation.JNeurosci33:9488–9497.HarrisonBJ,PujolJ,OrtizH,FornitoA,PantelisC,YucelM.2008.Modulationofbrainresting-statenetworksbysadmoodin-duction.PLoSOne13:e1794.HassonU,NusbaumHC,SmallSL.2009.Task-dependentorga-nizationofbrainregionsactiveduringrest.ProcNatlAcadSciUSA106:10841–10846.KellyC,BiswalBB,CraddockRC,CastellanosFX,MilhamMP.2012.Characterizingvariationinthefunctionalconnectome:promiseandpitfalls.TrendsCognSci16:181–188.LaceyS,StillaR,SathianK.2012.Metaphoricallyfeeling:compre-hendingtexturalmetaphorsactivatessomatosensorycortex.BrainLang120:416–421.LiZ,KadivarA,PlutaJ,DunlopJ,WangZ.2012.Test–reteststa-bilityanalysisofrestingbrainactivityrevealedbybloodoxy-genlevel-dependentfunctionalMRI.JMagnResonImagingMackeyAP,SingleyATM,BungeSA.2013.Intensivereasoningtrainingalterspatternsofbrainconnectivityatrest.JNeurosciMarRA.(2011).Theneuralbasesofsocialcognitionandstorycomprehension.AnnuRevPsychol62:103–134.MasonRA,JustMA.2009.Theroleofthetheory-of-mindcorticalnetworkinthecomprehensionofnarratives.LangLinguistCompass3:157–174.MurphyK,BirnRM,BandettiniPA.2013.Resting-statefMRIconfoundsandcleanup.Neuroimage80:349–359.RaichleME,MacLeodAM,SnyderAZ,PowersWJ,GusnardDA,ShulmanGL.2001.Adefaultmodeofbrainfunction.ProcNatlAcadSciUSA98:676–682.ShadishWR,CookTD,CampbellDT.2002.ExperimentalandQuasi-ExperimentalDesignsforGeneralizedCausalInferenceBoston:HoughtonMifin.ShehzadZ,KellyAMC,ReissPT,GeeDG,GotimerK,UddinLQ,LeeSH,MarguliesDS,RoyAK,BiswalBB,PetkovaE,Castel-lanosFX,MilhamMP.2009.Therestingbrain:unconstrainedyetreliable.CerebCortex19:2209–2229.StevensWD,BucknerRL,SchacterDL.2010.Correlatedlow-fre-quencyBOLDuctuationsintherestinghumanbrainaremodulatedbyrecentexperienceincategory-preferentialvi-sualregions.CerebCortex20:1997–2006.WeissenbacherA,KasessC,GerstlF,LanzenbergerR,MoserE,WindischbergerC.2009.Correlationsandanticorrelationsinresting-statefunctionalconnectivityMRI:aquantitativecom-parisonofpreprocessingstrategies.Neuroimage47:1408–1416.XiongJ,ParsonsLM,GaoJ-H,FoxPT.(1999).Interregionalcon-nectivitytoprimarymotorcortexrevealedusingMRIrestingstateimages.HumBrainMapp8:151–156.YanCG,CheungB,KellyC,ColcombeS,CraddockRC,DiMar-tinoA,LiQ,ZuoXN,CastellanosFX,MilhamMP.2013.Acomprehensiveassessmentofregionalvariationintheimpactofheadmicromovementsonfunctionalconnectomics.Neuro-image76:183–201.ZaleskyA,FornitoA,BullmoreET.2010.Network-basedstatis-tic:identifyingdifferencesinbrainnetworks.NeuroimageZuoX-N,KellyC,AdelsteinJS,KleinDF,CastellanosFX,MilhamMP.2010.Reliableintrinsicconnectivitynetworks:test–retestevaluationusingICAanddualregressionapproach.Neuro-image49:2163–2177.Addresscorrespondenceto:GregoryS.BernsDepartmentofEconomicsCenterforNeuropolicyEmoryUniversity36EagleRowAtlanta,GA30322BERNSETAL. Short-andLong-TermEffectsofaNovelonConnectivityintheBrainGregoryS.Berns,KristinaBlaine,MichaelJ.Prietula,andBrandonE.PyeWesoughttodeterminewhetherreadinganovelcausesmeasurablechangesinresting-stateconnectivityofthebrainandhowlongthesechangespersist.Incorporatingawithin-subjectsdesign,participantsreceivedresting-statefunctionalmagneticresonanceimagingscanson19consecutivedays.First,baselinerestingstatedatafora‘‘washin’’periodweretakenforeachparticipantfor5days.Forthenext9days,participantsread1/9thofanovel DepartmentofEconomics,CenterforNeuropolicy,EmoryUniversity,Atlanta,Georgia.GoizuetaBusinessSchool,EmoryUniversity,Atlanta,Georgia. *‘‘Story’’and‘‘narrative’’areoftenusedinterchangeably.Technically,