Sie sind auf Seite 1von 8

Human Brain Activity Time-Locked to Narrative Event Boundaries

Author(s): Nicole K. Speer, Jeffrey M. Zacks and Jeremy R. Reynolds


Source: Psychological Science, Vol. 18, No. 5 (May, 2007), pp. 449-455
Published by: Sage Publications, Inc. on behalf of the Association for Psychological Science
Stable URL: http://www.jstor.org/stable/40064637 .
Accessed: 18/06/2014 15:52

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Sage Publications, Inc. and Association for Psychological Science are collaborating with JSTOR to digitize,
preserve and extend access to Psychological Science.

http://www.jstor.org

This content downloaded from 195.34.78.191 on Wed, 18 Jun 2014 15:52:15 PM


All use subject to JSTOR Terms and Conditions
PSYCHOLOGICAL SCIENCE

ResearchArticle

Human Brain ActivityTime-


Locked to Narrative Event
Boundaries
M. Zacks, and JeremyR. Reynolds
Nicole K. Speer,Jeffrey

University
Washington

ABSTRACT - Readers structurenarrativetextintoa series ratedactivitieswhilelisteningto or readingdescriptions of


ofeventsin ordertounderstandand remember thetext.In everyday activities(e.g.,reading abouta personmaking cookies;
thisstudy,subjectsread briefnarrativesdescribingevery- Speer& Zacks,2005; Speer,Zacks,& Reynolds, 2006). These
day activitieswhile brain activitywas recorded with findingsarenotaltogether surprisinggiventheories ofdiscourse
functionalmagneticresonance imaging. Subjects later comprehension, which suggest that readers comprehend text
read thestoriesagain to dividethemintolarge and small as a seriesofmentalmodels(situation models) in order to un-
events.Duringtheinitialreading,pointslateridentified as derstand and remember thesituations described(vanDijk &
boundariesbetweeneventswereassociatedwithtransient Kintsch, 1983).
increasesin activityin a numberof brain regionswhose As eventsstructure visualactivity, situation modelsneces-
activitywas mediatedby changesin the narratedsitua- sarilystructure narrated activity.A readerconstructing a situ-
tion,such as changesin characters9goals. These results ationmodelbeginsby layinga foundation forthe mental
indicatethatthesegmentation of narratedactivitiesinto representation basedontheinitialinformation from thetextand
eventsis a spontaneouspart of reading, and that this thereader'spriorknowledge (Gernsbacher, 1990). If the in-
processofsegmentation is likelydependenton neural re- coming information is consistent withthe model currently under
sponsesto changes in the narratedsituation. development (e.g.,shares charactersand itis
locations), mapped
ontothecurrent model.If theincoming information does not
overlapwiththecurrent model,thereader willshift his orher
Although activityin theworldis presented to observers as a focusofattention tobeginbuilding a newstructure thatsatisfies
stream ofcontinuous information, peopleperceiveactivity as a theconstraints ofthecurrent information (Gernsbacher, 1990;
seriesofdiscrete units,orevents (Newtson,1973;Zacks,Tver- Zwaan& Radvansky, 1998).
sky,& Iyer,2001).Forexample, whenwatching a personmakea Researchon situation-model construction duringnarrative
batchofcookies,an observer might the
perceive activity as a has on behavioral
comprehension dependedprimarily variables,
sequenceofeventssuchas "making thebatter" and"bakingthe
particularlyreadingtime, toinferthat readers updatesituation
cookies."Thisabilitytoperceivecontinuous visualactivityas modelsin
response to narrativecues such as changesin a
discreteeventsis termed eventstructureperception, anditplays character's locationorgoals(e.g.,Zwaan,Magliano, & Graesser,
an important rolein comprehending everyday in the
activities
1995;Zwaan,Radvansky, Hilliard,& Curiel,1998).Therefore,
realworld(Newtson, 1973;Zacks,Tversky, & Iyer,2001). itis notclearwhether thisprocessoccursspontaneously during
Recentevidencesuggests thattheabilitytosegment activity relatively recentresearch
natural readingconditions. However,
intoa seriesof eventsis notconfined to visualperception. on event has demonstrated thatneuroimaging
understanding
like
Readers, viewers, are able to thestructure
identify ofnar- methods areable toobserveevent-segmentation processesin a
relativelynaturalistic setting withoutovert behavioral measure-
NicoleSpeerandJeremy are
Reynolds currently at the of
University et
ment(Zacks,Braver, al., 2001). By applying these event-
Coloradoat Boulder.Addresscorrespondence to NicoleK. Speer, methodsto readingcomprehension, one can
Department CampusBox 345, University
of Psychology, of Colo- segmentation
rado at Boulder,Boulder,CO 80309-0345,e-mail:nicole.speer® measure spontaneous segmentation processes during reading.
colorado.edu. In addition,theories ofsituation-model construction thatdetail

Volume18- Number
5 © 2007 Association
Copyright Science
forPsychological 449

This content downloaded from 195.34.78.191 on Wed, 18 Jun 2014 15:52:15 PM


All use subject to JSTOR Terms and Conditions
Narrative
EventBoundaries

the conceptual featuresreaders use to update theirsituation "The schoolband startedplayingoutsidein thestreet.Ray-
models (Gernsbacher,1990; Zwaan & Radvansky,1998) pro- mond, alongwithseveralother children,
jumpedupimmediately
vide an efficient,theoreticallygroundedmethodforexploring towatchthebandfrom hisseat.Raymond, seemingly satisfied
the influenceofthesefeatureson thissegmentation process. withwhathe hadseen,calmlysatdownfacingthefront."). The
The currentstudytook advantageof neuroimagingmethods fourscenesusedinthecurrent studydescribe Raymond getting
to test two specific predictionsthat arise fromthe literatures upandeatingbreakfast (Waking Up),playing withhisfriendson
on event understandingand reading comprehension:First,if theschoolgrounds (PlayBefore an
School),performingEnglish
readers spontaneouslysegmentthe activityin narrativetexts lessoninschool(ClassWork), andparticipatingina classmusic
into series of events (i.e., create new situationmodels), then All
lesson(MusicLesson). stimuli canbe found on-lineathttp://
therewill be transientincreases in neuralactivityat the points dcl.wustl.edu/stimuli.html.
subjects explicitly identifyas boundaries between events. An LCD projector was usedtoprojectstimuli ontoa screen
Second, ifthisprocess of segmentationis based on changes in positionedattheheadofthescannerbore,andsubjectsviewed
the narratedsituation,then the neural responses to event thestimulithrough a mirror connected to thehead coil. The
boundarieswill be mediatedby cues such as changes in char- stories
werepresented onewordata timein52-point, whitefont
acters,theirlocations,and theirgoals. on a black background, and all wordswerecenteredon the
We testedthese predictionsby measuringneuralactivityus- projectionscreen.Eachwordremained onthescreenfor200 ms
ing functionalmagneticresonance imaging(fMRI) while sub- and was followed by a delayof 150 ms timesthenumber of
jects read narrativesabout everydayactivities.Followingthe syllablesin theword.Stimulus presentationand timing were
initialreadingof the narratives,subjects divided the narrated controlledbyPsyScopesoftware (Cohen,MacWhinney, Flatt,&
activityintomeaningful units.These behavioraldata wereused Provost,1993) running on an ApplePowerMacG4 computer
to identifyregionswhose activityduringthe initialreadingof (Apple,Cupertino, CA). A PsyScopebuttonboxwas used to
the narrativesincreased at points subjects later identifiedas recordresponsesduring thebehavioraltesting session.
eventboundaries(cf.Zacks, Braver,et al., 2001). An additional
set ofanalyses testedthe predictionthatcues such as changes
Tasks and Procedure
in charactersand theirlocations would mediate the relation
betweenbehavior and brain activity(i.e., tested whetherthe Subjectsweregiven10 to 15 minofpracticewiththeword-by-
wordreading procedure priortothefunctional scans.Theythen
locationsofreaders'eventboundariesaccountedforadditional
read each ofthefourstories a separatefMRIscan; scans
in
variance in brain activityonce the effectsof changes in the
narratedsituationhad been takenintoaccount). rangedinlength from 8.5 to10.9min.Theorderofpresentation
ofthestorieswas counterbalanced acrosssubjects.The task
procedure fortheneuroimaging sessionis presented in detail
METHOD elsewhere (Speer,2005).
Subjectsreturned within4 daysofthescanningsessionto
Subjects readthestories againwhilesegmenting theactivityinthestories
Twenty-eight individuals (ages 19-34; 20 women)volunteered
(segmentation task).They were not toldaboutthistask untilthey
toparticipate in thisstudyfora cashpayment. All wereright-
returned afterthescanning session,andtherefore allwerenaive
handednativeEnglishspeakerswithno history oflanguageor
tothesegmentation taskduring theinitialreading ofthestories.
readingdisorders. For5, we haddatafrom onlytwo(n = 1) or In the task, wereasked toidentify
three(n = 4) narratives coarse-segmentation subjects
becauseofequipment error
orsubjects'
thelargestunitsofactivity thatseemednatural andmeaningful,
fatigue.Informed consentwasobtainedin accordancewiththe
andinthefine-segmentation task,theywereaskedtoidentify the
guidelinessetbytheHumanStudiesCommittee at theWash-
smallestunitsofactivity thatseemednaturalandmeaningful.
ington University SchoolofMedicine.
All subjectsperformed bothsegmentation tasks,andtheorder
ofthetaskswascounterbalanced acrosssubjects.Subjectswere
given a practicenarrative toacquaintthemselves withtheseg-
Materials
mentation task and wererequiredto identify at least three
Weused fourscenesfromthebookOneBoy'sDay (Barker&
(coarse-segmentation task) orsix (fine-segmentationtask)unit
Wright, 1951). One Boy'sDay is an observational recordof
boundaries beforeproceeding with theexperiment. orderof
The
theeveryday activitiesofa 7-year-old boy("Raymond Birch") thestorieswasthesameas in the
scanning session.
duringa singleday in the late 1940s. Observersrecorded
Raymond's activities at 1-minintervals;consequently, thereare
notemporal discontinuities andveryfewexplicitcohesioncues NarrativeCoding
(e.g.,"therefore" or "so that")in thenarrative. Although the Eachnarrative wasdividedintoclauses(a clausewasdefined as
bookis observational, thedescriptionsarequitearticulate, and a verbwithitsargument structure). The number ofclauses inthe
readmorelikestoriesthanstep-by-step listsofactivities(e.g., foursceneswasas follows: Waking Up,192;PlayBefore School,

450 Volume
18-Number
5

This content downloaded from 195.34.78.191 on Wed, 18 Jun 2014 15:52:15 PM


All use subject to JSTOR Terms and Conditions
NicoleK. Speer,Jeffrey
M. Zacks, and JeremyR. Reynolds

178;ClassWork, 172;MusicLesson,215.Tworaters codedeach BEHAVIORAL RESULTS


clauseforchangesineachofsixdimensions (mean Cohen's K =
.77; discrepancies wereresolvedbydiscussion): Data from thebehavioral sessionwereanalyzedto ensurethat
subjectsperformed the segmentation task appropriately by
• Although therewerenotemporal changesinthestories, each identifying moreeventboundaries inthefine-segmentation task
clausewascodedforthepresenceorabsenceofa temporal thaninthecoarse-segmentation task.Thenumber ofboundaries
reference (e.g.,"immediately"or"slowly"). identifiedin a narrative was dividedby thetotalnumberof
• Each clausewasalso codedforthepresenceorabsenceofa clausesinthatnarrative tomeasure thelength (inclauses)ofthe
spatialchange,thatis, a changein thenarrative location, unitsidentified.Alpha was set at .05 forthisand all following
suchas moving fromone roomin a houseto another, or a analyses.The averagelengthof theunitsin thecoarse-seg-
changein thelocationsofcharacters, suchas moving from mentation taskwasgreater (M = 15.43,SEM = 1.83)thanthe
onesideofa roomtoanother. averagelengthoftheunitsidentified in thefine-segmentation
• Objectchangeswerecodedanytimea character changedhis task(M = 3.47,SEM = 0.35),t(27)= 7.05,prep > .99,d = 1.38,
orherinteraction withan object(e.g.,Raymond pickedup a indicatingthatsubjectswereperforming thesegmentation tasks
candyEasteregg). appropriately.
• Character changeswerecoded whenever the subjectof a A seriesoflogisticregressions was carriedoutto determine
clausechanged(e.g.,ifRaymond wasthesubjectofclausen, whichsituation changespredicted thelocationsofindividuals'
andSusanwasthesubjectofclausen + 1,clausen + 1 would coarseandfineeventboundaries. Foreachindividual, separate
be codedas havinga character change). regressionsestimated theeffect ofthenarrative variablesonfine
• A clausewascodedas havinga causalchangewhenever the and coarsesegmentation, and theregression coefficients were
activitydescribed intheclausewas not caused
directly byan thenentered intoone-sample ttests,usingsubjectsas a random
activitydescribedin thepreviousclause (e.g.,a character factor. As Figure1 shows,thesituation-change variablesreli-
initiateda newaction). ablypredicted patterns ofsegmentation inboththecoarse-and
• Goalchangeswerecodedwhenever a characterbegana new thefine-grain tasks.In thecoarse-segmentation task,spatial,
goal-directed such as whena characterinitiated object,character,
activity, and
causal, goalchanges were each associated
speaking. with reliableincreases in the likelihood that subjectswould
identifya clause as an event boundary, smallest t(27) = 4.24,
= 0.80. In thefine-segmentation
In addition,clauseswerecodedforterminal Prep> -99,d task,object,
punctuation (e.g.,
character,causal,goal,andtemporal changes reliably increased
periodsandquestion marks)andnonterminal punctuation (e.g.,
commasandsemicolons).

Imaging
Imageswereacquiredon a 3-T SiemensVisionMRI scanner
(Erlangen,Germany). A pillow,washcloths, andtapewereused
tominimize headmovement, andheadphones andearplugs were
usedtominimize noisefrom thescanner. High-resolution (lxl
X 1.25 mm)structural imageswereacquiredusinga sagittal
MP-RAGE(magnetization-prepared rapid-gradient echo) Tl-
weighted sequence.Functional imageswereacquiredusinga
T2*-weighted asymmetric spin-echoecho-planarsequence,
with32 slices(4.0 x 4.0 mmin-plane resolution)acquiredevery
2.048 s. An additional T2-weighted fastturbo spin-echo scan
acquired structural data in the same planes as the functional
scans,inordertofacilitate mapping between thefunctional and
structuraldata. The functional data were preprocessed and
1. Mean odds ratiosforthecoefficientsin theregressionsanalyzing
warpedtoa standard stereotacticspace(Talairach Tournoux, Fig.
&
thebehavioraldata. Theseoddsratiosrepresentthechangein theoddsof
1988).Timing offsetsbetween sliceswerecorrected usingcubic identifying
a boundarygiventhepresenceofeach typeofsituationchange.
splineinterpolation, and slice intensity differences werere- For example,an odds ratioof 1.5 fora situation-change variablewould
moved.Alldatawererealigned withinandacrossrunsforeach indicatethattheoddsofidentifying a boundaryat a givenclausewere1.5
timesgreaterif thatclause containeda changeon thatdimensionthanif
subject,and imageintensity was normalized to a whole-brainthe clause containedno changeon thatdimension.Error bars indicate
modevalueof1,000. 95% confidence intervals.The dashedlinerepresentsthenullhypothesis.

5
18- Number
Volume 451

This content downloaded from 195.34.78.191 on Wed, 18 Jun 2014 15:52:15 PM


All use subject to JSTOR Terms and Conditions
NarrativeEventBoundaries

TABLE 1
RegionsShowingIncreasedActivationin Responseto PerceivedEventBoundaries
Coordinates Brodmann's Size
Region x y z area (voxels)
Rightmiddlefrontalgyms* 26 24 53 6/8 10
Rightanteriormiddletemporal
gyrus* 54 -13 -15 21 30
Rightposteriorsuperior
temporal
gyrus* 46 -57 26 22 13
Rightposteriorcingulate* 14 -54 20 23/31 131
Bilateralprecuneus* 3 -66 37 7 549
Bilateralprecuneus* 2 -49 67 7 16
Leftsubcallosalgyrus -9 17 -12 25 8
Leftposteriorcingulate* -7 -29 31 31 12

Note.Asterisks
indicateregionsthatshowedgreateractivation
forcoarsethanfineboundaries.

thelikelihoodthata clause wouldbe identified


as an event Regionssensitivetothepresenceofaneventboundary included
boundary,smallest = =
£(27) 2.36,pTep .92, d = 0.45. These a in
largeregion posterior medial cortex,includingposterior
resultsdirectly
replicateourpreviousfindings
(Speeret al., cingulatecortexand theprecuneus(Brodmann's areas,BA, 7,
2006).

IMAGING RESULTS

AnalysisofEventPerception
Toidentify brainareaswithtransient changesinactivity atevent
we
boundaries, followed theprocedure used byZacks,Braver,
et al. (2001). Behavioraldata fromthecoarse-and fine-seg-
mentation scanswereusedtoidentify "trials"during thefunc-
tionalscans.Eachtrialconsisted ofa 36.86-s(18-frame) window
aroundtheonsetofa wordthata givenindividual lateridentified
as an eventboundary in thecoarse-orfine-segmentation task.
Foreach subject,generallinearmodels(GLMs)wereused to
producetwo36.86-s timecoursesforeach voxel(one time
courseforeach ofthetwosegmentation conditions). Thirty-six
regressors codedfortheeffects ofinterest (2 segmentation grains
x 18frames), and8 regressors codedfortheeffects ofnointerest
(nuisancevariables;4 codingfordifferences acrossrunsand
4 codingforthelineartrendwithineach run).A voxel-wise
random-effects 2 x 18 analysisofvariance(ANOVA)withgrain
(coarse and fine)and time(Frames1-18) as theindependent
variablesidentified voxelswhoseactivity wasmodulated bythe
presenceofcoarseandfineeventboundaries identifiedduring
thesegmentation tasks.To control themap-wise falsepositive
rate at an alpha level of .05, we convertedthe voxel-wise
ANOVAF maptoz statistics anduseda z threshold of4.50 anda
2. Regionswhoseactivationincreasedin responseto eventbound-
contiguity criterionoffivecontiguous voxels (McAvoy, Ollinger, Fig.
aries. In (a), thetopimagesrepresentinflatedleftand rightlateralviews
& Buckner, 2001). of cortex,and the bottomimagesrepresentthe corresponding inflated
Table 1 liststheeightregionswhoseactivity changedtran- medialviews.The color codingindicatesthe z statisticforeach region.
in responsetoeventboundaries The characteristics of each regionare detailedin Table 1. These images
siently (i.e.,showeda statisti- were created
using the ComputerizedAnatomicalReconstructionand
callysignificant maineffect oftime)duringpassivereadingof EditingToolkit(CARET; Van Essen, 2002a, 2002b; Van Essen et al.,
the stories.Figure2a showsthe locationsof theseregions 2001; Van Essen, Drury,Harwell,& Hanlon, 2005). The graph(b) de-
mappedontothecorticalsurface, andFigure2b showsthemean pictstheaveragetimecourseofactivationacrossall eightregionsshowing
increased activationin response to event boundaries, separatelyfor
timecourseof activation duringthe passive-reading taskin coarse and fineeventboundaries The verticalline representsthe point
responsetocoarseandfineboundaries acrossall eightregions. wherean eventboundarywas identified.

452 Volume
18- Number
5

This content downloaded from 195.34.78.191 on Wed, 18 Jun 2014 15:52:15 PM


All use subject to JSTOR Terms and Conditions
M. Zacks, and JeremyR. Reynolds
NicoleK. Speer,Jeffrey

23,and31).Tworegions inrighttemporal cortex,oneinanterior The coefficients fortheevent-boundary predictorsin each


middletemporal cortex(BA 21) and theotherin posterior su- modelweresubjectedtoa 2 x 18 ANOVA,withGLM(first vs.
periortemporal cortex (BA 22), also showed changes in acti- second)and time (Frames1-18) as the independent variables.
vationin relationto eventboundaries.In frontal cortex,the The majority oftheregionsshoweda statistically significant,
middlefrontal gyrus(BA 6/8)andthesubcallosalgyms(BA 25) Greenhouse-Geisser-corrected (Winer,1971) interaction be-
showedevokedresponses toeventboundaries.Inall regions,the tweenGLM and time,forbothmodelsusingcoarse event
transientchangescorresponded to briefincreasesin activity. boundaries and modelsusingfineeventboundaries, smallest
In sevenregions, thetimecourseofactivation interactedwith F(17, 459) = 2.81,/?= .03 (see Fig.3). Theoneexception was
segmentation grain, such that coarse-grained boundaries the subcallosalgyrus,in whichthe interactionwas not statisti-
evokeda largerresponse thanfine-grained smallest callysignificant
boundaries, forcoarseeventboundaries, F(17,459) = 2.24,
= =
F(17,459) 2.00,/) .01,r\p2 = .07.The oneexception tothis =
p .07, or forfineeventboundaries, F(l 7,459) = 1.90,/)= .13.
patternwas thesubcallosalgyrus, whichfailedto showa sta- Theseresultsdid notchange when thesecondGLMincluded
tistically
significantdifference betweencoarseand fineevent punctuation variables,in additionto the situation-change
boundaries, F(17, 459) = 0.72,p = .79, r}/ = .03. variables.Thus,including thesituation-change variablesinthe
GLM attenuated the relationbetweeneventboundariesand
brainactivityin sevenoftheeightregions.
Sensitivity ofRegionsto SituationChanges
Activityin each oftheregionsthatshowedevokedresponses to DISCUSSION
eventboundarieswas subjectedto an additionalANOVAto
determine whether thesituationchangescouldaccountforany These resultsdemonstrate thatreadersare sensitiveto the
oftheincreasesin activation associatedwithcoarseand fine stmctureofeventsina narrative,
andsuggestthattheperception
eventboundaries. Foreachsubject,twoGLMswerecreatedfor ofthisstructure is based on theprocessingofchangesin the
eachoftheeightregions inTable1,forboththecoarse-andthe narrated situation.A network ofbrainregionsresponded se-
fine-segmentation conditions.
ThefirstGLMcodedforthenui- lectively at eventboundarieswhenpeoplereadabouteveryday
sancevariablesand theeventboundaries evenwhenthereaderswerenotexplicitly
(coarseorfine),and activities, attending
the secondGLM coded forthe nuisancevariables,situation to thoseboundaries. Thisnetwork of
includeda largeportion
changes(coded as assumed responses at
beginning clauseonset; posteriormedial cortex,anteriorand superiortemporalgyri,
Boynton, Engel,Glover, & Heeger,1996),andeventboundaries. and rightmiddlefrontal - a subsetofthe
and subcallosalgyri

Fig. 3. Signal change based on the coefficientsused in the general linear models for a representative region that
== magnetic resonance.
responded to event boundaries (bilateral precuneus, 3, -66, 37- see Table 1). MR

Volume18- Number
5 453

This content downloaded from 195.34.78.191 on Wed, 18 Jun 2014 15:52:15 PM


All use subject to JSTOR Terms and Conditions
Narrative
EventBoundaries

network ofbrainregionsthatrespondstoeventboundarieswhile situation(Gernsbacher,1990; Zacks et al., 2007; Zwaan &


subjectsare viewingmoviesofeverydayevents(Zacks, Braver, Radvansky,1998): The situationchanges not only predicted
et al., 2001), as well as a subsetofthe networkofbrainregions reader'sexplicitperceptionof narrativestructure, but also ac-
thatis moreactive when subjects are readingcoherentstories countedfora portionofthe influenceof narrativestructure on
than when theyare readingunrelatedor weakly related sen- brainactivity.
tences (Ferstl & von Cramon,2001; Huettner,Rosenthal,& The currentresultshave several implicationsfortheoriesof
Hynd,1989; Mason & Just,2004; Robertsonet al., 2000). The discoursecomprehensionand eventunderstanding. First,they
majorityofregionsthatrespondedtonarrativeeventboundaries provide evidence not onlythat readersare able to identifythe
had responses that were modulated by segmentationgrain: structureof narratedactivities,but also that this process of
Coarse boundaries were associated with significantlylarger segmentingcontinuoustextintodiscreteeventsoccurs during
evoked responsesthanfineboundaries.This largerevoked re- normalreading.When readerswere readingthe storiesin the
sponseassociatedwithcoarse boundariesis also observedwhen scanner,theydid notknowtheywouldlaterbe asked tostructure
people viewmoviesofeverydayactivities(Zacks, Braver,et al., the activityin the text,and yeta networkof brainregionsex-
2001). These resultssuggestthatthe regionsidentifiedin this hibitedincreased activationat pointsreaders later identified
studyare integrallyinvolvedin detectingthe structureof nar- as event boundaries.These regionswere a subset of regions
ratedactivities,and thatthisprocess ofsegmentation is partof previouslyimplicatedin readingcomprehension(Ferstl& von
a larger,modality-independent systeminvolvedin the compre- Cramon,2001; Huettneret al., 1989; Mason & Just,2004;
hensionofeverydayactivities(i.e., involvedin comprehending Robertsonet al., 2000), and the currentresultssuggestthat
real-worldvisual eventsas well as narrateddescriptionsofthose these regionsmay be involvedspecificallyin the process of
events). updatingreaders'situationmodels.
The behavioral data indicated that changes in characters, Second, these data suggestthatthe segmentation ofnarrated
theirlocations,theirgoals, theirinteractionswithobjects, the activitiesis drivenin partby changesin thenarratedsituation,
causal relationshipsbetweenclauses, and temporalreferences such as changes in characters'goals. The neural responseto
all increasedthelikelihoodthatsubjectswouldidentify a clause eventboundariesin narrativeswas partiallyaccountedforby
as containingan eventboundary(replicatingSpeer et al., 2006). changesin thenarratedsituation.This resultconvergeswiththe
Theories of discourse comprehension(Zwaan & Radvansky, findingthatthe explicitperceptionofeventstructure in narra-
1998) state thatreaders and viewersconstructmentalmodels tivesis correlatedwithsuch changes(Speer et al., 2006), and is
ofthe currentactivityusing a combinationofinformation from consistentwiththeoriesproposingthattheperceptionofstruc-
the environment and knowledgeabout the typicalstructureof turein discourseis drivenby discrepanciesbetweenincoming
eventsin theworld.As incominginformation aboutthe current information and the currentmentalmodelofthe visual or nar-
eventbecomes less consistentwiththe currentmentalmodel, ratedsituation(Gernsbacher,1990; Zacks et al., 2007; Zwaan &
readers are more likely to update the currentmodel to more Radvansky,1998).
accuratelyrepresentthe stateoftheworld(Gernsbacher,1990; Third,the spatiotemporalpatternof brain activationin re-
Reynolds,Zacks, & Braver,in press;van Dijk & Kintsch,1983; sponse to event boundaries in narrativeswas similarto the
Zacks, Speer, Swallow, Braver,& Reynolds,2007; Zwaan & patternthathas been observedin studies using moviesrather
Radvansky,1998). Therefore,readers may explicitlyperceive than narrated descriptionsof real-worldactivities (Zacks,
thepointswherecharacters,goals,and otherelementschangeas Braver,et al., 2001). In lightofrecentdiscourse-comprehension
eventboundariesbecause these changes drivethe updatingof theoriesproposingthatreaderssimulateperceptualand motor
readers'mentalmodels at thosepoints. experiences during text comprehension(Barsalou, 1999;
The designofthecurrentstudyallowedus totestthisclaim. If Glenberg,1997; Pulvermuller, 2005; Zwaan,2004), we believe
readers'segmentationof continuoustextis dependenton per- this similaritybetweenprocessingof visual and narratedac-
ceivingthatincominginformation is no longerconsistentwith tivitiesmaybe morethanmerecoincidence,and mayreflectthe
thecurrentmentalmodel,thenchangesin variousaspects ofthe existenceofa generalnetwork forunderstanding eventstructure
narratedsituationwill mediatethe relationbetweenbrain ac- (Zacks et al., 2007). Future studies will need to address the
tivityand eventboundaries,leading to a severe attenuationof relationbetween processingof visual and narratedactivities
the relationbetweeneventboundariesand brain activityonce moreconcretely.
theseconceptualcues are takenintoaccount.This attenuation is Finally,this studydemonstratesthe feasibilityof studying
preciselywhatwas observed:The relationshipbetweenevent discoursecomprehensionin vivo,withoutinterrupting thenor-
boundariesand brain activationin the majorityof regionsre- mal processofreadingbyrequiringovertresponsesto measure
sponsive to event boundarieswas significantly reduced once cognitiveprocesses.Usingbrainactivityas a covertmeasureof
changesin thenarratedsituationweretakenintoaccount.These the cognitiveprocess of segmentingnarrativetextintoevents,
resultsgive furtherweightto theoriessuggestingthatreaders we testedtwohypothesesabout discoursecomprehensionin a
segmentnarrativetexton the basis of changes in the narrated relativelynaturalisticreading situation.These data not only

454 Volume
18-Number
5

This content downloaded from 195.34.78.191 on Wed, 18 Jun 2014 15:52:15 PM


All use subject to JSTOR Terms and Conditions
NicoleK. Speer,Jeffrey
M. Zacks,andJeremy
R. Reynolds

showthatreadersspontaneously segmentnarratedactivities Speer,N.K. (2005). Textcomprehension in thebrain.Un-


processes
intoevents, butalsoprovide into
insight how readersusehigher- published doctoral
dissertation,
Washington St.Louis,
University,
level(i.e., at thesituation-model MO.
level)features
ofactivity
to
narrative textduring Speer, N.K., & Zacks, J.M. (2005). Temporalchangesas event
segment comprehension. boundaries: and ofnarrative
Processing memory
consequences
timeshifts.
Journal
ofMemory Language, and 53, 125-140.
Acknowledgments - The NationalInstitutes
ofHealth(NIH- Speer,N.K.,Zacks,J.M.,& Reynolds, J.R.(2006).Situationchanges
RO1MH70674) and the AmericanPsychologicalAssociation predictthe perception
of eventboundaries,readingtime,andper-
ResearchAward)supported thisresearch.The ceivedpredictability in narrativecomprehension. Manuscript
(Dissertation
submitted forpublication.
authorsthankRebecca Hedden,Jennifer Scott,and Carol Talairach,J.,& Tournoux,P. (1988). Co-planar
stereotaxic
atlasofthe
McKennaforassistancewithdatacollection;BrettHydeand humanbrain.Stuttgart, Germany: Thieme.
RolfZwaanforassistancewithstimulus
development;andDave van Dijk,T.A.,& Kintsch,W.(1983). Strategies ofdiscoursecompre-
BalotaandRandyBuckner forcomments ona previousdraft
of hension. New York:Academic Press.
VanEssen,D.C. (2002a). Surfacemanagement system [Computer da-
themanuscript. Retrieved from .
tabase]. April 13, 2005, http://pulvinar.wusd
edu:8081/sums/search.do?filename= ATLAS
REFERENCES VanEssen,D.C. (2002b).Windows onthebrain:Theemerging roleof
atlasesanddatabasesinneuroscience. Current
OpinioninNeuro-
Barker, R.G.,& Wright, H.F.(1951).Oneboysday:A specimen record 12, 574-579.
biology,
of behavior. New York:Harper & Brothers. VanEssen,D.C, Dickson,J.,Harwell, J.,Hanlon,D., Anderson,
C.H.,
Barsalou,L.W.(1999). Perceptualsymbolsystems. Behavioraland & Drury, H.A. (2001).Anintegrated software
systemforsurface-
BrainSciences, 22, 577-660. based analysesof cerebralcortex.Journalof the American
Boynton, G.M.,Engel,S.A., Glover,G.H., & Heeger,DJ. (1996). MedicalAssociation,41, 1359-1378.
Linearsystems analysisoffunctional magnetic resonanceimag- Van Essen, D.C, Drury,H.A., Harwell,J., & Hanlon,D. (2005).
ingin humanVI. Journal ofNeuroscience, 16, 4207-4221. CARET: Computerized anatomicalreconstruction and editing
Cohen,J.D.,MacWhinney, B., Flatt,M., & Provost, J. (1993). Psy- toolkit[Computer software and manual].RetrievedApril 13,
Scope:An interactive graphicsystem fordesigning and control- 2005,from http://brainmap.wustl.edu/caret
lingexperiments in thepsychology laboratory usingMacintosh Winer,B.J.(1971). Statistical in experimental
principles design(2nd
computers. Behavior Research Methods, Instruments, & Comput- ed.). NewYork:McGrawHill.
ers,25, 257-271. Zacks,J.M.,Braver,T.S., Sheridan,M.A.,Donaldson,D.I., Snyder,
Ferstl,E.C., & vonCramon, D.Y. (2001). The roleofcoherenceand A.Z., Ollinger,J.M.,et al. (2001). Humanbrainactivity time-
cohesionin textcomprehension: An event-related fMRIstudy. lockedto perceptualeventboundaries. NatureNeuroscience,
4,
Cognitive BrainResearch, 11, 325-340. 651-655.
Gernsbacher, M.A. (1990). Language comprehension as structureZacks,J.M., T.S.,& Reynolds,J.R.
Speer,N.K.,Swallow, K.M.,Braver,
building. Hillsdale,NJ:Erlbaum. (2007). Eventperception: A mind/brain perspective. Psycholog-
Glenberg, A.M. (1997). Whatis memory for?Behavioraland Brain ical Bulletin,
133, 273-293.
Sciences, 20, 1-55. Zacks,J.M.,Tversky, B., & Iyer,G. (2001).Perceiving, remembering,
Huettner,M.I.,Rosenthal, B.L.,& Hynd,G.W.(1989).Regionalcere- andcommunicating structurein events.Journal ofExperimental
bralbloodflow(rCBF)innormal readers:Bilateralactivationwith
Psychology: General, 130, 29-58.
narrative text.Archives ofClinicalNeuropsychology, 4, 71-78. Zwaan,R.A. (2004).Theimmersed experiencer: Towardan embodied
Mason,R.A.,& Just,M.A. (2004). Howthebrainprocessescausal theory oflanguagecomprehension. In B.H. Ross (Ed.), Thepsy-
inferences in text:A theoretical accountofgeneration and inte-
chologyof learningand motivation (Vol.44, pp. 35-62). New
gration component processes utilizingbothcerebralhemispheres. York:AcademicPress.
Psychological Science,15, 1-7. Zwaan,R.A.,Magliano, J.P.,& Graesser, A.C. (1995). Dimensions of
McAvoy, M.P.,Ohinger, J.M.,& Buckner, R.L. (2001). Clustersize situationmodelconstruction innarrative comprehension. Journal
thresholds forassessmentof significant activationin fMRI. ofExperimental Psychology: Learning, Memory, and Cognition,
Neurolmage, 13, S198. 21, 386-397.
Newtson, D. (1973).Attribution andtheunitofperception ofongoing Zwaan,R.A.,& G.A.(1998).Situation modelsinlanguage
Radvansky,
behavior. Journal ofPersonalityand SocialPsychology, 28, 28- comprehension and memory. Psychological Bulletin,123, 162-
38. 185.
Pulvermiiller,F. (2005). Brainmechanisms linkinglanguageand ac- Zwaan,R.A.,Radvansky, G.A.,Hilliard,A.E., & Curiel,J.M.(1998).
tion.NatureReviews Neuroscience, 6, 576-582. Constructing multidimensional situationmodelsduringreading.
Reynolds, J.R.,Zacks,J.M.,& Braver, T.S.(inpress).A computational StudiesofReading,2, 199-220.
Scientific
modelof eventsegmentation fromperceptualprediction. Cog-
nitiveScience.
Robertson,D.A., Gernsbacher, M.A., Guidotti,S.J., Robertson,
R.R.W.,Irwin,W, Mock,B.J.,& Campana,M.E. (2000). Func-
tionalneuroanatomy ofthecognitive processofmapping during (Received3/17/06; Revisionaccepted5/15/06;
discoursecomprehension. Psychological Science,11, 255-260. Final materialsreceived3/5/07)

5
Volume1&- Number 455

This content downloaded from 195.34.78.191 on Wed, 18 Jun 2014 15:52:15 PM


All use subject to JSTOR Terms and Conditions

Das könnte Ihnen auch gefallen