ERIC ED612379: Building a Knowledge Base: Predicting Self-Derivation through Integration in 6- to 10-Year-Olds

2018·0.64 MB·English

by  ERIC

JournalofExperimentalChildPsychology176(2018)55–72
ContentslistsavailableatScienceDirect
Journal of Experimental Child
Psychology
journal homepage: www.elsevier.com/locate/jecp
Building a knowledge base: Predicting
self-derivation through integration in
6- to 10-year-olds
⇑
Alena G. Esposito , Patricia J. Bauer
DepartmentofPsychology,EmoryUniversity,Atlanta,GA30322,USA
a r t i c l e i n f o a b s t r a c t
Articlehistory: Self-derivation of new factual knowledge through integration of
Received16January2018 separateepisodesoflearningisonemeansbywhichchildrenbuild
Revised10July2018 knowledge. Content generated in this manner becomes incorpo-
Availableonline10August2018
ratedintotheknowledgebaseandisretainedovertime;successful
self-derivationpredictsacademicachievement.Yetthecomponent
Keywords:
processes on which self-derivation through integration depends
Self-derivation
are as yet unknown. In parallel studies with 6- and 8-year-olds
Semanticknowledge
Integration (N=41; Experiment 1) and 8- and 10-year-olds (N=40;
Verbalcomprehension Experiment2),wetestedanumberofpredictorsrelatedtoother
Academicachievement productiveprocessesandlearning(reasoning,executivefunctions,
Transfer verbal comprehension, and long-term retrieval). Across studies,
withdifferentmethods,onlyverbalcomprehension,ameasureof
accumulatedsemanticknowledge,accountedforuniquevariance
in self-derivation through integration performance. The results
indicate that self-derivation through integration of separate epi-
sodes relates to accumulation of knowledge and the ability to
recruit the knowledge in the service of specific task demands.
Implicationsforcognitivetrainingandtransferarediscussed.
(cid:1)2018ElsevierInc.Allrightsreserved.
⇑
Correspondingauthor.
E-mailaddress:[email protected](A.G.Esposito).
https://doi.org/10.1016/j.jecp.2018.07.011
0022-0965/(cid:1)2018ElsevierInc.Allrightsreserved.
56 A.G.Esposito,P.J.Bauer/JournalofExperimentalChildPsychology176(2018)55–72
Introduction
Constructingaknowledgebaseisoneofthemostimportanttasksindevelopment.Knowledgelays
afoundationforlanguageandthought,andknowledgesupportstheacquisitionofmoreknowledge
(Goswami,2011).Childrenbuildknowledgethroughdirecttuitionandexperiences.Becauselearning
experiences are distributed over time and modalities, for knowledge to accumulate, separate but
relatedepisodesmustbeintegratedwithoneanother.Importantly,elaboratingtheknowledgebase
alsodependsonproductiveprocessesthatpermitself-derivationofnewsemanticcontentbasedon
integrated representations (see Bauer, 2009, 2012). Research has revealed that new facts resulting
from integration and self-derivation become incorporated into the knowledge base (Bauer &
Jackson,2015)andareretainedovertime(Varga&Bauer,2013;Varga,Stewart,&Bauer,2016).Crit-
ically, performance on tasks that test self-derivation of new factual knowledge through integration
predicts academic achievement in both reading and math (Esposito & Bauer, 2017). Currently
unknownarethecomponentcognitiveabilitiesonwhichself-derivationthroughintegrationdepends.
Inthe current research,in separatestudies with6- and 8-year-olds(Experiment1) and 8- and 10-
year-olds(Experiment2),wetestedcandidatepredictorssuggestedbyresearchonotherproductive
processes,specificallyreasoning,executivefunctions,verbalcomprehension,andlong-termretrieval.
Productiveprocessesandcomponentcognitiveabilities
Productive processes such as analogy, deduction, and induction allow learning to proceed effi-
cientlybecauselearnerscanextendbeyondinformationexplicitlyprovidedinalessontoderivenovel
understandings(e.g.,Goswami,1992,2011;Perret,2015).Productiveprocesseshavebeenobserved
across the lifespan (see Mandler & McDonough, 1996, for examples even during infancy) and are
assumed to be a major mechanism of cognitive development (Bauer & Varga, 2017; Bauer, 2012;
Brown,Bransford,Ferrara,&Campione,1982;Goswami,2011;Siegler,1989).
Thecurrentliteratureonthecomponentcognitiveabilitiesonwhichproductiveprocessesdepend
hasexaminedonlyalimitednumberofcandidates,withparticularemphasisoninhibitorycontroland
workingmemory.Thenumberofdirecttestsofthesecandidateprocessesissmall,yetthereiscon-
sensusthatbothinhibitorycontrolandworkingmemoryplayarole(seeGoswami,2011,fordiscus-
sion). In terms of direct tests, Moutier (2000b) studied inhibitory control, training in logic, and
deductivereasoningin10-and11-year-oldchildren.Inoneconditionchildrenweretrainedtoinhibit
acommonperceptualmatchingerror,andinanotherconditiontheyreceivedlogicaltraining.Atpostt-
est,onlythechildrenwiththetrainingininhibitorycontrolshowedimprovementinthedeductive
reasoningtask.Inarelatedstudy,Moutier(2000a)tested10-and11-year-olds’deductivereasoning
andperformanceonanumberofinhibitorycontroltasks(StroopTask,TrailMaking,andWisconsin
CardSort).Higherperformanceonthedeductivereasoningtaskwasrelatedtohigherinhibitoryper-
formance(seealsoMoutier,Plagne-Cayeux,Melot,andHoudé,2006,foradditional,indirectevidence
ofrelationsbetweencognitiveinhibitionand8-to10-year-olds’performanceondeductivereasoning
problems).
Otherstudieshavenotmeasuredinhibitorycontroldirectlybuthaveimplicatedit,workingmem-
ory, or both based on patterns of task performance. For example, Richland, Morrison, and Holyoak
(2006)investigatedanalogicalproblemsolvinginchildrenaged3–14years.Theyfoundthatyounger
childrenweremoresusceptibletothepresenceoffeature-leveldistractors,relativetoolderchildren,
andhadlowerlevelsofperformance.Theyinterpretedthepatterntoindicatethatyoungerchildren
did not yet have sufficient inhibitory control to ignore the distractors and maintain the task-
relevantinformationinworkingmemory.
Inadditiontoinhibitorycontrolandworkingmemory,accumulationanduseofpriorknowledge
hasbeenimplicatedasafactoraccountingforage-relatedchangesininductiveanddeductivereason-
ing (e.g., Goswami, 2011). Yet this conjecture has not been subjected to strong empirical testing
becausetherearefewstudiesthathaveincludedameasureofverbalcomprehensionorothermea-
sures of acquired knowledge among the candidate predictor variables. An exception is Fisher,
Godwin,andMatlen(2015),inwhichmeasuresofverbalcomprehensionandreasoningwereexam-
A.G.Esposito,P.J.Bauer/JournalofExperimentalChildPsychology176(2018)55–72 57
inedalongwithworkingmemoryandinhibitorycontrolaspredictorsofinductivegeneralizationwith
familiarcategories. Yetin thisstudytheverbalcomprehension andreasoningmeasureswerecom-
binedintoageneralintelligencecomposite,therebyprecludingexaminationoftheiruniquecontribu-
tionstoperformance.
Anotherliteraturethatbearsonthequestionofrelationsbetweenaccumulationanduseofprior
knowledgeandproductiveprocessesisresearchoninferencemakinginthecontextofreadingcom-
prehension.Readingcomprehensionisstronglycorrelatedtoverbalcomprehension(e.g.,Riedel,2007;
Yovanoff,Duesbery,Alonzo,&Tindal,2005),andbotharedescribedasmeasuresofsemanticknowl-
edgeanditsrecruitmentanduse(e.g.,Akshoomoffetal.,2013;Schrank,2010).Consistentwithsug-
gestionsof relationsbetweenprior knowledgeand successfulinference,severalstudies havefound
correlationsbetweenreadingcomprehensionandinferencemakingwithinasingletextsuchasdraw-
inganinferencefromtwoconsecutivesentencesinastory(e.g.,Cain&Oakhill,1999;Cain,Oakhill,&
Bryant,2004;Oakhill,Cain,&Bryant,2003).Interestingly,inthisliteraturethequestioniswhether
inferencemakingpredictsreadingcomprehension,notwhetherreadingcomprehension—asanindex
ofaccumulationanduseofpriorknowledge—predictsinferentialability.Thisorientationisapparent
inCainetal.(2004),forexample,whereitwasfoundthatbothworkingmemoryandinferencemaking
predicteduniquevarianceinreadingcomprehensioninchildrenaged7–9years.Thisandothersuch
studiessuggestthattherearerelationsbetweeninferencemakingandaccumulatedknowledge,yet
they do not directly inform relations between acquired knowledge and inference as a productive
process.
In summary, productive processes such as induction and deduction are supported by inhibitory
controlandworkingmemory.Readingcomprehensionisrelatedtothesecomponentcognitiveabili-
tiesaswellasinferencemaking,suggestingthataccumulatedknowledgemayalsoberelatedtopro-
ductive processes. Thus, both reasoning and verbal comprehension are indicated in productive
processing but have not been directly tested, and it is not clear whether they both explain unique
variance.
Self-derivationthroughintegrationandcomponentcognitiveabilities
Thespecificproductiveprocessthatisthesubjectofthecurrentresearchisself-derivationofnew
factualknowledgethroughintegrationofseparateyetrelatedepisodesofnewlearning.Thisprocessis
ofparticularsignificancebecauseitisaviablemodeloflearningasitoccursoutsidethecontrolofthe
laboratory.Initialinvestigationsofthisspecificproductiveprocessusedastorypassageparadigmin
whichchildrenwerepresentedwithtruebutpreviouslyunknownfacts(e.g.,dolphinsliveingroups
calledpods)embeddedwithinrichlycontextualizedstorypassages.Afterabriefdelay,arelatedtrue
butpreviouslyunknownfact(e.g.,dolphinstalkbyclickingandsqueaking)waspresentedviaasep-
aratestorypassagethathadauniquestoryline.Afteranotherbriefdelay,childrenwereaskedques-
tions (e.g., how do pods talk?), the answers to which could be generated by integrating the
informationfromtheseparateepisodes(e.g.,Bauer&SanSouci,2010).Performanceoncontroltrials
(inwhichonlyoneofthetwostoriesispresented)makesclearthatbothfactsarenecessarytoanswer
thequestions(Bauer&Larkina,2017;Bauer&SanSouci,2010).
Consistent with the suggestion that self-derivation through integration is importantto learning,
new factual knowledge derived through this process is retained over time. For example, in Varga
and Bauer (2013), 6-year-olds correctly produced the novel facts on 63% of open-ended trials and
recalled60%ofthefacts1weeklater(seeVargaetal.,2016,andVarga&Bauer,2017,forsimilarfind-
ings with 4-year-olds and college students, respectively). Moreover, performance on tests of self-
derivationthroughintegrationrelatestoacademicachievementinbothreadingandmathasindexed
by measures that determine grade promotion in kindergarten through Grade 3 (state standardized
performanceevaluations,classroomgrades,andnationallynormedassessments).Thisresearchsug-
geststhattheskillsandabilitiesonwhichself-derivationofnewfactualknowledgethroughintegra-
tiondependssharevariancewiththosethatsupportacademicachievementbroadly.
Especiallyinlightofevidenceofrelationsbetweenself-derivationthroughintegrationtaskperfor-
manceandacademicachievement,itisimportanttodeterminetheskillsandabilitiesonwhichthe
process depends. Like other productive processes such as analogy, deduction, and induction, self-
58 A.G.Esposito,P.J.Bauer/JournalofExperimentalChildPsychology176(2018)55–72
derivationthroughintegrationmaydependongeneralcognitiveabilitiessuchasinhibitorycontrol
and working memory (Goswami, 1992, 2011). These abilities tend to be interrelated (e.g., Fukuda,
Vogel,Mayr,&Awh,2010;Nettelbeck&Burns,2010)andespeciallyduringearlychildhoodmayalso
relatetoothergeneralcognitiveabilities,includingabstractreasoningandprocessingspeed(Mungas
etal.,2013).Measuresoftheseabilitiesalsorelatetoacademicperformance(i.e.,highachieversshow
highperformance;e.g.,Bhat,2016;Colom,Escorial,Shih,&Privado,2007;Floyd,Evans,&McGrew,
2003; Lynn, Meisenberg, Mikk, & Williams, 2007). As such, we would expect positive correlations
betweenmeasuresoftheseabilitiesandself-derivationthroughintegration.Consistentwiththissug-
gestion,inastudywith4-and6-year-olds,VargaandBauer(2014)foundthatmeasuresofreasoning
andverbalcomprehensionbothwerecorrelatedwithself-derivationperformance.Yetonlytherea-
soningtasksignificantlypredictedperformanceinaregressionmodel.
Inadditiontogeneralcognitiveabilities,self-derivationofnewfactualknowledgethroughintegra-
tionmayalsodependonaccumulatedsemanticknowledgeandthelong-termretrievalthereof.This
possibilityissalientbecausethematerialoverwhichthecognitiveprocessesarecarriedoutisreal-
worldsemanticknowledge.Thefactsthatsupportself-derivationarepreviouslyunknown,yetthey
relatetoconceptsabout whichchildrenhavesomefamiliarity(e.g.,animals ingeneral, dolphinsin
particular,howanimateobjectscommunicate).Assuch,long-termretrievalandsemanticknowledge
could be expected to relate to self-derivation through integration of separate episodes. This logical
possibility is supported by the reading comprehension literature that points to a relation between
inferencemaking and knowledge (e.g., Cain & Oakhill, 1999). It is further supported by the finding
of a correlation between reading comprehension and self-derivation of new factual knowledge
throughintegrationin7-to11-year-olds(Bauer,Blue,Xu,andEsposito,2016,Experiment2).Thecor-
relationwasobservedbothwhenthefactswerereadbythechildrenthemselvesandwhentheywere
readbyanexperimenter.Giventhatreadingcomprehensionisstronglycorrelatedtoverbalcompre-
hension and both are considered measures of semantic knowledge and the ability to recruit it, the
relationisconsistentwithsuggestionsofrelationsbetweensemanticknowledgeandself-derivation
throughintegration.
Insummary,self-derivationofnewfactualknowledgethroughintegrationofseparateepisodesof
newlearningisaparticularlypromisingareaofstudyduetotheecologicalvalidityofthestimuliand
paradigmandtherelationwithacademicperformance.Yettodate,fewstudieshaveexaminedthe
relation between self-derivation through integration and other cognitive abilities. The available
researchisnotsufficienttodeterminewhatcognitiveabilitiesrelatetoself-derivationthroughinte-
gration,althoughtheextantresearchofferssupportforgeneralcognitiveskillssuchasexecutivefunc-
tionsandreasoningaswellasaccumulatedknowledgeandtheabilitytouseittoreason.
Thecurrentresearch
Inthecurrentresearch,weconductedtwoexperimentstotesttheskillsandabilitiesonwhichself-
derivationthroughintegrationdepends.Weaddressedthequestioninchildrenacrosstheearlyschool
years,whentheyareheavilyengagedinthedevelopmentaltaskofaccumulatingaknowledgebase.In
Experiment1,weassessedself-derivationthroughintegrationin6-and8-year-oldsusingastorypas-
sageparadigm(Bauer&Larkina,2017;Bauer&SanSouci,2010).InExperiment2,weassessedself-
derivationthroughintegrationin8-and10-year-oldsusinganadaptationofthesingle-sentencepara-
digmusedbyBauerandcolleagues(Baueretal.,2016;seealsoVarga&Bauer,2017,foruseofthe
paradigm with adults). Thus, across the experiments, we examined self-derivation performance in
childrenaged6–10years.
Inbothexperiments,childrencompletedacognitivebatterythatassessedabstractreasoning,exec-
utivefunctions(includinginhibitorycontrol),verbalcomprehension,andlong-termretrieval.Weused
the Woodcock–Johnson III Tests of Cognitive Skills (COG; Woodcock, McGrew, & Mather, 2001) to
assess reasoning (Concept Formation), verbal comprehension (Verbal Comprehension), and long-
term retrieval (Visual Auditory Learning). Given the specific hypotheses regarding inhibition, we
added measures of inhibitory function to supplement the Woodcock–Johnson III COG battery. We
expected positive associations between self-derivation through integration and reasoning, long-
termmemoryretrieval,andexecutivefunctionsbecauseself-derivationthroughintegrationisapro-
A.G.Esposito,P.J.Bauer/JournalofExperimentalChildPsychology176(2018)55–72 59
ductiveprocessand,thus,logicallycouldbeexpectedtofollowthesamepatternofrelationasother
such processes, including analogy, deduction, and induction (e.g., Moutier, 2000a; Richland et al.,
2006). Based onpriorresearchonself-derivation throughmemoryintegrationspecifically(Varga &
Bauer,2014),weexpectedrelationswithabstractreasoninginparticular.Wealsoexpectedpositive
associations with semantic knowledge because the content over which the processes involved in
self-derivationthroughintegrationoperateisreal-worldfactualknowledge.Previousworkhasshown
a relation between reading comprehension and inference making (e.g., Cain & Oakhill, 1999) and
betweenreading comprehension and self-derivation throughintegration (Baueret al., 2016). Given
thatreadingcomprehensionisameasureofsemanticknowledge,itlogicallycouldbeexpectedthat
self-derivationthroughintegrationwouldrelatetothedepthandbreadthofaccumulatedsemantic
knowledgeandtheskillstorecruitittoreasonondeclarativetasks.
Evenasweadvancethesepredictions,wenotethatthecognitivefunctionsunderstudyoverlap.
Yetgiventhepsychometricstructureoftheinstrumentsweusedtomeasurethem,itisreasonable
to anticipate that we could identify relatively unique contributions. Specifically, in factor analyses,
the subtests of the Woodcock–Johnson III COG battery (Woodcock et al., 2001) load onto different
clusters and, thus, are interpreted to measure separate constructs; Concept Formation loads to the
Fluid Reasoning cluster (.81), Verbal Comprehension loads to Comprehension Knowledge (.95), and
VisualAuditoryLearningloadstoLong-TermRetrieval(.71)(Taub&McGrew,2004).Therearenosig-
nificant residual correlations between the measures. Although the measures of executive functions
(includinginhibitoryfunction)arenot partoftheWoodcock–JohnsonIII COGbattery andhave not
beensubjectedtofactoranalyseswiththeabovemeasures,similarmeasuresfromtheDelis–Kaplan
ExecutiveFunctionSystemhavebeencomparedwiththeWoodcock–JohnsonIIICOGbatteryclusters,
withresultsrevealingonlymoderatecorrelations,indicatingthatthesemeasuresalsorepresentasep-
aratefactor(Floydetal.,2006).Thus,althoughwecannotaccomplishcompleteisolationofthediffer-
ent cognitive components under study, we can identify relative unique contributions. The effort is
important because understanding the factors that contribute to successful self-derivation through
integrationcanguidethedevelopmentofinterventionsandeducationalmethodsaimedtofacilitate
productiveknowledgegeneration.
Experiment1
Method
Participants
Theparticipantswere20 6-year-olds(9girlsand11boys;M =6years7months,range=6;2–
age
6;10 [years;months] and 21 8-year-olds (11 girls and 10 boys; M =8years 6months, range=8;
age
2–8;10). Children participated in two sessions approximately 1week apart, each lasting approxi-
mately1h.ThechildrenwererecruitedfromavolunteerpoolconsistingofEnglish-speakingfamilies
fromalargemetropolitanareainthesoutheasternUnitedStateswhohadexpressedinterestinpar-
ticipating in child development research. Based on parental report, the sample was 24.4% African
American, 7.3% biracial or multiracial, and 68.3% Caucasian, with 2% of the sample self-identifying
asHispanic.Intermsofeducation,5%ofreportingparentsindicatedthattheprimarycaregiverhad
somecollege,2.5%reportedhavinganassociate’sdegree,40%reportedhavingacollegedegree,and
52.5% reported having education beyond a bachelor’s degree. One family did not report education
level. Parents provided informed written consent and children gave verbal assent at the beginning
ofSession1.Forthisandthesubsequentexperiment,auniversityinstitutionalreviewboardapproved
theprotocolandprocedures.Attheendofeachsession,childrenreceivedanage-appropriatetoyto
acknowledgetheirparticipation.AtthecompletionofSession2,parentsweregivena$10giftcard.
Stimuli
The stimuli were novel ‘‘stem” facts from three relational structure groups: ‘‘used to make,”
‘‘located in,” and ‘‘talks by.” Therewere two different domains within each relationalstructure, for
atotalofsixuniquedomains:palmtreesandseaweed(usedtomake),desertsandforests(located
60 A.G.Esposito,P.J.Bauer/JournalofExperimentalChildPsychology176(2018)55–72
in),anddolphinsandhippos(talksby).Withineachrelationalstructure,onedomainwaspresentedat
eachtestingsession,resultingintwosetsofsimilarbutuniquestimuli.Witheachdomain,eachstory
pairfeaturedtwonovelstemfactsthatcouldbecombinedtoderiveanovelintegrationfact.Thestem
facts and integration facts from the domains of palmtrees, deserts, and dolphinshad been used in
prior related research (e.g., Bauer & Larkina, 2017); the stem facts and integration facts from the
domains of seaweed, forests, and hippos were developed for this research. All facts were accurate
anddeterminedbypilottestingtobenoveltochildreninthetargetagerange.Furthermore,pilottest-
ing confirmed that both stem facts were necessary to derive the integration fact (see also Bauer &
Larkina,2017;Bauer&SanSouci,2010).
Eachstemfactwaspresentedinthecontextofatextpassagereadaloudbyanexperimenter(see
Appendixfor an example text passage).The passages were 81–89 wordsin lengthdistributed over
fourpages.Eachpageconsistedofahand-drawncolorillustrationdepictingthemainactionsofthe
text;thetext was notvisible on thepage. The passageswere similarin structure; in each passage,
acharacteruniquetothatpassage(e.g.,ladybug)learnedanewfactduringthecourseofan‘‘adven-
ture.”Onlythestemfactswereincludedinthepassages;theintegrationfactswerenotpresented.
Measures
Weassessedanumberofcognitivedomains.
Reasoning. TheConceptFormationsubtestoftheWoodcock–JohnsonIIICOG(Test5)wasusedasa
measureofcategoricalreasoningbasedoninductivelogic(medianreliability=.94from5to99years)
(Woodcock et al., 2001). Prior research has revealed relations between abstract reasoning and self-
derivationthroughintegration(Varga&Bauer,2014).Participantssawarowofshapeandcolorpat-
ternsandwererequiredtoderivetherulethatgovernseachsequence.Participantsreceived1point
foreachcorrectlyanswereditem;pointsweresummedforatotalscore.Thetotalscorewasthenused
todeterminethestandardizedscoreusingtheWoodcock–JohnsonIIICompuscoreandProfilespro-
gram.Theresultingscorewasbasedonchildren’sperformancecomparedwithacomparisongroup
ofsame-agedpeers,withameanof100andastandarddeviationof15.
Executivefunctions. Executivefunctionsweremeasuredwiththreetasks:theTrailMakingTask,Biva-
lent Shape Task, and Go/No-Go task. All three were administered on a Dell Optiplex 755 computer
attached to a Dell S2240Tb 21.5-inch color touchscreen monitor using the Psychology Experiment
Building Language (Mueller, 2010). These tasks were included due to relations between executive
functionsandotherproductiveprocesses(Goswami,1992,2011).
WeadministeredaversionoftheTrailMakingTask,ameasureofcognitiveflexibility,developed
specifically for children(e.g., Bowie & Harvey, 2006; Delis, Kaplan, & Krames, 2001; Mueller,2010;
Reitan,1971).Childrencompletedthreetrailsbyputtingitemsinorder.Thefirsttrailrequirednumer-
ationfrom1to16(1–2–3).Inthesecondtrail,thesequencewasalphabeticalfromAtoK(A-B-C).In
thelasttrail,thesequencealternatedbetweenanumberandaletter(1-A-2-B-3-C).Responsetimefor
thelasttrailisthemeasureofcognitiveflexibility.TheBivalentShapeTask,ameasureofinhibitory
control, requires children to match the shape between objects and ignore the highly salient color
(Esposito,Baker-Ward,&Mueller,2013).Thedependentvariablewasmeanreactiontimeoncorrect
incongruenttrials(colorandshapemismatch).TheGo/No-Goparadigm(Archibald&Kerns,1999),a
secondmeasureofinhibitorycontrol,wasadaptedfortouchscreenuse(Mueller,2010).Participants
wereaskedtotapthescreeninresponsetothetargetstimulus.Thetargetstimulusappearedwith
80%frequencyandnontargetwith20%frequency.Thedependentvariabletakenfromthismeasure
wastotalerrors.
Long-term retrieval. The Visual Auditory Learningsubtest of the Woodcock–Johnson III COG (Test 2
and Test 10) served as a measure of long-term memory for associated stimuli (median reliability=
.86 from 5 to 99years) (Woodcock et al., 2001). This measure was included to ascertain whether
self-derivationthroughintegrationisrelatedtothestoringandreactivationofinformationinlong-
termmemory.Inthistask,participantsareshownaseriesofrebuses(pictographicsymbolsofwords)
and are later asked to recall the visual–auditory associations from long-term memory. Specifically,
A.G.Esposito,P.J.Bauer/JournalofExperimentalChildPsychology176(2018)55–72 61
oncetherebusesarelearned,participantsarepresentedwithseveralrebusesformingasentenceand
areaskedtospeaktheassociatedwordsaloud(Test2).Oneweeklater,participantswerepresented
withnewseriesofrebusesto‘‘read”withoutbeingremindedofthewordassociations(Test10).In
both Test 2 and Test 10, participants received 1 point for each incorrectly answered item, defined
asafailuretoidentifythecorrectwordortodosowithin5sofviewingarebus.Totalerrorswere
enteredintotheWoodcock–JohnsonIIICompuscoreandProfilesprogram,whichreturnedstandard
scoresforeachmeasurebasedonchildren’sperformancecomparedwithsame-agedpeers(again,a
meanof100andastandarddeviationof15).
Verbalcomprehension. TheWoodcock–JohnsonIIITest1isanorm-referencedverbalcomprehension
measure(medianreliability=.90from5to99years)(Woodcocketal.,2001).Thetestincludesfour
subtests—PictureVocabulary,Synonyms,Antonyms,andAnalogies—whichassesscomprehensionof
individual words as well as the relation among words. The test represents context-free knowledge
andsematicmemory(Schrank,2010).Thismeasurewasincludedtoascertaintherelationbetween
self-derivation through integration and acquired semantic knowledge and is based on evidence of
relations between self-derivation and academic performance and reading comprehension (Bauer
etal.,2016;Esposito&Bauer,2017).Childrenreceived1pointforeachcorrectlyanswereditemin
eachsubtest.ThesubtestscoreswerethenenteredintotheWoodcock–JohnsonIIICompuscoreand
Profilesprogram,whichreturnedacompositestandardscorebasedonchildren’sperformancecom-
paredwithsame-agedpeers(again,ameanof100andastandarddeviationof15).
Procedure
The procedure was modeled after related research (Bauer & Larkina, 2017; Bauer & San Souci,
2010).Childrenweretestedindividuallyinalaboratoryroombyoneoftwofemaleexperimenters;
childrenweretestedbythesameexperimenterateachofthetwosessions.Eachsessionwasapprox-
imately1hlong.Sessionswerevideo-recorded,andtherecordingswereregularlyreviewedtoensure
protocolfidelity.
Session1. Atthefirstsession,childrenwereexposedtothreestorypassagepairscontainingrelated
facts,onestorypassagepairfromeachoftherelationalstructuregroups(i.e.,usedtomake,located
in,andtalksby).Eachstorypassagewasreadtwicewhilechildrenwereencouragedtolistenandlook
atthepictures.Thefirstpassagefromeachpairwasread,followedbyverbalcomprehensionSubtests
AandB.Thesecondpassagefromeachstorypairwasthenread.Followingthesecondpassage,chil-
dren participated in the computerized measures of executive functions in random order. Children
were then tested for self-derivation through integration performance in an open-ended format.
Following open-ended testing, for trials on which children did not provide a correct answer in
open-endedtesting,theywerepresentedwithaforced-choicequestionwiththreealternatives,one
of which was correct. Forced-choice testing was included to ensure that there would be sufficient
variability if open-ended testing resulted in a restricted range. We concluded the session with the
long-termretrievaltask.
Session 2. Children returned to the laboratory approximately 1week after their initial visit
(M=7.2days).ThesessionproceededinmuchthesamewayasSession1.Previouslyunseenpassage
pairsfromeachrelationalstructurewerepresented.Storypassageswithineachpairwereseparated
by the verbal comprehension Subtests C and D. Computerized executive function tasks again
separated the passages from self-derivation through integration performance. Children were first
askedself-derivationquestionsinanopen-endedformatfollowedbyaforced-choiceformat(again,
onlywhenchildrendidnotproduceacorrectanswerinopen-endedtesting).Afterthetests,children
completedthereasoningmeasureandthedelayedlong-termretrievalmeasure.Fortheself-derivation
through integration task, across participants, story passage pairs were presented approximately
equally often in Session 1 and Session 2. In addition, within a story passage pair, stories were
presentedfirstandsecondapproximatelyequallyoftenacrossparticipants.
62 A.G.Esposito,P.J.Bauer/JournalofExperimentalChildPsychology176(2018)55–72
Scoringofself-derivationthroughintegration
Theexperimentersrecordedchildren’sanswerstoboththeopen-endedandforced-choiceintegra-
tion questions. Participants were awarded 1 point for each correctly provided answer in the open-
ended and forced-choice formats. Thus,there were two measures of performance: open-ended and
totalscore(calculatedbyaddingopen-endedandforced-choicescores).Scoringwascheckedbyeach
experimentertoensurefidelity.
Results
The results are reported in two sections. First, we report the self-derivation through integration
performance and the relation betweensessions. Second,we report the relations betweenmeasures
ofcognitiveabilitiesandself-derivationthroughintegrationperformance.Analyseswereconducted
usingSPSSVersion24.Allstatisticaltestsweretwo-tailed.
Self-derivationperformance
Self-derivation through integration performance for both Sessions 1 and 2 is represented in
Fig.1forboth6-and8-year-olds.Overall,the6-year-oldsshowedevidenceofself-derivationthrough
integration at both sessions without ceiling or floor effects. The 8-year-olds showed variability in
open-endedperformanceatbothSessions1and2butreachedceilinginthetotalscoreatbothsessions.
PerformancefromSession1toSession2didnotdiffersignificantlyforeither6-year-olds(open-ended:
t=(cid:1)1.19,p=.25;totalscore:t=(cid:1)1.19,p=.65)or8-year-olds(open-ended:t=(cid:1)0.96,p=.35;total
score: t=1.45, p=.16). Across the sample, Session 1 and Session 2 were significantly correlated
(r=.45,p=.004).Giventhattherewerenodifferencesinperformanceandthatperformanceacross
the two sessions was correlated, Session 1 performance and Session 2 performance were summed,
effectivelyincreasingtherangefrom0to3to0to6.Inaddition,giventhattherewassufficientvariance
in open-ended testing (M=3.51, SD=1.86, 59% correct, range=6) but not in total score (M=5.28,
SD=1.34,88%correct,range=5),onlyopen-endedperformancewasexaminedfurther.
Relationswithcognitiveabilities
PerformanceonthemeasuresofcognitiveabilitiesisreportedinTable1,PanelA.InTable2,Panel
A, wereportthe correlations.Reasoning and verbal comprehension weresignificantlycorrelated to
self-derivationthroughintegrationperformance.Noothercognitiveabilitieswerecorrelatedtoself-
derivationperformance.
e
c
n
a
m
r
o
rf
e
p
n
a
e
M
6-year-olds 8-year-olds 6-year-olds 8-year-olds
Total score
Par(cid:2)cipant age
Fig.1. Meanself-derivationperformanceacrosstestsessions:Experiment1.
A.G.Esposito,P.J.Bauer/JournalofExperimentalChildPsychology176(2018)55–72 63
Table1
Performanceoncognitivemeasures.
PanelA:Experiment1 PanelB:Experiment2
Verbalcomprehension,standardizedscore 115.49(14.35) 115.58(13.87)
Long-termmemory,standardizedscore 112.93(15.91) 104.15(11.90)
Long-termmemorydelay,standardizedscore 118.74(13.53) 103.87(10.58)
Reasoning,standardizedscore 107.29(13.53) 108.47(11.87)
TMT,totalcompletiontime 48.48(29.31) 35.60(14.08)
BST,meanreactiontime 1.14(0.21) 1.05(0.19)
GNG,meanerrors 4.37(2.60) 3.68(1.61)
Note.Standarddeviationsareinparentheses.TMT,TrailMakingTask;BST,BivalentShapeTask;GNG,Go/No-Gotask.
Wenextconductedaregressionanalysistoascertainwhetherreasoningandverbalcomprehen-
sion, the measures that were significantly correlated with self-derivation performance, explained
uniquevariance(seeTable3).Ageofparticipantswasalsoincludedinthemodel.Themodelwassig-
nificantandexplained51%ofthevarianceinself-derivationthroughintegrationperformance.Only
ageandverbalcomprehensionweresignificantpredictorsofperformance;reasoningdidnotexplain
asignificantportionofthevariance.
Discussion
Both6-and8-year-oldsshowedevidenceofself-derivationofnewknowledgethroughintegration
ofseparateyetrelatedepisodesofnewlearning.PerformanceatSession1andperformanceatSession
2werenotsignificantlydifferent,andperformanceatthetwosessionswassignificantlycorrelated.
Ofthemeasuresofgeneralcognitiveabilities,onlyreasoningandverbalcomprehensionwerecor-
relatedtoself-derivationthroughintegrationperformance.Yetintheregressionmodel,reasoningdid
notpredictuniquevariance.Instead,verbalcomprehension—ameasureofthedepthandbreadthof
accumulatedsemanticknowledge—andparticipantageexplainedsignificantanduniquevariance.
Themethodusedinthecurrentexperimentaffordedavalidtestofthequestionofthespecificityof
relations between measures of cognitive and semantic development and self-derivation of new
knowledgethroughintegration.Yetonelimitationoftheapproachwasthatwemeasuredchildren’s
performanceononly6trials,yieldingasomewhatrestrictedrangeofscores.Inopen-endedtesting,
the performance range was 6 and in total score the range was 5. Accordingly, in Experiment 2 we
adopted a new method of testing that more than quadrupled the number of trials in each session.
Wealsoincreasedtheagerangetoinclude10-year-olds.Wedid notinclude6-year-olds,however,
inanticipationthattheywouldnotbeabletomeettherequirementthattheyreadthestimulusfacts
themselves(seebelow).
Experiment2
Method
Participants
The participants were 21 8-year-olds (8 girls and 13 boys; M =8years 5months, range=
age
8;2–8;11)and1910-year-olds(9girlsand10boys;M =10years5months;range=10;1–10;10).
age
Children participated in two sessions, each 1h long, approximately 1week apart. Children were
recruitedfromthesamesourceandrepresentthesamepopulationasinExperiment1.Noneofthe
children had participated in Experiment 1. Based on parental report, the sample was 25% African
American, 18% biracial or multiracial, 3% Asian, and 50% Caucasian, with 3% of the sample self-
identifying as Hispanic. Two parents did not respond to the request for racial self-classification. In
terms of education, 15% percent of parents reported the primary caregiver having a high school
degree, 10% reported having an associated degree, 22.5% reported having a college degree, and
52.5% reported havingeducationbeyond a bachelor’s degree. Asin Experiment 1, parentsprovided
6
4
Table2
Correlationsbetweenself-derivationthroughintegrationandcognitivemeasures.
A
.G
1.Self- 2.Verbal 3.Long-term 4.Long-termmemory 5. 6. 7.BST 8. .E
derivation comprehension memory delay Reasoning TMT GNG sp
o
PanelA:Experiment1 sito
,
1. Open-endedself-derivation P
2. Verbalcomprehension,standardized .45** .J.B
score au
3. Long-termmemory,standardizedscore .28 .33* er/
4. Long-termmemorydelay,standardized .17 .30 .77** Jou
score rn
5. Reasoning,standardizedscore .33* .59** .25 .18 al
o
6. TMT,totalcompletiontime .20 (cid:1).21 .15 .05 (cid:1).002 fE
7. BST,meanreactiontime (cid:1).05 (cid:1).007 (cid:1).02 .09 (cid:1).20 (cid:1).002 xp
e
8. GNG,meanerrors (cid:1).006 .01 .11 .001 .02 .30 (cid:1).007 rim
e
PanelB:Experiment2 n
1. Open-endedself-derivation tal
2. Verbalcomprehension,standardized .42** Ch
score ild
P
3. Long-termmemory,standardizedscore .17 .30 sy
4. Long-termmemorydelay,standardized .05 .15 .59** ch
o
score lo
g
5. Reasoning,standardizedscore .39* .44** .45** .58** y
1
6. TMT,totalcompletiontime (cid:1).22 (cid:1).08 .14 .41* .007 76
7. BST,meanreactiontime (cid:1).06 .29 .07 .13 .18 .28 (2
8. GNG,meanerrors (cid:1).11 (cid:1).11 .002 .16 .005 .23 (cid:1).16 01
8
)
Note.TMT,TrailMakingTask;BST,BivalentShapeTask;GNG,Go/No-Gotask. 55
* p<.05. –7
2
** p<.01.