50
AppliedPsycholinguistics20:4
578
Nittrouer&Miller:Developmentofphonemiccodingstrategies
theyhadaccesstothatstructure.Nonetheless, the results fortheserialrecall
taskforthenormalreadersshowednoevidencethataphonemiccodingstrategy
hadbeen used to agreaterextent by them m thanbythepoorreaders tostore
wordsinworkingmemory.Severalresultssupportthisconclusion.Thenormal
readersmadesomewhatfewererrorsoverallthanthepoorreaders(althoughthe
effectdidnotreachstatisticalsignificance),buttheyshowednomoreofarhym-
ingeffectthanthepoorreadersshowed:thedifferencesinthenumberoferrors
madebetweenthenormalandpoorreaderswereroughlyequalforbothrhyming
andnonrhyminglists.Ifobserveddifferencesbetweenthereadinggroupswere
attributabletodifferencesintheextenttowhichaphonemiccodewasusedto
storewordsinworkingmemory,thelargestgroupdifferencewouldhavebeen
observedfornonrhymingmaterials, aswasthecasefortheadultsandthe11-
year-olds:meannumberoferrors weresimilarforrhymingmaterials forthe
adultsandthe11-year-oldsbutsignificantlydifferentfornonrhymingmaterials.
Also,therelationbetweenphonemicawareness(asmeasuredbyboththepho-
nemedeletionandpigLatintasks)andtheserialrecalldifferencescorewasnot
statisticallysignificant.Thus,nogroupdifferencewasobservedintheextentto
whichphonologicalcodingstrategies wereused,norwas there arelationbe-
tweenphonemicawarenessandtheuseofphonologicalcodingstrategiesacross
therangeofreadingabilities.Itseemsthattheuseofaphonemiccodingstrategy
forstoringitemsinworkingmemorydoesnotfollowautomaticallyasaresult
ofdevelopingaccesstothe syllable-internal, phonemicstructureoflanguage.
Instead,itseemsthatlearningtostoreitemsinworkingmemoryusingaphone-
miccodetakesplaceovertimeanddevelopmentallytrailstheabilitytoretrieve
phonemicinformationfromthelinguisticsignal.
Care was takenin this study tominimize task requirements. . Nonetheless,
theseresultsshowedlittledisparitybetweentheperformanceofthenormaland
poorreadersontherecalltask.Therefore,theconclusioncouldbereachedthat
thepoorreaders’serialrecallimprovedinthisstudy,relativetothatofearlier
studiesshowingdifferences inserialrecallbetweennormalandpoorreaders,
suchthattheyperformedsimilarlytothenormalreaders.Whileminimizingtask
requirementsmayhaveaccountedforthedecreaseindisparitybetweenthenor-
mal-andpoor-reading8-year-oldstosomeextent, neithergroupperformedas
well as s the e 11-year-olds oradults. . Thus, it seems fairto suggest that some
additionalskillmustbeneededbesidesbeingabletoaccessphonemicstructure
inordertomakeuseofthatstructureinworkingmemory.
Atthesametime,theserialrecalltaskusedinthisexperimentmayhavebeen
too difficultgenerallyfor8-year-olds, , thus degrading g theperformanceofthe
normal-reading8-year-olds.Itwasnotuncommonfor8-year-olds, eventhose
withnormalreadingabilities,tomake10errorsonsomeitemsintheintermedi-
atelistpositionsintheserialrecalltask,assuggestedbythehighmeanerror
rates forthesepositionsseeninFigures1and2. Inaddition, someofthe8-
year-olds were e unable to obtain n any y correct answers on the pig Latin task.
Therefore,ourabilitytodetectsignificantgroupdifferencesandsignificantcor-
relationsmayhavebeenconstrained.Experiment2wasdesignedtoseeifthe
resultsofthisfirstexperimentwouldbereplicatedwhenthesepotentialprob-
lemswerecorrected.
46
AppliedPsycholinguistics20:4
579
Nittrouer&Miller:Developmentofphonemiccodingstrategies
EXPERIMENTII:SIX-ITEMLISTS
Asecondexperimentwasconductedasacheckonthepossibilitythatthefailure
to find the anticipated effects s in n Experiment 1 fornormal andpoor readers
wasduetowhatistraditionallytermedceilingandflooreffects.Specifically,a
significantReadingAbility·Rhymeinteractionwasexpectedfortheserialre-
calltaskbutwasnotfound.Inaddition,significantcorrelationswereexpected
bothbetweenthereadingscoresandtheserialrecalldifferencescoresandbe-
tweenthe phonemic awareness scores andtheserial recalldifference scores.
Noneofthoseexpectedcorrelationswasobserved.Inthissecondexperiment,
severalproceduralchangesweremade.First,theserialrecalltaskwasconducted
with six-itemlistsinhopesthatanyceilingeffectsforthe numbers s oferrors
wouldbeavoided.Also,all48itemswereusedonthepigLatintaskinhopes
thatflooreffectsforthenumbersofitemscorrectonthattaskwouldbeavoided.
Thissecondchangewouldnotbeexpectedtohavemucheffectonitsownfor
the8-year-oldswhosimplycouldnotdothepigLatintask:thosechildrendid
notgetfurther thanthefirst sixitems. . Nonetheless, it t seemeda worthwhile
attempttospreadoutscores on n the pigLatintask. . Finally, the agerangeof
childrenincludedinthecomparisonofgoodandpoorreaderswasincreasedto
include8-,9-,and10-year-olds.Inadditiontoimprovingthechancesthatmost
ofthechildrenwouldnotscoreneartheflooronthepigLatintask,thischange
meantthatthenumberofchildren participatingwouldbeincreased,thus s im-
provingthepossibilityoffindingsignificantgroupdifferencesandcorrelations,
iftheyactuallyexistinthegeneralpopulation.
METHOD
Participants
Childrenbetween8and10yearsofagewereenlistedforthissecondexperi-
ment.Onlyonechangewasmadetothecriteriaforparticipationfromthefirst
experiment.TheblockdesignoftheWISC-IIIwasusedtoscreenthechildren
fornonverbalabilitiesinsteadoftheCPM.Thischangewasmadebecausethe
meanscoresforthe8-year-oldsinthefirstexperimentwerehigherthanwould
beexpectedforarandomlyselectedgroupofchildren, ifthetestnormswere
appropriateforthesesamplesofchildren.Atotalof73childrenmetthecriteria
forparticipation. Of these, 57children fit the e description ofnormal readers
(standardscoresforthereadingsubtestoftheWRAT-Rof95orbetter)and16
children fitthe descriptionof poorreaders s (standard d scores s of85 5 orpoorer)
usedinthe firstexperiment. Mean ageofthe participants ineachgroupwas
9;3.MeanstandardscoresonthereadingsubtestoftheWRAT-Rwere108for
thenormalreaders(SD=7)and76forthepoorreaders(SD=9).Asinthefirst
experiment, these scoresmeantthatthe normal l readerswere readingroughly
halfayearaboveexpectationsfortheirchronologicalage,andthepoorreaders
werereadingroughlyayearandahalfbehindexpectationsfortheirchronologi-
calage. Unlikethefirstexperiment,though,slightdifferenceswerefoundbe-
tweenthenormal-andpoor-readinggroupsonthecriterionmeasuresofgeneral
45
AppliedPsycholinguistics20:4
580
Nittrouer&Miller:Developmentofphonemiccodingstrategies
andlanguageabilities.OntheblockdesignoftheWISC-III,themeanscorefor
normalreaderswas.33standarddeviationsabovethemean,whereasthemean
score forpoor readers was .33 standard deviations below the mean. . Within-
groupstandarddeviationswerethesameas inthegeneralpopulationforboth
groups.Thebetween-groupdifferencewasstatisticallysignificant,t(71)=2.26,
p=.03. Onthe PPVT-R, themeanstandardscoreforthenormalreaderswas
106(SD=13),andthemeanscoreforthepoorreaderswas93(SD=11).This
differencewasalsostatisticallysignificant, t(71)=3.65,p<.001. Thesegroup
differenceswerenotconsideredproblematicinthisexperiment,largelybecause
theywereactuallysmallinmagnitude:groupmeansforbothgroupswerevery
close to the population means. . Furthermore, , such h differences couldonly in-
crease theprobabilityoffindinggroupdifferencesontherecalltask,andthe
prediction in this experiment was that no such differences s would d be e found.
Specifically,theeffectsofinterestweretheReadingAbility·Rhymeinterac-
tion,thecorrelationbetweenserialrecalldifferencescoresandreadingscores,
andthe correlations between each ofthe phonemic awareness measures and
serialrecalldifferencescores.Failuretofindtheseeffectsstatisticallysignifi-
cant, even though h slight differences s in n general l and language e abilities exist,
wouldonlyprovideparticularlystrongsupportforthecontentionthattheeffects
donotexistinthegeneralpopulation.
Stimuliandprocedures
Withtwoexceptions,thestimuliandprocedureswerethesameasinExperiment
1.First,all48itemswereusedinthepigLatintask.Second,thelistsofwords
fortheserialrecalltaskconsistedofsixitemsinsteadofeight.Forthenonrhym-
ing lists, , the words teen and d seedwere excluded; forthe rhyming lists, the
wordsPatandvatwereexcluded.Forthetraininglists,thelettersKandLwere
excludedfromthenonrhyminglists,andGandBwereexcludedfromtherhym-
inglists.
RESULTS
Phonemicawareness
Forthenormal-readinggroup, themeannumberofitemscorrectonthe pho-
nemedeletiontaskwas23.9(SD=6.3).Forchildreninthepoor-readinggroup,
the meannumberofitemscorrectwas13.4(SD=7.2).Thisgroupdifference
wasstatisticallysignificant, t(71)=5.71, p<.001. ForthepigLatintask, the
mean numberofitems s correctfor r the normal-readinggroupwas s 29.3 3 (SD=
14.1). The mean n number of items correct for the poor-reading group was
9.3(SD=12.3).Thisgroupdifferencewasstatisticallysignificant,t(71)=5.16,
p<.001.
Serialrecall
Table2showsmeanerrorscoresforthenormalandpoorreadersaswellasthe
meandifferencescores.AsinExperiment1, these scores are summedacross
list positions s but t provided separately for the e rhyming g and nonrhyming lists.
55
AppliedPsycholinguistics20:4
581
Nittrouer&Miller:Developmentofphonemiccodingstrategies
Table2.Meanerrors(outof60)acrossalllistpositions
forthenormalandpoorreadersbyrhymeconditionand
meandifferencescores
Normalreaders
Poorreaders
Rhyming(errors)
M
30.3
34.8
SD
(7.8)
(5.0)
Nonrhyming(errors)
M
24.2
29.4
SD
(8.8)
(6.6)
Differencescores
M
6.1
5.3
SD
(7.4)
(6.0)
Figure 4showsthe meannumberoferrorsforeachlist position. Atwo-way
ANOVAwasperformedonthesummederrorscoresacrosslistpositions,with
reading ability as the between-subjects s factor r and rhyming condition n as s the
within-subjectsfactor.AsinExperiment1,themaineffectofrhymingcondition
wasstatisticallysignificant,F(1,71)=32.53,p<.001. Thistime,theeffectof
readingabilitywasclearlysignificant, F(1, 71)=5.99,p=.017. Again, how-
ever,theReadingAbility·Rhymeinteractionwasnotsignificant.
Correlations
ThesamePearsonproduct-momentcorrelationcoefficientswerecomputedon
thesedataasonthoseofExperiment1,withthesameresults.Thecorrelations
betweenscores on n each phonemicawarenesstaskwithscores s onthe reading
subtestoftheWRAT-Rwerestatisticallysignificant:forphonemedeletion(r=
.70,p<.001),forpigLatin(r=.57,p<.001).Thesecorrelationsarestrikingly
similartothosecomputedforthedatainExperiment1. Thus,evenwithmore
itemsonthepigLatintaskandawiderrangeofparticipantages,thecorrelation
betweenthisphonemicawarenesstaskandreadingabilitywasnotasgreatas
betweenphonemedeletionandreadingability.AsinExperiment1,thecorrela-
tion betweentheserialrecalldifferencescores s andthescores onthe reading
subtestoftheWRAT-Rwasnotsignificant.Finally,thecorrelationscomputed
betweenscoresoneachphonemicawarenessmeasureandtheserialrecalldiffer-
encescoreswerenotsignificant.
DISCUSSION
ThepurposeofthesecondexperimentwastocheckthefindingsfromExperi-
ment1toensurethatobservedeffectswerenotattributabletoprocedurallimita-
tions. Inspiteofthechangesinprocedures,identicaltrendswereobservedin
thesecondexperiment:thepoorreadersmadesomewhatmoreerrorsonserial
24
AppliedPsycholinguistics20:4
582
Nittrouer&Miller:Developmentofphonemiccodingstrategies
Figure4.Numberoferrorsmadebythegoodandpoorreadersforserialrecallofsix-word
listsofrhymingandnonrhymingmaterials.
recallthanthegoodreaders,butthedifferencesbetweenthegroupsweresimilar
for the rhyming g and nonrhyming g materials. Consequently, no o evidence e was
found thatgood readers use e a a phonemic code to a greater extent than poor
readers. In addition, the e serial recall difference score was foundto correlate
neitherwithreadingabilitynorwithphonemicawareness.
GENERALDISCUSSION
Acentralquestionaddressedbythisstudywaswhetherchildrenwhoarecapa-
ble ofaccessingphonemicstructureinthespeech signalnecessarilyusethat
structureforstoringitemsinworkingmemorytoagreaterextentthanchildren
whohavedifficultyaccessingphonemicstructure.Inotherwords,doestheuse
ofphonemiccodingstrategiesinworkingmemoryemergeautomaticallywith
theabilitytoaccessthatphonemicstructure?Basedontheresultsoftwoexperi-
ments, the answerto o thisquestionisapparently“no.” Young normal readers
showednoevidenceofusingphonemiccodingstrategiesforstoringitems in
workingmemorytoagreaterextentthanyoungpoorreaders.First,nodiffer-
enceswerefoundbetweenthereadinggroupsinthemagnitudeoftherhyming
effect. Second, norelationwasfoundbetweenthe children’s readingabilities
andtheextenttowhichtheyusedaphonemiccodeforstoringitemsinworking
memory.
Documents you may be interested
Documents you may be interested