Cognitive Processing

8/12/2019 Cognitive Processing

1/12

Cognitive Processing in Oral and Silent Reading ComprehensionAuthor(s): Aita SalasooReviewed work(s):Source: Reading Research Quarterly, Vol. 21, No. 1 (Winter, 1986), pp. 59-69Published by: International Reading AssociationStable URL: http://www.jstor.org/stable/747960.

Accessed: 18/04/2012 02:43

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 [email protected].

International Reading Associationis collaborating with JSTOR to digitize, preserve and extend access to

Reading Research Quarterly.
http://www.jstor.org/action/showPublisher?publisherCode=irahttp://www.jstor.org/stable/747960?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/747960?origin=JSTOR-pdfhttp://www.jstor.org/action/showPublisher?publisherCode=ira


2/12

Cognitiveprocessingin oraland silent readingcomprehension

AITA ALASOOIndiana UniversityTHERELATIONetweenoral and silent readingcomprehensionwas examinedin an informa-tion-processing ramework. n this framework,comprehension nvolves constructionof andaccess to a hierarchicalknowledgestructureby the reader.Readingratesandcomprehensionmeasures thatprobedrecognitionof variouslevels of text structurewere collected for pas-sages readorally and silently by 16 college students.Oral readingrates were slower thansilent readingrates. More encompassingor higherlevel representationswere verifiedmoreslowlythan exical andlower level textrepresentations.Differencesdue to readingmodewerefoundonly for low- andhigh-level propositions hat occurred n the textexplicitly;for these,silent readingof the text led to slower verificationresponsesthan oral reading.The resultssuggest that memorytraces of text microstructure reated in (slower)oral readingare ac-cessed fasterduring memory-basedcomprehension asks than traces establishedby fasterprocessesthat occurduringsilentreading.

Processuscognitifde la comprdhension e l'crit en lectureorale etsilencieuse ENSEBASANTur le processusde traitement e l'information n a compare a comprehensionde l'6criten lecture orale et en lecturesilencieuse. Selon cette base, la comprehensionmpli-que l'61aborationt l'utilisationd'uneorganisationhierarchiquedes connaissances(precon-naissances)parle lecteur. En faisant ire oralementpuis silencieusementdiffdrentspassagespar seize.61lvesde college, on a pu recueillir des donnees concernant a vitesse de d6chif-frage et I'6valuation e la comprehensionen postlecture,a partirdesquelleson peut deter-miner la reconnaissancedes niveauxdiff6rentsde la structuredu texte. La lecture orales'effectuaitplus lentementque la lecturesilencieuse. Ona puobserver a theoriede l'effetdesniveaux(levelseffect)de Kintsch(1974) dans la capacit6d'utiliser es preconnaissances:esrepresentations lus globales ou d'unniveausup6rieur e v6rifiaientplus lentementque lesrepresentationsexicaleset de niveauinf6rieur.Des diff6rencesattribuables u mode de lec-turen'ont6t6constatiesque pourles propositionsdes haut et basniveauxretrouv6es xplici-tementdans le texte: la lecturesilencieuseentrainaitune v6rificationplus lente des r6ponsescomparativement la lecture orale du texte. Les r6sultatsobtenusrevelentque le mode delecture choisi produitun effet qui affecte le courstemporeldes processusde comprehensionau niveausuperieur.Lorsd'exercicesm6morielsde comprehension, l est plus facile d'acc6-der aux donnees sur la microstructure u texteenregistrees ors de la lecturesilencieuse, ouile processusd'"enregistrement"es donneess'effectueplus rapidement.Finalement,on pro-poseraquelquesorientationsaux recherchesulterieures.

Note. This manuscriptwasacceptedfor publicationunder heeditorshipof S. JaySamuelsandP. DavidPearson.

59


3/12


4/12

1980; LaBerge & Samuels, 1974; Rumelhart,1977). During reading comprehension,wordsare identified (e.g., Glushko, 1981; Levy,1981), larger meaning structuresor proposi-tions are constructed and integratedto priorknowledge (e.g., de Beaugrande, 1980;Kintsch, 1974, 1977), and, accordingto someresearchers, inferencesor macrostructures redrawn from information in the printed text(Frederiksen,1981; van Dijk, 1979).4) The reader in this framework comesfrom the populationof college studentsand ispracticedat readingfor comprehensionand forlater retrieval of various levels of knowledgefrom the text. Cautionin extendingthe follow-ing discussionto otherpopulations s advised.Numerousprevious studies of oral and si-lent reading have used both on-line measure-ments of readingbehavior andperformanceonsubsequent memory and comprehensiontests.However, this body of research has not beenconclusive about the extent and locus of differ-ences between oral and silent reading proc-esses. Two examplessuffice: First, differencesbetween silent and oral readinghave been ob-served in eye movements (e.g., Anderson &Swanson, 1937; Fairbanks,1937; Wanat,1976)and in reading rates (e.g., Juel & Holmes,1981; Mead, 1915, 1917; Rogers, 1937), butthe role of these differenceshas yet to be clari-fied in a model of readingcomprehension.Sec-ond, the implications for silent readingprocesses from oral readingerrors or miscues(e.g., Danks & Hill, 1981; Goodman, 1969,1970a, 1970b; Goodman & Goodman, 1977;Levy, 1981; Weber,1968) are not clear. Skilledsilent readingis covert in natureand errorsofomission can be inferredonly indirectlyfromregressive eye movements (Rayner &McConkie, 1976). Nevertheless,miscue analy-sis is the empirical foundationof Goodman'stheory of reading (e.g., Goodman, 1970a,1970b;Goodman& Burke, 1973). Miscuedatafrom oral readingcan be informativeaboutsi-lent readingonly if one assumes a unitarycog-nitive basis forcomprehensionn bothmodesofreading.And this assumption s the one underinvestigationhere: How similar are the cogni-tive components of comprehension followingoral andsilentreading?

Historically, two kinds of evidence havebeen held to support the existence of sharedcognitive processes in oral and silent reading.First, the process of phonologicalrecodinghasbeen implicated n both modes of reading(e.g.,Baron, 1977; Glushko, 1981; Kleiman, 1975;Spoehr& Smith, 1975). The functionof phono-logical recodingof visual print may be to keepseveral words (their meanings and structuralconstraints)n short-termmemoryduring luentreading (Banks, Oka, & Shugarman, 1981).Recodingoccursrelativelyearlyin thetemporalcourse of reading,when compared o the finalproduct of comprehension. Second, a sizablenumberof studies have failed to find compre-hension differencesafter oral and silent reading(e.g., Anderson & Dearborn, 1952; Anderson& Swanson, 1937; Gray, 1956; Jones, 1932;Juel& Holmes, 1981; Poulton& Brown, 1967;Rogers, 1937). Because researchershave iden-tified an on-line readingprocess (phonologicalrecoding) n bothmodesof readingand becausethey have failed to observe differencesbetweenthe two comprehension ests, it may seem ap-propriateto conclude that a central cognitivebasis is commonto both oral and silentreading.Sucha conclusionwouldbe premature ndwouldconstitutea misinterpretationf the liter-ature;research n both fields is still surroundedby controversyand steeped in methodologicalproblems. Phonological recoding may interactwith properties of the text structure, such ascontextual constraint(e.g., Wanat, 1976), andwith measuresof readability hat are functionsof both the reader and the text (e.g., Coke,1974). Evidencefor such interactions rom vo-calizationsuppressionstudiesis difficult to as-sess, since under such conditionssubjects mayuse unnatural eadingstrategies.Indeed,if pho-nological recoding nteractswithsome text-andreader-basedvariables, it may also be affectedby readingmode.Thebodyof literaturenoraland silentread-ing comprehensions inconclusive.Several lass-room studies of oral and silent readingcomprehension avereported ralreadingadvan-tages (Collins, 1961; Elgart, 1975; Swalm,1973);othershavefound silentreading o be su-perior for comprehension(e.g., Mead, 1915,1917).Thus,one cannotdrawstrongconclusions

Oralandsilent reading SALASOO 61


5/12

fromthe null findings n postreading ests.One methodologicalproblemwith the oraland silent readingliteraturehas been an oftenloose, global definition of comprehension(ifone is attemptedat all). Inappropriater insen-sitive measurement techniques often followfrom this type of conceptualshortcoming.An-other problem with comprehension measure-ment tools has been a wide range ofcomponents used to produce the requiredre-sponse. The task demandsposed by multiple-choice comprehensionquestions (e.g., Collins,1961), standardized written comprehensiontests (e.g., Elgart, 1975), and oral cloze proce-dures(e.g., Swalm, 1973) varygreatly.Solutions to these problemsin measuringcomprehensionof readingtexts may lie in re-cent developments n the field of text and dis-courseprocessing:namely,in the proposalof anumberof detailedmodels of knowledgestruc-ture andcomprehension e.g., Anderson, 1976;de Beaugrande, 1980; Kintsch & van Dijk,1978; Minsky, 1975, 1980; Schank, 1975,1981). In one model (Kintsch & van Dijk,1978), readingpassages and the meaningsde-rivedfromthemby readersare viewedin termsof hierarchical propositional structures andtheir interrelations and associative links withthe centralthemeof the text. By carryingout apropositionalanalysisof a reading ext basedonsuch models (e.g., Turner& Greene, 1977),one can identifylevels of knowledgestructuresin relation o the central heme. Thelowestlevelin this hierarchy s given to lexical (or surface)knowledgeof the words that occur in the text;concepts related to the meaning of perceivedwords compose the propositionalcomprehen-sion structures.Largerunitsof meaningexplic-itly presented in the text comprise the text'smicrostructure f low-level andhigh-levelprop-ositions. High-levelmicrostructure ropositionsare more complex and encompass lower levelpropositions.The same knowledge structures that de-scribe the text itself are important or the proc-esses involved in reading, understanding,andremembering he passage. Kintsch (1977) has

suggested that higher level structuresrequiremore processing and leave more enduringmemory traces than lower level structures.Insupport of this notion, strong microstructurelevels effects have been found in readingtimeand recall measures n studiesof silent reading(Cirilo& Foss, 1980;Kintsch& Keenan, 1973;Kintsch, Kozminsky, Streby, McKoon, &Keenan, 1975): More time is spent readinghigherlevel propositions han lowerlevel ones,and afterwards, the former are recalled withgreaterprobability hanthelatter.This studyexaminestwo extensionsof pre-vious propositionalevels effects in comprehen-sion. First, in traditionalmemorystudies(e.g.,Tulving,1976), itemsthat arerecalledbetterarerecognizedmorepoorly.It is buta small steptorelate hiswell-documentedesult o thesuperiorrecall of high-level propositionsover low-levelpropositions.The prediction ollows thathigherlevel statements rom the text might be recog-nized with less accuracy, peed, andconfidencethan lower level knowledgestructures.Second,comprehending text also entailsthe abilitytoderive inferencesconsistent with the thematiccontent or the meaningmacrostructures f thetext (Frederiksen,1981;van Dijk, 1979). Usingpropositional nalysis,one maybe ableto com-pare comprehensionof inferences with that ofhigh-level propositions hatoccur in the text. Insum, recent advances n text analysisandcom-prehensionmodelinghave enabledmore accu-ratemeasurement f variousaspectsof readingcomprehension.'These tools maymake t possi-ble to understandetter he nature f comprehen-sion in oral andsilentreading.In this spirit, the present study askswhetheroral and silentreadingprocessesdiffer-entially affect variouslevels of comprehension,namely,lexical knowledge, low- andhigh-levelpropositions, and inferences. Four dependentvariables were used to investigate readingandcomprehension; he first variablewas chosentoconfirm the relation between oral and silentreading rates, and the other three variables tomeasurecomprehensionat each of four differ-ent levels.

62 READING ESEARCHQUARTERLY* Winter 1986 XXI/1


6/12

MethodSubjectsSixteen IndianaUniversity students weretested individually.Subjectswere right-handednative English speakers with normal or cor-rectedvision andno knownreadingdisabilities.Subjectsreceived creditas partial ulfillmentofthe requirements or an introductorypsychol-ogy course.MaterialsTwelvetestpassagesandthreepracticepas-sages between 150 and 300 words in length,previouslyused in studiesof listening compre-hension (Blank, Pisoni, & McClaskey, 1981;Brunner& Pisoni, 1982), were chosen as ex-positoryreadingtexts. Foreach text, four veri-ficationstatementswere constructed o evaluatevariouslevels of comprehension.A sample pas-sage andits statementsare shown in Table1.Memory for the lexical knowledge in thetextwas testedwithquestionsthatrequiredrec-ognitionof test wordsas havingoccurred n thereading passage or not. The questionswere ofthe standardorm, "Didthe word'XXXX' oc-cur in this story?"Either correct words or dis-tractorwords synonymousto wordswhich hadoccurred in the test passage (and of similarlength and frequency) occurred in the targetposition.

Three additionalquestiontypes addressedhigher levels of comprehension.The proposi-tionalrepresentationonstructedduringreadingand comprehensionof a text was examinedbyaskingreaders o verify (a) high- and low-levelpropositions, and (b) inferential statements.The verificationquestionsoccurred n two ver-sions, each of which required either a "Yes/True"or "No/False"response from the reader.Low-level propositionspresentedfor recogni-tion either were exact repetitionsof one-clausesentencesfromthe test passageor hadone sub-stitutedwordthat rendered he statement ncon-gruous with the text. Similarly, high-levelproposition test items repeated or misrepre-senteda clausecentral o thepassagetheme,butone which did not necessarily occur within asingle sentence. Inferencesrequiredsubjectstosynthesize informationexplicitly conveyed inthepassage;false inferenceswerecontradictoryto accurate yntheses.DesignReadingmode and level of comprehensionquestionwere within-subjectsvariables. Read-ing mode was blocked-each subject read sixstoriesaloudand six silently.Theorderof silentand oral responseblocks was counterbalancedacross subjects.Twocounterbalanced rders ofthe 12testpassageswereused, each with 8 sub-jects. Withineach of these two subjectgroups,

Table1 Sampletest readingpassageIn ancientRome, JuliusCaesarbannedchariotdrivingatnight.It seems the thundering hariotwheels made toomuch noise. Now-over 2,000 years later-people arestarting o realize that noise isn'tgood for them. It affects theirhearing,theirpeace of mind,theirabilityto workefficiently,and,as some doctorspointout, theirgeneralhealth.Mostpeoplestill acceptnoise as a routinepartof theirdaily lives: sirens,horns,airplanes,householdappliances,powermowers, ackhammers.Some even seek out noise in the form of loud rockmusic. Peoplecan see a smog-filled skyor a filthylake andtheyrecognizepollution,but noise is not usuallyregardedwithequalconcern.Noise is a form of pollutionand, like otherforms, it'sgettingworse.A U.S. government tudy says that noise pollu-tion is doublingeverytenyears. Says Dr. VernO. Knudsen,a noise expertat the Universityof California:"Ifnoise con-

tinuesto increasefor the next 30 yearsas it has for the past 30, it couldbecomelethal."Questions

Lexicalknowledge:Did the word "doctors" ccurin thisstory?Low-levelproposition:Noise pollutionaffects one's work.High-levelproposition:Peopleare starting o realize the harmfuleffectsof noise.Inference:Moderntechnologyhas contributed o the rise in noise pollution.

Oral and silent reading SALASOO 63


7/12

"Yes/True" esponses were made on the left-handsideof theresponseboxby halfof thesub-jects and on the right-handside by the otherhalf. Inaddition,correctand incorrectquestionversions were counterbalanced among eachgroup. The orderof the four questions, one ateach level, was also randomfor every passagepresentation.Procedure

Subjects were tested individually in asound-attenuatedroom. The instructions, thepromptsduring the course of the experiment,the reading passages, and the comprehensionquestionswere all presentedon a GBC Standardcathode-ray-tube (CRT) display monitor(Model MV-10A) placed at eye level about40 cm in frontof thesubject.Stimuluspresenta-tion andresponsecollectionwerecarriedoutbya PDP-11/34 computer.The subject interactedwith the visual promptsby pressingappropriatebuttons on a seven-buttonresponse box con-nected to the computer.The start of each readingpassage was an-nouncedon the centerof the CRT screenby aprompt: "Attention New Story Coming Up.Please press READY button to begin."Whenthe subjecthad indicatedhis or her readiness,Iannounced he readingmode (oralor silent)forthe passage. The passage appearedon the CRTscreenone sentenceat a time and was advancedby the reader'sbutton-press ontrol.Subjectswere instructed o readeach sen-tence once andthento pressthe "Ready" uttonin orderto continuereadingas fluentlyas possi-ble. The presentationtechnique preventedre-gressive eye movements to previous sentencesand also allowedfor the collection of sentence-by-sentencereadingrates for each test passage.At the end of each passage, the question phasewas announced on the screen by a centeredprompt: "Attention Questions. Please pressREADY button to begin." Subjects initiatedeach questionpresentationhemselvesandwereinstructed o respondas quicklyand accuratelyas possible once the questionshad appearedonthe screen. After making their "Yes/True" r"No/False" esponsesto each question, subjectsenteredconfidenceratingsof theirresponsesona scale from 1 to 7. A ratingof 7 indicated a

highly confidentresponseand a ratingof 1, aguessing response. A ratingscale reminderonthe screenat this time was terminatedwhensub-jects presseda button o select a confidencerat-ing. Then, feedbackaboutthecorrectanswer othecomprehensionquestionwasprovided n theform of a flashinglight immediatelyabovethecorrectbutton on the responsebox. Followingthe fourquestions,thenextreadingpassagewasannounced.Thus, for each passage, sentence-by-sentence reading rates and comprehensiondatain the form of the numberof errors,ques-tion-answering latencies, and confidence rat-ings foreach question ypewerecollected.Three passages served as practice for allsubjects at the beginning of the experimentalsession; the first was readsilently,andthe sec-ond and third, orally. Instructionsemphasizedreading n order o understand hecontentof thepassages. Oral fluency and intonationof read-ing aloud were not mentioned in the instruc-tions. I attempted o minimizeany feelings ofperformance anxiety during the oral readingblocks.

ResultsUnless otherwisenoted,all reported esults

arestatistically ignificant,p < .01.Reading LatenciesA two-wayanalysisof variance(ANOVA),with readingmode and passages as fixed fac-tors, was performed on subjects' mean sen-tence-by-sentence reading times for eachpassage. As expected, subjects took longer toreadpassagesaloud, M = 8.71 s, than to readthem silently,M = 6.95 s, F(1, 15) = 22.06.This difference,M = 1.76 s, was foundfor all16 subjects.Therewere significantdifferencesin reading imesbetween the passages,F(5, 65)= 10.28, reflecting differences in sentencelengths, but these differences did not interactwith the readingmodeeffect, F(5, 65) < 1.0.ComprehensionFor initial ANOVAs, the four questiontypes (lexical, low-level andhigh-level proposi-tional, and inferential)were treatedas a single

64 READING ESEARCHUARTERLY Winter 1986 XXI/1


8/12

factor,question evel. An extensionof thelevelseffect would predictslower, less accurate,andless confident recognition responses to thequestionswithhighercomprehensionevels.Foreach subject,the total numberof errorsfor each question evel were summedfor silentand for oral reading(for a possible total of sixeach). In addition, the mean response latencyand confidence rating for each question levelover all reading passages were computed foreach reading mode condition. SeparateANOVAswiththe orderof readingmodeblocksas a between-subjects actor and questionlevelandreadingmode as fixed, within-subjectsac-tors were performedon the number of errors,mean responselatencies, and mean confidenceratings.Comprehensionrrors.No significant ffectof orderof readingmode was found,so this fac-tor was not included n furtheranalyses.As ex-pected,an overallquestion-level ffect indicateddifferentialerrorpatterns or the various com-prehension evels of the questions,F(3, 45) =8.18. However,no significantdifferenceswerefoundin the numberof errorsmade in oral andsilentreadingconditions,F(1, 15) = 3.54, p >.07. In addition,no significant nteractionwasfound betweenthe level of comprehension ues-tion andthereadingmode,F(3, 45) = 2.22, p >.09. Furtheranalysisof the propositionalmicro-structureyielded the expected levels effect forlow-level and high-level propositionsexplicitlyrepresented n the text, F(1, 15) = 8.04, p .10, further analysessummedover this factor. Significantmain ef-fects of both readingmode, F(1, 15) = 11.52,andquestionlevel, F(3, 45) = 32.13, were ob-tainedin the subsequentanalysiswith the fourquestion types as one factor. Furthermore, ndata for response latencies, readingmode and

questionlevel interacted ignificantly,F(3, 45)= 4.85.No differences between silent and oralreading were found in the latencies for re-sponses to questionsprobinglexical and infer-entiallevels of knowledgegainedfromthe text,F(1, 15) = 2.02, p> .17, andF(1, 15)< 1.0,respectively. In contrast, oral reading led tofasterverificationof propositionsreflectingthemicrostructure nowledge representation f thetexts, F(1, 15) = 27.95. The failureto obtainasignificant difference between response laten-cies for low- and high-level propositionswasunexpected, F(1, 15) = 2.96, p> .10. Thecomprehensionquestionlatencyand errordataare shown in Figure 1.Confidenceratings. Contrary o what onemightexpectfromtheiruse in listeningcompre-hension tasks (e.g., Brunner& Pisoni, 1982),confidence ratingsfailed to reflect meaningful

Figure1Response atencydata(top panel)and errordata(bottompanel)for fourlevels ofcomprehensionquestionsaftersilentreading(filledbars)and oralreading openbars)E SILENTI ORALg 5.0-z0w

I,u 4.0-z

R- 3.0c)W

0.50S 0 . 2 5

LEXICAL LOW HIGH INFERENTIALPROPOSITIONPROPOSITION

LEVELOF COMPREHENSIONQUESTION



9/12

differences between silent and oral readingprocesses.An overallANOVA evealeda signif-icant effect of question level, indicatingdifferences betweenthe four types of compre-hension-levelprobes, F(3, 45) = 9.20, but nosignificant differences due to reading modewereobserved,F(1, 15)< 1.0. The mean con-fidenceratingwas 5.70, wherea ratingof 7 in-dicateda very confidentresponseand a ratingof 1, a guessing response.Subsequentanalysesindicatedno significantquestion-level ffectbe-tween low- andhigh-levelmicrostructure rop-ositions. Thus, subjects' confidence in theirrecognitionor verificationresponses to infor-mation rom textstheyhad ust readappeared obe minimally related to cognitive processesused in readingcomprehension.

DiscussionThe present study investigateddifferencesbetween oral and silent readingusing a struc-turedconceptof text comprehension.Fromtheresults, one can identify specific levels ofknowledge structuresused in text comprehen-sion which were affected by prior readingmode: Subjects correctly recognized low- andhigh-level propositional statements from thetext more quickly after oral readingthan aftersilent reading.However, his effect was not re-flected in subjects'errors in or confidence rat-ings of comprehensionquestions. In contrast,as predictedby previousresearch,the readingmode did affect reading rate. The observedreadingtime variedinverselywith responsela-tencies on low- and high-level propositionalstatements: The mean oral sentence-by-sen-tence reading time was 1.76 seconds longerthanthe mean silentreading ime.Two hypotheses in the literaturerelate totheseresults.According o the first hypothesis,the ongoing vocalizationresponsein oral read-ing requiresattentionalcapacitythat is sharedwith othercognitive processesinvolved n com-prehension(Goodman, 1970b; Wanat, 1976).This view predicts slower and less accuratecomprehension performance in oral reading.

Also, in this view, higherlevels of comprehen-sion should suffermore than lowerlevels in theoral readingmode. In fact, the results of thisstudy contradictboth of these predictionsandmay be taken as support for views that adultreadershaveautomaticpronunciation esponsesthat do not demandprocessing capacityin oralreading(e.g., Danks& Hill, 1981).According to the second hypothesis, theslowerspeedof oralreading compared o silentreading)is accompaniedby greaterrelianceoncontextrather han the printed ext, resulting npoor comprehension.Stanovich 1981) has sug-gested that faster silent readingenables moreeffective use of low- and high-level text infor-mation, leading to superior comprehension.This prediction, again, was not confirmed bythedata.The most importantresult in this study isthat fasterresponsesrelatedto text microstruc-ture followed oral readingthan silent reading.Textdifficultyis implicated n this finding;themean error rate for comprehension questionswas 21%, and many subjectsremarkedon thedifficulty of the experimental ask. It appearsthat the readingmaterialsand/or the compre-hension task in the present studywere difficultfor the subjects.In manytasks, difficultyleadsto slowerhumanperformance.In reading,too,when comprehensions difficult,the perceptualand cognitive processes may be slowed down(Levy, 1981). Additionalsupport or this argu-mentcomes fromreportsof spontaneousvocal-ization during(silent)readingof difficult texts(e.g., Hardyck& Petrinovich,1970). Either hevocalizationof the oral reading responseor itsslower ratemay compensatefor the otherwisefasterencodingof text microstructureor com-prehension n silent reading.These hypothesesconcerningthe role of text difficultyremaintobe testedwitheasy readingpassages.

Oralreadinghad the greatesteffect on textmicrostructure.This suggeststhatthe processesthat integrateinformationwithin and betweenclauses are more influenced by the readingmode than either lexical or inferential proc-esses. Workingmemory processes involved insyntacticparsingandthedevelopmentof propo-sitional knowledge structuresmay receive the

66 READING ESEARCHUARTERLY Winter1986 XXI/I


10/12

benefitsarisingfromoralreading.I suggestthatthe speedof integratingexical (andotherstruc-tural)knowledgeinto higher levels of compre-hension is critical in processing differencesbetween silent and oral reading. Presumably,phonological recoding provides the entry intothis working memory store in both readingmodes. How does this (albeit broad)proposalrelateto the two specific accounts of oral andsilent reading hatdominate he literature?Because Goodman's model of reading(e.g., Goodman, 1969, 1970a, 1970b; Good-man & Goodman, 1977) rests almost entirelyon oral reading error data, Goodman has at-tempted o specify theprocessesin bothreadingmodes in some detail. One point of confusionhas already been alluded to, but at this timeGoodman'sperspectiveshould be examinedingreater depth. For most adult oral readers,Goodmansuggests that "primarily, ral outputis producedafter meaning has been decoded"(1970b, p. 483). Thus, oral reading requiresboth the decoding of the meaningcomponent,which Goodman identifies with silent reading,and then, a derivativerecodingprocess, inter-pretableas phonologicalrecoding, "toproducean orallanguageequivalentof the graphic nputwhich is the signalin reading" 1970a, p. 502).The underlyingcognitiveprocessesleadingto meaning reconstruction (i.e., comprehen-sion) arepresumed o be the sameprocesses inboth silent and oral reading, according toGoodman (1970a). This appears inconsistentwith his assertion of active sampling, predic-tion, andprocessingspeed differences betweenthe two reading modes (cf. 1970a, p. 502).Thus, while Goodman indicates awareness ofdifferences ntherelativeratesof processing,henowhereassociatesthem with readingcompre-hension processes. Therefore, Goodman im-plicitlypredictsno differencesbetweenoral andsilentreading n comprehensionperformance.A more flexible account of differences inoral and silent reading has been proposed byDanks and Fears (1979). Two alternativemodels of the oral readingprocesses are postu-lated-the decoding and comprehension hy-potheses. Both models necessarily includephonological recoding or, as Danks and Fears

name it, decoding (not to be confused withGoodman's imilarmechanismreferred o as re-coding). The Danks and Fears (1979) modelsdifferin thatcomprehensionoccurspriorto thespokenoral reading responseonly in the com-prehensionmodel. Whichof the two models isused dependson variablessuch as the readingskill and motivationof the reader,the specifictask, and the textdifficulty.The presentresultsmay exemplify comprehension-guided oralreading, accordingto Danks and Fears.Again,text difficulty would appear to underlie theresult of faster recognition of microstructurestatements read orally than of those read si-lently.Inorderto gain further nsightintothesemodels, it is necessarybothto specifyin greaterdetailthe processing stagesbetweenphonologi-cal recodingandcomprehensionandto test outthe predictionsabouttext difficultywith easierreadingpassages.Insum, the resultsof this studysuggestthattemporaldifferencesexist betweenthe compre-hension processes for silent and for oral read-ing. This readingmode effect (in both readingrates and in response latencies to comprehen-sion questions) may occur after words havebeen recognized, in the assignmentof meaningand case roles, and in the text-unit ntegrationprocesses that takeplace in workingmemory.Iproposethat the additional ime spent in work-ing memoryduringcompletionof these higherlevel comprehensionprocesses in oral reading(compared to silent reading) may result inmemorytraces that are retrieved aster n later,memory-basedcomprehensionprobes of thesehigher evel unitsof the textmeaning.The studyfound no differences between oral and silentreadingin responses to lexical questions. Onemust be cautious n interpretinghis null result;it may suggest thatlexical access is not the lo-cus of the readingmode effect. This result alsosuggests thathypothesesof oral reading super-iority thatrely on the possibilityof double en-coding into working memory, i.e., from boththe visual and auditory (vocalization) inputs,are untenable:Additionalworking memoryac-tivation would predictlargest oral readingad-vantages for the lexical probes. The lack ofsignificantdifferences found between oral and



11/12

silentreading n response atencies for inferenceprobes suggests that inferences may not bedrawn automatically during on-line readingcomprehensionand may instead be computedfrom stored information upon demand(Frederiksen,1981). As they read for compre-hension, readers construct and store proposi-tional structures of the levels specificallyprobedby the microstructure erificationstate-ments used in this study (cf. Kintsch & vanDijk, 1978). For this reason, additionalwork-ing memory storage duringthe slower clausalintegration processes of oral reading appearslater to facilitate asterverificationonly for low-andhigh-levelmicrostructuresromthetext.In conclusion, the datasuggest thatdiffer-ences dueto readingmode areprimarilya func-tion of the speed of higherlevel integrationandcomprehensionprocessesthatoccur when sub-jects read texts for comprehension. Slowerreading rates in oral reading led to faster re-sponses to comprehensionitems on low- andhigh-level propositional structures from thetext. These results supportmodels of oral andsilentreadingwith commonearlystagesof pho-nological recoding (e.g., Danks & Fears,1979). Such modelsobviouslyhaveyet to iden-tify the detailsof theprocessingandrepresenta-tional levels where time differences n oral andsilent reading will be reflected in memory-based comprehension ests. The present studyalso pointsto the importanceof two theoreticalendeavors:first, extendingunitaryprocessingmodels of reading comprehensionto accountfor temporaldifferencesdue to readingmode,and second, viewing comprehensionas a struc-turedhierarchyof component evels of meaningand structure ather hanas a globalprocessthatsomehowreveals tself aftera readerencountersa printed ext.REFERENCESANDERSON, I.H., & DEARBORN, W.F (1952). Thepsychologyof teaching reading.New York:RonaldPress.ANDERSON, I.H., & SWANSON, D.E. (1937). Common actorsineye movementsn silent and oralreading.Psychologi-cal Monographs,48, 61-69.ANDERSON, .R. (1976). Language, memory, and thought.Hillsdale,NJ: Erlbaum.

BANKS, W.P., OKA, E., & SHUGARMAN, S. (1981). Recodingof printed words to internal speech: Does recodingcome before lexical access? In O.J.L. Tzeng & H.Singer(Eds.), Perceptionofprint:Readingresearch nexperimental sychology (pp. 137-170). Hillsdale, NJ:Erlbaum.BARON, . (1977). Mechanismsfor pronouncing printed

words: Use and acquisition.In D. LaBerge& S.J. Sa-muels(Eds.),Basicprocesses n reading:Perception ndcomprehension.Hillsdale,NJ: Erlbaum.BLANK, M.A., PISONI, D.B., & MCCLASKEY,C.L. (1981). Ef-fects of target monitoring on understanding fluentspeech. Perception& Psychophysics,29, 383-388.BRUNNER, H., & PISONI, D.B. (1982). Some effects of per-ceptual load on spokentext comprehension.JournalofVerbalLearningand VerbalBehavior,21, 186-195.CIRILO, R.K., & FOSS, D.J. (1980). Textstructure ndreadingtime for sentences. Journalof VerbalLearningand Ver-bal Behavior, 19, 96-109.COKE, .U. (1974). The effects of readabilityon oral andsilentreadingrates.Journalof EducationalPsychology,66,406-409.COLLINS,R. (1961). The comprehension of prose materialsby college freshmen when readsilentlyand when readaloud. JournalofEducationalResearch,55, 79-82.DANKS, J.H., & FEARS, R. (1979). Oralreading:Does it re-flect decoding or comprehension? In L. Resnick &P. Weaver Eds.), Theoryandpracticeof early reading(Vol. 3). Hillsdale,NJ: Erlbaum.DANKS, J.H., & HILL, G.O. (1981). An interactiveanalysisoforal reading.In A.M. Lesgold& C.A. Perfetti(Eds.),Interactiveprocesses in reading (pp. 131-154). Hills-dale, NJ: Erlbaum.DE BEAUGRANDE, . (1980). Text, discourse, and process:Toward multidisciplinarycience of texts. Norwood,NJ: Ablex.ELGART,D.B. (1975). Oral reading, silent reading, and lis-tening comprehension:A comparative tudy.JournalofReadingBehavior,10, 203-207.FAIRBANKS, . (1937). The relation between eye-movementsand voice in the oral reading of good and poor silentreaders.Psychological Monographs,48, 78-101.FREDERIKSEN, J.R. (1981). Sources of process interactionsin reading.In A.M. Lesgold& C.A. Perfetti Eds.), In-teractiveprocesses in reading (pp. 361-386). Hillsdale,NJ: Erlbaum.GLUSHKO,R.J. (1981). Principles for pronouncing print:The psychology of phonography. n A.M. Lesgold &C.A. Perfetti (Eds.), Interactiveprocesses in reading(pp. 61-84). Hillsdale,NJ:Erlbaum.GOODMAN,K.S. (1969). Analysis of oral reading miscues:Applied psycholinguistics. Reading Research Quar-terly,5, 9-30.GOODMAN, K.S. (1970a). Reading: A psycholinguisticguessinggame. InH. Singer& R. Ruddell(Eds.), The-oretical models andprocesses of reading(pp.497-508).Newark,DE: InternationalReadingAssociation."3OODMAN,.S. (1970b). Behind the eye: What happens inreading.In H. Singer& R. Ruddell(Eds.), Theoretical

68 READING ESEARCHUARTERLY Winter 1986 XXI/1


12/12

models and processes of reading (pp. 470-496). New-ark,DE: InternationalReadingAssociation.GOODMAN,K.S., & BURKE,C. (1973). Theoretically basedstudies of patterns of miscue in oral readingperform-ance. Washington, DC: U.S. Departmentof Health,Education& Welfare.GOODMAN,K.S., & GOODMAN,Y.M. (1977). Learningabout

psycholinguistic processes by analyzing oral reading.HarvardEducationalReview,47, 317-333.GRAY,W.S. 1956). Theteaching of readingand writing:Aninternational urvey.Paris: UNESCO.HARDYCK, C.D., & PETRINOVICH, L.F. (1970). Subvocal

speech and comprehension level as a function of the dif-ficulty of readingmaterial.Journal of VerbalLearningand VerbalBehavior,9, 647-652.JONES,E.E. (1932). A comparison of comprehension resultsin oral and silent reading. Peabody Journal of Educa-tion, 9, 292-296.JUEL, C., & HOLMES,B. (1981). Oral and silent reading ofsentences.ReadingResearchQuarterly,16, 545-568.JUST, M.A., & CARPENTER, .A. (1980). A theory of read-

ing: From eye fixations to comprehension. Psychologi-cal Review,87, 329-354.KINTSCH,. 1974). Therepresentation f meaning n mem-ory. Hillsdale, NJ: Erlbaum.KINTSCH,w. (1977). Memory and cognition. Hillsdale, NJ:Erlbaum.

KINTSCH,W., & KEENAN,J.M. (1973). Reading rate as afunction of the number of propositions in the base struc-ture of sentences. CognitivePsychology,5, 257-274.KINTSCH,W., KOZMINSKY,., STREBY,W., MCKOON,G., &KEENAN,J.M. (1975). Comprehension and recall of atext as a function of content variables. Journal of VerbalLearningand VerbalBehavior, 14, 196-214.

KINTSCH,W., & VANDIJK,T.A. (1978). Toward a model oftext comprehension and production. Psychological Re-view, 85, 363-394.KLEIMAN,G.M.(1975). Speech recoding in reading. Journalof VerbalLearningand VerbalBehavior, 14, 323-339.LABERGE,D., & SAMUELS, .J. (1974). Toward theory ofautomatic nformation-processingn reading.CognitivePsychology,6, 293-323.LEVY,B.A. (1981). Interactive processing during reading. InA.M. Lesgold& C.A. Perfetti Eds.), Interactive roc-esses in reading(pp. 1-36). Hillsdale,NJ: Erlbaum.MEAD, .D.(1915). Silent versus oral readingwith one hun-dred sixth-grade children. Journal of Educational Psy-chology, 6, 345-348.MEAD,C.D. (1917). Results in silent versus oral reading.Journalof EducationalPsychology,8, 367-368.MINSKY,M. (1975). A framework for representing knowl-edge. In P Winston(Ed.), Thepsychology of computervision. New York: McGraw-Hill.MINSKY,M. (1980). K-lines: A theory of memory. CognitiveScience, 4, 117-133.POULTON, .C., & BROWN,C.H. (1967). Memory after read-

ing aloud and reading silently. British Journal of Psy-chology,58, 219-222.RAYNER,K., & MCCONKIE,.W. (1976). What guides a read-er's eye movements? Vision Research, 16, 829-837.

ROGERS,M.V. (1937).Comprehension in oral and silentreading.Journalof GeneralPsychology,17, 394-397.RUMELHART, .E. (1977). Toward an interactive model ofreading.In S. Dornic (Ed.), Attentionandperformance(Vol. 6, pp. 573-603). Hillsdale, NJ: Erlbaum.SCHANK,R. (1975). Conceptual dependency theory. In R.Schank,N. Goldman,C. Rieger& C. Riesback(Eds.),Theoretical ssues in natural language processing: Aninterdisciplinary workshop. Cambridge, MA: Bolt,Beranek & Newman.SCHANK,. (1981). Language ndmemory. nD.A. Norman(Ed.), Perspectiveson cognitivescience (pp. 105-146).Norwood, NJ: Ablex.

SPOEHR,K., & SMITH,E. (1975). The role of orthographicandphonotactic ules in perceiving etterpatterns.Jour-nal of Experimental sychology:HumanPerceptionandPerformance,1, 21-34.STANOVICH, . (1981). Attentional and automatic contexteffects in reading. In A.M. Lesgold & C.A. Perfetti(Eds.), Interactiveprocesses in reading(pp. 241-268).Hillsdale,NJ: Erlbaum.SWALM,.E. (1973). A comparisonof oral reading,silentreading, and listening comprehension. Education, 92,111-115.TULVING,. (1976). Ecphoricprocesses in recall and rec-ognition. In J. Brown (Ed.), Recall and recognition.London: Wiley.TURNER,A., & GREENE,E. (1977). The constructionanduse of a propositional text base (Tech. Rep. No. 63).Boulder,CO: Universityof Colorado,Institutefor theStudyof IntellectualBehavior.VAN DIJK, T.A. (1979). Macro-structures. Hillsdale, NJ:Erlbaum.WANAT,.E (1976). Readingreadiness.VisibleLanguage,10, 101-127.WEBER,R.M. (1968). The study of oral reading errors: Asurveyof the literature.ReadingResearchQuarterly,4,96-119.

FootnotesThis research was supported by NIMH research grantMH 24027 to Indiana University in Bloomington. I amgratefulto David B. Pisoni for the use of the Speech Re-searchLaboratoryacilities and for insightfulcommentsonan earlier draft, to Jerome C. Harste for helpful discus-sions, and to Hans Brunner for access to the experimentalmaterials. Reprint requests may be sent to Aita Salasoo,Department of Psychology, SUNY-Binghamton, Bingham-ton, NY 13901.'Unlike many experimenters who have employed nonspe-cific or low-level comprehension tests, which test primarilymemory for text vocabulary (e.g., Swalm, 1973) or ambig-uous levels of comprehended knowledge such as "informa-tion in the story" (Collins, 1961), one early researcher,Mead (1915, 1917), adopted a measure reported as "per-centage of points reproduced of points read." This measureis a precursor, albeit nonspecific, of the current use of prop-ositional text-based analyses to measure comprehension.


Documents

Cognitive Processing