Generative Learning of Reading Comprehension

8/6/2019 Generative Learning of Reading Comprehension

http://slidepdf.com/reader/full/generative-learning-of-reading-comprehension 1/15

The Teaching of Reading Comprehension according to the Model of Generative LearningAuthor(s): Michele Linden and M. C. WittrockSource: Reading Research Quarterly, Vol. 17, No. 1 (1981), pp. 44-57Published by: International Reading AssociationStable URL: http://www.jstor.org/stable/747248 .

Accessed: 14/04/2011 19:57

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless

you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you

may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at .http://www.jstor.org/action/showPublisher?publisherCode=ira. .

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed

page of such transmission.

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 Association is collaborating with JSTOR to digitize, preserve and extend access to

Reading Research Quarterly.

http://www.jstor.org

http://www.jstor.org/action/showPublisher?publisherCode=ira

http://www.jstor.org/stable/747248?origin=JSTOR-pdf

http://www.jstor.org/page/info/about/policies/terms.jsp



http://www.jstor.org/page/info/about/policies/terms.jsp

http://www.jstor.org/stable/747248?origin=JSTOR-pdf




44

The teaching of reading comprehension

according to the model of generative learning

MICHELE LINDEN

California tate University,Northridge

M.C. WITTROCK

University f California,LosAngeles

AN INSTRUCTIONALEQUENCEerived from Wittrock's model ofgenerativelearningwas presentedin classrooms for three days to 58

ten-year-old children.The purpose of the studywas to determine the

utility of the model for teaching reading in elementary schoolclassrooms. Employing a unifactor, four-treatment design with

participants individually and randomly assigned to the treatments,the data supported (p<.01) the hypothesis that, with time held

constant, children instructed to generate associations for the text

during reading show greater comprehension of that text than dochildren not instructed to generate the associations. The findings

also confirm the hypothesis that these instructions to generateassociations for the text increased the number of text-relatedassociations produced during learning, which apparently led toincreases in fact retention and story comprehension. The data

support predictions from the model of generative learning andindicate its utility for improving the teaching of reading in

elementary schools.

L'enseignementde la comprdhension de lecture d'apres le

modled'dtudegdneratriceONA PRPISENT•I

ans des classes une sequence d'enseignementd6riv6eedu module d'6tude g6n6ratrice de Wittrock pendant trois joursdevant 58 enfants ag6s de dix ans. Le but de cette etude 6tait de

d6terminer 'utilit6du modulepour l'enseignementde la lecture dansles cours 616mentaires.En employant un modulede facteur uniquede quatretraitementsavec des participantsassign6saux traitementsindividuellement et au hasard, les donn6es soutiennent (p < .01)

l'hypoth6se suivante: dans une dur6e maintenue constante, lesenfants auxquels on a demand6 de g6n6rer des associationsconcernant le texte

pendantla

lecture,montrent une

plus grandecomprehension du texte que les enfants auxquels on n'avait pas

demand6 ces associations. Les d6couvertes confirment 6galementl'hypoth6sesuivante:ces instructionsencourageantdes associationsconcernant le texte ont augment6 le nombre d'associations ayantrapportau texte produites pendant la lecture, qui apparemmentont

men6 a une augmentation de la m6morisation des faits et de la



The teaching of reading comprehension LINDEN WITTROCK 45

comprehension u r'cit. Les donn'es soutiennent es pr6visionsgtpartirdu moduled'etudeg6neratricet indiquent on utilit6pourl'ameliorationde l'enseignementde la lecture dans les cours

616mentaires.

La ensenianzade comprensidn de lectura

segtunel modelo de aprender generativo

SE PRESENTOuna secuencia de instrucci6n, derivadadel modelodeWittrockde

aprender enerativo,a 58

nifiosde 10 afios

durante3dias en clase.El objetivodel estudioeradeterminara utilidaddelmodelo para la ensefianzade lectura en la escuela elemental.Empleandoun solo factorcon disefiode tratamiento uadruplecon los participantes signadosindividualmente al azar en lostratamientos,os datosrespaldaronP < .01)la hip6tesisque,conel tiempo mantenidoconstante, los nifios instruidosa generarasociacionesparael texto durante a lectura,demostraronmayorcomprensi6ndel texto que los nifios no instruidos a generarasociaciones.Los resultadostambi6nconfirman a hip6tesisqueestas instruccionespara generar asociaciones para el texto,aumentaronel n(imerode asociacionesrelacionadas on el textoproducidasduranteel aprender,que aparentementelevarona unincrementode retenci6nde los hechos y de comprensi6nde lahistoria. Los datos sostienen las prediccionesdel modelo deaprender enerativo indican uutilidadparamejoraraensefianzade lecturaen las escuelaselementales.

In Wittrock'smodel of generativelearning(1974, 1978a, 1980,1981), reading comprehension occurs when readers build relationships

(1) betweenthe text and theirknowledgeandexperience,and(2)amongthedifferent parts of the text. Severalstudies with school childrenandadults

support the utility of this model of generativelearningforunderstandingreading comprehension, fact retention, and their facilitation (Bull &

Wittrock, 1973;Doctrow, Wittrock, & Marks, 1978;Wittrock & Carter,1975; Wittrock, Marks, & Doctorow, 1975).

According to this model of learning, teachers can facilitate

reading comprehension by inducing the readers to attend to the text, torelate their knowledge and experience to it, and to build associations,

abstractions, and inferences from it. The generation of associations andrelations can be taught in a variety of ways, such as by inducing thelearners to generate text-relevant summary sentences, headings,inferences, main ideas, criticalcomments, and evaluations. In addition to

generating verbal representationsfor the text, the readerscan be taughtto construct imaginal representations such as pictures, images, graphs,



46 READINGRESEARCH UARTERLY Number 1, 1981 XVII/I

illustrations, diagrams, and drawings (Kaplan, 1971). In the earlierexperiments these verbal and imaginal generative activities have

consistently produced gains in comprehension and fact retention when

they have been studied in the laboratory or in the classroom (e.g., Bull

and Wittrock, 1973; Doctorow, Wittrock, & Marks, 1978; Wittrock,

Marks, & Doctorow, 1978). However, these earlier studies were usually

designed to isolate the effects of generative learning activities, not to

approximate the complexity of a sequence of several days of classroom

teaching.

In the following experiment we again studied the teaching ofreading comprehension according to the model of generative learning.However, we also included in the experiment several conditions that

closely approximated those often found in elementary school class-

rooms. We included classroom teachers, instead of written instructional

materials and written directions only. In addition, the instruction in the

children's school extended over several days, instead of lasting only a

single day. The instruction also progressedacross severaldays from one

typeof

generativeactivity to another type, such as from imaginal to

verbal activities.These proceduresprovided the opportunity to examine

the power and the utility of generative teaching activities in realistic,

complicated school settings.At the same time we wanted to retain control across the

treatmentgroups of the teachervariable, the stories readbythe children,the tests over the stories, the time to learn,andthe children'sability level,socioeconomic background, and ethnic composition. For these reasons

we randomly assigned individual children to the treatments and held

constant the teacher, stories, tests, and school setting. Only thegenerativeactivities and their sequencevaried in conceptuallydistinctive

treatments that extended over several days of instruction in the

classroom.

We also retained a procedure for quantifying generative

learning. The model of generative learning maintains that teaching can

be understood only by knowing the generations it induces in the learners

at the time they learn, not by the nominal stimuli of instruction, i.e., not

by the text or the teacher'sactions alone. Consequently, we devised and

used a way to record the numberand quality of each child'sgenerationsin the two experimental groups, as they read and learned the stories

presented during the three days of instruction. These recordsof student

generations during instruction defined the functional treatments in the

study, and insured that the teacher actually induced the intended

differencesamong the treatments. In educational researchon classroom



The teaching of reading comprehension LINDEN& WITTROCK 47

teaching, it is not adequate to assume, without appropriate measures,that instruction produces the intendedstudent learningactivities, suchas

the generation of summaries,pictures,and inferences.For this reasonwe

obtained objective measures of the number and quality of learner

generations. They index the existence of functional treatments,

independently of their effects upon learning, retention, and compre-hension.

From the model of generative learning and from the studies

cited earlier,four hypotheses were derived.The first hypothesis was that

children of about 10yearsof age increase theirreadingcomprehensionoftext when they generate text-relevant images, illustrations, analogies,

metaphors, or summary sentences as they read. Text-relevant genera-tions include the development of relations among the parts of the text or

between the text and the learners'knowledge and experience.To test this

hypothesis it is necessary to establish that the treatments varied the

number of text-relevant generations constructed by the learners.

The second hypothesis is that childrenapproximately 10yearsold construct more text-relevant

generations and,as a

result,comprehend text better when teaching proceeds from imaginal to verbal

generative activities, rather than from verbal to imaginal generativeactivities. We have no relevant data to support this hypothesis directly.However, after age 8, children use generated images to enhance their

learning (Pressley, 1977). Imagery makes learning specific and concreteand is a preferredmode of storage for childrenat this age (Paivio, 1970;Rohwer, 1970). The addition of verbal labels to images or picturesfacilitates learning for children at this age (Paivio, 1970), although the

reversesequence may facilitate learning as well. Although none of thesefindings leads to the second hypothesis, taken together they suggest that

imaginal generations followed by verbal generations approximate theorder in which children of this age acquire text-relevant associations.

The third hypothesis is that, when comparedwith generationsnot relevant to the text, text-relevant generations enhance readingcomprehension. This hypothesis is based upon earlier findings andrelatedliterature(Wittrock, 1981)that indicatethat generationsfunction

by inducing the learners to construct relations among the parts of the

text, such as between events and characters in a story, betweenparagraphs and the theme of the story, or between the text and their

knowledge and experience.The fourth hypothesis is that the number of text-relevant

generations correlates positively with reading comprehension. This

hypothesis closely relates to the first hypothesis and provides separate



48 READINGESEARCHUARTERLYNumber 1, 1981 XVII/1

tests, within the two experimental treatments and across the fourtreatments,of the relation betweenthe amount of studentgenerationand

reading comprehension.

Method

Experimental DesignA unifactor, four-treatment design with individual random

assignmentof participantsto conditions was used. Instructionaltime was

held constant across all groups.

Participants

A total of 64 children, 34 boys and 30 girls, the entire

population of two fifth-grade public school classrooms, comprised the

sample of learners who participated in the study. Most of the children

were Hispanic. About 75% of them spoke Spanish and English. Both

classes came from the same school, located in a lower-middleto middle

socioeconomic level neighborhood in Los Angeles, California.Six childrenwere dropped from the studybecauseof absences

from one or more sessions during the experiment, resulting in a final

sample of 58 children, 30 boys and 28 girls.

Procedure

The following procedures were developed from pilot studies

using a different sample of fifth-graders from a nearby school. The

childrenwereindividually randomly assigned

within sex to thefollowingfour treatmentgroups:(1) Imaginal to Verbal Generations, (2) Verbalto

Imaginal Generations, (3) No Instructions to Generate, and (4)Classroom Teacher Taught Control Group. The time given to teachingand to testing was held constant acrossthe four treatmentgroupsat three

hours, given in three one-hour sessions during the first or second class

periods, over three consecutive days. One week intervenedbetween each

treatmentgroup. To control and to monitor carefullythe administration

of each treatment, the children were taught in groups of eight or less.

The children'sregularreadingteacherpresented he instructionin the teacher-taught control group, Treatment 4. In the other three

treatmentgroups, the senior author, who is also an experiencedreading

teacher, provided the instruction.

On Day 1, the firstday of instruction andtesting, each child in

each treatment read one of the following three stories, describedlater in




the materials section: (1) "The City Man Who Looked for a Farm,"(2)"Alone on MiseryIsland,"and(3) "Tekana'sPup." Immediatelyafterthe

instruction in each treatment group, two tests, a test of factual

information and a test of story comprehension, were given to each child.

These tests are also described in the materials section. Time to learn the

story, 45 minutes, and time to take the tests, 15 minutes, were held

constant across all groups, on all days.On Days 2 and 3 the childrenin each treatmentgroupreadthe

two remainingstories, one per day, and were given the respectivefactual

information and story comprehension tests under the same conditionsdescribed for Day 1.

The differences among the four treatments consisted of the

generative activities the children performed each day as they read each

story duringthe instructional period. A detailed outline of the generativeactivities performedby the learners and of the procedurefollowed bythe

experimenter appears in Linden (1979). In the first treatment group,Imaginal to Verbal Generations, on Day 1 the children generated,

illustrated,and named

imagesfor

StoryI

immediatelyafter

theyread it.

On Day 2, the children in this treatment composed and wrote summarysentences as they read Story 2 listed above. On Day 3 the children

generated and wrote metaphors and analogies as they read Story 3.More precisely, the procedure followed on each day of the

Imaginal to Verbal treatment was as follows. On Day 1the teachergaveeach child a story to read. She said to the group of children, "Here is a

story for you to read. As you read it, make pictures in your mind of

everything that is happening in the story."

After the children had read the story, the teacher askedeach ofthem to draw the images or pictures they made as they read. The

drawings were collected, and the fact-retention and comprehensiontestswere administered.An example of a fact-retentiontest item is as follows."State one reason Father gave for not wanting to stop at the truckfarm."One of the comprehensiontest items was:"Did Father reallywant to buya farm? How can you tell?"

On Day 2 the same procedure was followed prior to and

duringthe reading.After the childrenhad read the story, the teachersaid,

"Today we will divide the story into three sections. Write one or twosentences to summarize each section of the story."She then gave several

examples of summaries."Let'sdo the summaries now. Remember, each

summary sentence tells what that section of the story is about. Eachsentence is like a title." The group then divided the story into sections.Then each pupil composed the summary sentences. As in Day 1, a



50 READING RESEARCHQUARTERLY * Number 1, 1981 XVII/1

comprehension and a fact-retention test were given over the story justread.

On Day 3, again the same procedure used on Day 1 was

followed prior to and duringthe readingof the story. After the storyhad

been read, the teacher said, "Yesterdaywe talked about a fog and how it

was like a darkenedroom. This description is very interesting,because a

dark room and a fog are really not the same thing. They are alike,

however, in several ways-very often a description of one thing will

remind you of something else, something you may know more about."

First orally and then in writing, the pupils identified and describedanalogies and metaphors involving the story and their own experiences.

Again, as on Days 1 and 2, tests of comprehension and fact-retention

were administered.In the second treatment group, Verbal to Imaginal Genera-

tions, the three-day sequence of activities and the order of the readingof

the three stories describedfor Treatment 1were reversed.On Day 1, the

children in the second treatment group generatedand wrote metaphorsand

analogiesas they readStory 3. On Day 2, these childrenreadStory 2

and constructedand wrote summarysentences for the paragraphsof the

text as they read it. On Day 3 the children in Group 2 read Story 1 and

generated, illustrated, and titled imagesfor the text aftertheyreadit. The

order of the stories was reversed to match the reversed order of the

generative activities to hold constant across these two treatments the

stories paired with each type of generative treatment. This procedureinsured that these two treatments differed from each other only in the

order of generativeactivities, not also in the story used as the context for

each task. In the third treatmentgroup, No Instructionsto Generate,the

same stories were used, and in the same sequence, as in Treatment 1.

None of the instructionsregardinggenerativeactivitieswas mentioned to

the children.Instead, the instruction focused upon conventional reading

techniques and objectives: identification of main ideas, events, and

characters;vocabulary;categorization skills;discriminationof syllables;and phonetic analysis. Instruction in phonetic analysis occurredon each

of the three days. The remainingactivities occurredin at least one of the

three days of instruction. Treatment Group 3 provides a control tomeasurethe effects of generativeactivities, using the same teacher as was

used in Treatments 1 and 2.In the fourth treatment group, the Control Treatmenttaught

by the children'sregular readingteacher, the instructionwas deliberatelyleft to the classroomteacher'sdiscretion.The intent of this procedurewas



The teaching of reading comprehension LINDEN & WITTROCK 51

to provide a basis for comparing the results of the experimentalprocedures with more conventional teaching techniques, even thoughthese treatments might differ from one another in multiple ways.

The procedure used by the regular classroom teacher was

observed and recordedby the senior author throughout the threedays of

teaching and testing of Treatment Group 4. On Day 1, the childrenread

Story I aloud and then answered aloud the teacher'squestions about the

story, its main characters,and its moral. On Day 2, the same procedurewas followed, except that reviews of Story 2 were interspersedbetween

the readings of sets of two or three paragraphs.As part of these reviewsthe childrenwere asked what they thought would happennext, andwhat

they would like to have happen next in the story. A total of 17 student

generations of story summaries were recordedin this session. On Day 3,the classroom teacher markedly changed her style of teaching. The

posttest for the day's story, Story 3, was handed to each child at the

beginning of the hour along with the story. Each child was instructed to

read the test first, the story second, and then to answer the test questions

directlyfrom the text of the

story.With the exception of Day 3 of Treatment 4, the appropriatefactual information test and comprehension test were given to each childon each day after the instruction was completed. Each childwas given 15minutes to complete both tests.

Materials

The three stories mentioned in the procedure section were

derived from the reading series entitled Be a Better Reader:Foundation

A (Smith, 1968). Each story was approximately 1100 to 1300 words inlength and was written at the fourth-grade level. The data from a pilotstudy for this experiment showed that fourth-grade reading materialswere best suited to the learners in this study.

The two types of posttests used in this study, multiple-choicetests of factual information and completion tests of reading compre-hension, were prepared as follows. The senior author wrote six tests, afact-retention test and a comprehensiontest for each of the threestories.Each test was then given to three judges, who were asked to rate the

relevanceor congruence of each test item to its respectivestory. All itemsused in each test were rated as congruent to the text by all threejudges.The scores on the three fact-retention tests, consisting of 9, 12, and 17

items, given to each child, were summed to provide one measure of factretention that had a possible score of 38 correct.Similarly, the three testsof comprehension, consisting of 14, 15, and 22 items, were summed to



52 READINGRESEARCHUARTERLY Number 1, 1981 XVII/1

provide one measure of story comprehension with a possible score of 51correct.

In addition to the fact-retention and comprehension tests,measures of the number and type of generations the learnersproduced

during instruction were also obtained. The number consisted of a count

of the total sentences and pictures each child wrote or drew. The

relatednessof each of these generationsto the text was evaluatedbythree

additional judges who scored each generation as relevant only if it met

the following three conditions: (1) It contained at least one elementof the

text, such as a character or object;(2) it contained at least one event oractivity of the text; and (3) it described a relationship between the text

and the child's knowledge and experience. All generations produced bythe children were scored by all three judges as relevant to the text

according to these three criteria.

Results

Table 1 gives the means and standard deviations of

generationsproduced duringinstruction, the scores on the fact retention

test, and the scores on the comprehension test. A planned comparisontest on the data inTable 1indicatesthat the treatmentsvaried the number

of generations produced and measuredduringteaching, as planned (df=

54, t = 3.54, p <.01). The two experimental treatments each produced a

mean of 10 or more generations.The two control groups each produced1.2 or fewer generations.The threejudges found all of the generationsto

be text-relevant, using the three criteria describedin the method section.

Table1. Means and standarddeviationsof the generation,retention,and

comprehensioncoresof the treatmentgroups

Means and Standard Deviations

Treatments Number of Fact Retention ComprehensionGenerations (38 items) (51 items)

Imaginal to Verbal 13.0 27.63 28.63

Generations (n = 16) (4.8) (3.34) (6.32)

Verbal to Imaginal 10.7 23.29 31.28

Generations (n = 14) (2.5) (3.22) (5.73)

No Instructions to 0.0 25.14 17.71

Generate (n = 14) (0.0) (3.11) (8.59)

Classroom Teacher 1.2 21.57 21.57

Taught Control (n = 14) (1.9) (7.12) (12.65)




Because the treatmentgroups substantiallyvaried the numberof generations during instruction, it became possible to test the

hypotheses about the effects of generations upon comprehension. The

following statistical analyses were chosen to test three of the hypothesesstated in the introduction, all of which were supported.

A one-way analysis of variance indicated a statistically

significant difference among the treatment groups on the fact retention

test, F(3, 54) = 5.01, p<.01, and on the comprehension test, F(3, 54) =

7.39, p <.01. The correlation between the number of generations and

comprehension was .44 (n = 58, p<.01) for all treatments combined.The first hypothesis stated that the learners' generation of

text-related images, illustrations, metaphors, or summary sentences

during reading facilitates comprehension of the text. A planned

comparison test contrasting the combined mean comprehension scores

of the two experimental groups with the combined mean comprehensionscores of the two control groups clearly supported this hypothesis (df=54, t = 4.51, p<.01). From Table 1, the data indicate that, withoutinstructions to

generatesummariesand the

like,about 18items out of 51

were answeredcorrectlyon the average bythe childrenin this study. Withthe generation of text-related summaries, pictures, analogies, and

metaphors the average comprehension score was about 29 or 31,

depending upon the sequence of the instruction. The childrentaught bytheir regular reading teacher averaged about 22 items correct on thissame test. Instructional time was held constant across all four treatment

groups.The second hypothesis stated that a sequence of learner

activities that proceeds from imaginal to verbal generations producesmore text-related elaborations and greater comprehension than does a

comparable sequence that proceeds from verbal to imaginal activities. A

planned comparison test supportedthe first half this hypothesis (df= 54,t = 2.9, p <.01), although the mean difference between the two

experimental groups' number of generations was only 2.3. The secondhalf of this hypothesis was not supported. The imaginal to verbal

generative treatment did not increase comprehension.The third hypothesis predicted that text-related generations

enhance comprehension more than do generations not related to text.This hypothesis was not testable because the judges rated all of thelearners'generations in all treatment groups to be text-related.

The fourth hypothesis was that the number of text-related

generations correlates positively with reading comprehension. Acrossthe four treatment groups, the correlationbetweengenerationsand com-



54 READINGESEARCHUARTERLYNumber 1, 1981 XVII/1

prehension was .44 (p<.01), indicating support for this hypothesis.Within each of the two experimental groups, the correlation between

generation and comprehensionwas .57 and .24respectivelyfor the verbalto imaginal and imaginal to verbal groups.

No predictions about the retention of facts produced by thelearner activities of the treatment groups were made. However, asmentioned earlier,the data indicated a statistically significant difference

among the four treatment groups in fact retention (p< .01). The patternof the mean differences in retention

amongthe treatments

differs fromthe respective pattern found on the comprehension test. On the fact-retention test the group taught by the classroom teacher produced the

lowest mean, while the group given no instructionsto generateproduceda mean score between the means of the two experimental groups. The

difference in pattern of results is itself an interesting finding that,

contrary to several earlier studies, indicates that certain generativeactivities in some circumstances differently influence retention and

comprehension.

Discussion

The purpose of this study was to investigate in elementaryschool classrooms the effects of several days of sequenced generative

teaching activities, given by a classroom teacher,upon children'sreading

comprehension. According to the model of generative learning

(Wittrock, 1974) we predicted that teaching which induces learners to

perform generations relatingthe partsof the text to one another, or to the

readers' background and experience, enhances comprehension. In two

experimental treatments, a classroom teacher successfully inducedchildren to generate metaphors, analogies, summaries, pictures, and

inferences as they read three stories. Compared with the results of two

control groups, each of these generativeteaching procedures, imaginalto

verbal generations or verbal to imaginal generations, increased the

number of student generations and enhanced comprehension (p<.01)from a mean of about 18 for the control group to means of 29 and 31 for

the experimental groups. These sizable mean differences in compre-hension

compare closelyto the

50%or

greatermean increases found in

earlierexperiments on readingcomprehension with generativelearning.However, in this study the teaching occurred under more realistic

classroom teaching conditions, with a teacher presentingthe instruction

to small groups of children. Across the four treatment groups, the

number of generations correlated positively, as predicted, with reading



The teaching of reading comprehension LINDEN & WITTROCK 55

comprehension (r = .44, p <.01), again supporting the hypothesis that

generations and comprehension are related to each other.

The improvement in comprehension attributedto generative

teaching cannot be explained by factors such as the teacher variable,

length of instruction, time of day of instruction, stories read by the

children, types of generation, such as pictures or summary sentences ininteraction with the stories, and differences in the ability level of the

students. All of these factors, which commonly confound the results of

many studies in classrooms, were held constant across the treatmentgroups, to which the children were randomly assigned individually.

Another part of the procedure of this study that differs from

that of many studies of classroom learning is the measurement of the

differences among the treatments. As in several earlier studies (e.g.,Doctorow, Wittrock, and Marks, 1978)we defined the treatmentsbythe

quantified generative activities performed by the learners duringinstruction, not by the teachers' actions, directions in the text, or other

nominal stimuli presentedto the participants.This single modification in

procedure enables educational researchers n many contexts to establishobjectively that their intended treatmentsactually existed. One does not

have to infer or to guess whether the learners followed the directions,

performed the behavior, or engaged in the cognitive processes that the

treatments were intended to induce.The measuresof learnergenerations

clearly indicate, independently of the retention, comprehension,or other

dependent measures obtained later on separatetests, the existence of the

functional treatments. Especially in educational research focused on

cognitive learning, or in research concerned with explaining how

learning occurs in response to the environments of instruction, thesemeasures of learner activities are useful to obtain.

We also examined the effects of two different sequences of

instruction, imaginal to verbal and verbal to imaginal generations, uponthe number of student generations and upon comprehension. Althoughthe model of generative learning led to no prediction about the

differences in comprehension or number of generations produced bythese two treatments, related studies implied that the imaginal to verbal

sequence enhances performance on these two measures. Only theprediction about the number of generations was statisticallysupported.However, the actual difference in numberof generations,about two, wassmall. From these results, it seems that the generation of text-relevant

associations, in either order,was the moreimportantfactor in enhancingcomprehension in this study.



56 READINGESEARCHUARTERLY Number 1, 1981 XVII/ 1

The retention of facts was not enhanced, nor lowered, bygenerative teaching activities. In severalearlierstudies (e.g., Wittrock et

al., 1975) generativeteaching activities have also enhanced the retentionof factual information. This difference in results between the classroom

based, teacher-taught treatments of this study and treatments used inearlier studies is not easily explained. The teacher of this study mighthave emphasized generations of a differentsort than those inducedin the

earlier studies. The resolution of this dilemma depends upon the

development of a method of quantifying the generationsproducedin thelearnersin control groups not asked to generate or to make explicit therelations they construct as they read text.

The children's regular classroom teacher taught one of thetreatments in this study. Each of the two experimental treatments

produced considerablygreatercomprehensionand retentionthan did the

children's regular classroom teacher, who used whatever instructional

techniques she chose to includein herteaching. Because the time to teach,the stories read, and the ability levels of the randomly assigned children

were held constant in this study, the results indicate that the generativeteaching techniques have the potential of improving instruction in

comprehension as it now exists in some schools. One teacher's

performance in one control group does not prove the relative

effectiveness of generative teaching, but it offers a possibility worth

further study.Last of all, it might be argued that the differences among the

treatments occurred because the teacher of the two experimentalgroupsand one of the control groups knew the design of the study, including its

predictions, and may have inadvertently positively influenced the

experimental groups and negatively influenced the first control group.The measures of student generations establish the existence of the

functional experimentaltreatments;but theycannot refutethis argument

regarding the first control group. However, the performance of the

second control group does refute this argument. The argument is not

highly plausible because of the similarityin comprehension between the

first control group and the second control group, in which the teacher

washighly

motivated tosucceed,

but did not enhancecomprehensionnor, as quantified by the measures of generative teaching, frequently

induce the learners to engage in generative learning.Taken together, the data of this study indicate that without

any increase in the time given to instruction, reading comprehension

among 10-year-oldchildrencan, sometimes at least, be enhancedsizably




in the classroom by generativeteaching activitiesthat inducethe learnersto construct analogies, summaries,pictures, and inferences as they read.

These data support Wittrock's model of generativelearningand indicate

that it may have practical utility in elementary school classrooms when

the instruction is presented by a teacher.

Now that this study and its predecessors (e.g., Doctorow,

Wittrock, & Marks, 1978; Wittrock, Marks, & Doctorow, 1975)indicate that generative teaching procedures facilitate reading compre-hension, sometimes at least, in elementaryschools, a next and logical stepis to try to teach children strategies for generating text-relevant

representations as they read new stories or text. For future studies of

generativeteachingwe suggesttreatmentsdesignedto teach children how

to learn and to apply general strategies for constructing text-relevant

metaphors, pictures, summaries, titles, and similar representations as

they read new materials.

REFERENCES

BULL, B.L., & WITTROCK, M.C. Imageryn the

learningof verbaldefinitions.British Journal

of Educational Psychology, 1973, 43, 289-293.

DOCTOROW,M.J., WITTROCK,M.C.,& MARKS,C.B.

Generative processes in reading compre-hension. Journal of Educational Psychol-ogy, 1978, 70, 109-118.

KAPLAN,.A.Elaboration and reading achieve-ment as factors in children'slearningof textmaterials. Report No. C-03, WisconsinResearch and Development Center for

Cognitive Learning, University of Wiscon-sin, 1971.

LINDEN,M. An instructional sequence forteachingreadingderivedfrom thegenerativemodel of learning. Unpublished doctoral

dissertation, UCLA, 1979.

PAIVIO,A. On the functional significance of

imagery. Psychological Bulletin, 1970, 73,385-392.

PRESSLEY,. Imagery and children's learning:Putting the picture in developmental per-spective. Review of Educational Research,1977, 47, 585-622.

ROHWER, .D.Images and picturesin children's

learning. Psychological Bulletin, 1970, 73,393-403.

SMITH,N.B. Be a better reader:Foundations A.

Englewood Cliffs, N.J.: Prentice-Hall, Inc.,1968.

WITTROCK,M.C. Learning as a generativeprocess. Educational Psychologist, 1974,11,87-95.

WITTROCK,M.C. The cognitive movement ininstruction.Educational Psychologist, 1978,13, 15-30. (a)

WITTROCK,M.C. Developmental processes in

learningfrom instruction.Journal of Genet-ic Psychology, 1978, 132, 37-54. (b)

WITTROCK, .C. Learning and the brain. InM.C. Wittrock (Ed.), The brain and psy-chology. New York: Academic Press, 1980.

WITTROCK, .C.Reading Comprehension. InF.J. Pirozzolo & M.C. Wittrock (Eds.).Neuropsychological and cognitiveprocessesin reading. New York: Academic Press,1981.

WITTROCK,M.C., & CARTER, . Generative

processingof hierarchically rganizedwords.American Journal of Psychology, 1975, 88,489-501.

WITTROCK,M.C., & GOLDBERG, . Imagery and

meaningfulness in free recall: Word attri-butes and instructional sets. Journal ofGeneral

Psychology, 1975, 92,137-151.

WITTROCK,M.C., MARKS,C.B.,& DOCTOROW,M.J.

Reading as a generative process.Journal ofEducational Psychology, 1975, 67, 484-489.

Documents

Generative Learning of Reading Comprehension