ds.

THE °EFFECTS OF DEEP .STRUCT,UREVARIATIONS IN SENTENCES:FREE RECALL ANDPAIRED 7- ASSOCIATE LEARNING

,

e

/ l.

at

A

WISCONSIN RESEARCH AND DEVELOPMENT

CENfER FOR

COGNITIVE LEARNING i

o

et

ED 039 617

AUTHORTITLE

INSTITUTION

SPONS AGENCY

REPORT NOBUREAU NOPUB DATECONTRACTNOTE

EDRS PRICEDESCRIPTORS

DOCUMENT RESUME

24 CG 005 640

Davidson, Robert E. Dollinger, Laurel E.The Effects of Deep Structure Variations inSentences: Free Recall and Paired Associate Learning.Wisconsin Univ., Madison. Research and DevelopmentCenter for Cognitive Learning.Office of Education (DHEW) , Washington, D.C. Bureauof Research.TR-110BR 5-0216Dec 69OEC-5-10-15417p.

EDRS Price MF-$0.25 HC-$0.95College Students, *Deep Structure, Elementary SchoolStudents, Grade 2, *Learning, *Paired AssociateLearning, *Recall (Psychological) , *SentenceStructure

ABSTRACTThe deep structure description of a sentence marks

the actual grammatical relationships that exist among the words. Itis in the deep structure that "meaning" is rendered. Two sentencesthat are marked differently in deep structure might give rise to thesame description in surface structure. Three experiments examined the"psychological reality" of deep structure variations for sentences ofthe language. Analyses of a variety of dependent variables from aariety of experimental paradigms with two different populations

suggested that deep structure variations produce differential effectsin learning and recall. Two experiments with university studentsu sing free recall and cued recall paradigms indicated thattransitional error probabilities could detect different recallpatterns reflecting the deep structure variations. A third experimentw ith second grade children revealed that relatively greaterfacilitation of learning characterized sentences where an integralsubject-object relationship could be specified between the paireditems in some deep structure form. (Author)

Nabrm4 Technical Report No. 110%CO

re\ THE EFFECTS OF DEEP STRUCTURE VARIATIONS IN

SENTENCES: FREE RECALL AND PAIRED-ASSOCIATE LEARNINGLi.J

By Robert E. Davidson and Laurel E. Dollinger

Report from the Project on Situational Variablesand Efficiency of Concept Learning

Herbert J. Klausmeier and Robert E. DavidsonPrincipal Investigators

Wisconsin Research and DevelopmentCenter for Cognitive LearningThe University of Wisconsin

Madison, Wisconsin

December 1969

Published by the Wisconsin Research and Development Center for Cognitive Learning, supportedin part as a research and development center by funds from the United States Office of Education,Department of Health, Education, and Welfare. The opinions expressed herein do not necessarilyreflect the position or policy of the Office of Education and no official endorsement by the Officeof Education should be inferred.

Center No. C-03 / Contract OE 5-10-154

NATIONAL EVALUATION COMMITTEE

Samuel BrownellProfessor of Urban EducationGraduate SchoolYale University

Launor F. CarterSenior Vice President on

Technology and DevelopmentSystem Development Corporation

Francis S. ChaseProfessorDepartment of EducationUniversity of Chicago

../r

Henry ChaunceyPresidentEducational Testing Service

Martin DeutschDirector, Institute for

Developmental StudiesNew York Medical College

Jack EdlingDirector, Teaching Research

DivisionOregon State System of Higher

Education

Elizabeth KoontzWage and Labor Standards

Administration, U.S.Department of Labor,Washington

Roderick McPheePresidentPunahou School, Honolulu

G. Wesley SowardsDirector, Elementary EducationFlorida State University

Patrick Suppe,ProfessorDepartment of MathematicsStanford University

*Benton J. UnderwoodProfessorDepartment of PsychologyNorthwestern University

RESEARCH AND DEVELOPMENT CENTER POLICY REVIEW BOARD

Leonard BerkowitzChairmanDepartment of Psychology

Russell J. HosierProfessor, Curriculum

and Instruction

Stephen C. KleeneDean, College of

Letters and Science

B. Robert TabachnickChairman, Department

of Curriculum andInstruction

Archie A. Buchmiller Clauston Jenkins Donald J. McCarty Henry C. WeinlickDeputy State Superintendent Assistant Director Dean Executive SecretaryDepartment of Public Instructit Coordinating Committee for School of Education Wisconsin Education Association

` Higher Education

Robert E. Grinder Herbert J. Klausmeier Ira Sharkansky M. Crawford YoungChairman Director, R & D Center Associate Professor of Political Associate DeanDepartment of Educationc,i Professor of Educational Science The Graduate School

Psychology Psychology

EXECUTIVE COMMITTEE

Edgar F. BolattaBrittingham Profer,or of

Sociology

Anne E. Buci:iananProject SpecialistR & D Center

Robin S. ChapmanResearch AssociateR & D Center

Robert E. DavidsonAssistant Professor,

Educational Psychology

Frank H. FarleyAssociate Professor,

Educational Psychology

Russell J. HosierProfessor of Curriculum and

Instruction ancl of Business

*Herbert J. KlausmeierDirector, R & D CenterProfessor of Educational

Psychology

Wayne OttoProfessor of Curriculum and

Instruction (Reading)

Robert G. PetzoldAssociate Dean of the School

of EducationProfessor of Curriculum and

Instruction and of Music

FACULTY Oil PRINCIPAL INVESTIGATORS

Verivein L. AllenProfessor of Psychology

Ted CzajkowskiArcistant Professor of Curriculum

and Instruction

Robert E. DavidsonAssistant Professor of

Educational Psychology

Gary A. DavisAssociate Professor of

Educational Psychology

M. Vere DeVaultProfessor of Curriculum and

Instruction (Mathematics)

Frank H. FarleyAssociate Professor of Educational

Psychology

Lester I S. GolubLecturer In Curriculum and

Instruction and in English

John G. HarveyAssociate Professor of

Mathematics and of Curriculumand Instruction

Herbert J. KlausmeierDirector, R & D Center

Professor of EducationalPsychology

Donald LangeAssistant Professor of Curriculum

and Instruction

James MoserAssistant Professor of Mathematics

Education; Visiting Scholar

Wayne OttoProfessor of Curriculum and

Instruction (Reading)

Milton 0. PellaProfessor of Curriculum and

Instruction (Science)

Thomas A. RombergAssociate Director, R & D Center

Professor of Mathematics andCurriculum and Instruction

B. Robert TabachnickChairman, Department

of Curriculum andInstruction

Richard L. VenezkyAssistant Professor of English

and of Computer Sciences

Alan VoelkerAssistant Professor of Curriculum

and Instruction

Larry WilderAssistant Professor of Curriculum

and Instruction

Peter WolffAssistant Professor of Educational

of Psychology

MANAGEMENT COUNCIL

Herbert J. KlausmeierDirector, R & D Center

V.A.C. Henmon Professor ofEducational Psychology

Mary R. QuillingDirector

Technical Development Program

Thomas A. RombergAssociate Director

II

James WalterDirector

Dissemination Program

Dan G. WoolpertDirector

Operations and Business

* COMMITTEE CHAIRMAN

STATEMENT OF FOCUS

The Wisconsin Research and Development Center for Cognitive Learningfocuses on contributing to a better understanding of cognitive learning bychildren and youth and to the improveinent of related educational practices.The strategy for research and development is comprehensive. It includesbasic research to generate new knowledge about the conditions and processesof learning and about the processes of instruction, and the subsequent develop-ment of research-based instructional materials, many of which are designed foruse by teachers and others for use by students. These materials are tested andrefined in school settings. Throughout these operations behavioral scientists,curriculum experts, academic scholars, and school people interact, insuringthat the results of Center activities are based soundly on knowledge of subjectmatter and cognitive learning and that they are applied to the improvement ofeducational practice.

This Technical Report is from the Situational Variables and Efficiency ofConcept Learning Project in Program 1. General objectives of the Program areto generate new knowledge about concept learning and cognitive skills, tosynthesize existing knowledge, and to develop educational materials suggestedby the prior activities. Contributing to these Program objectives, the ConceptLearning Project has the following five objectives: to identify the conditionsthat facilitate concept learning in the school setting and to describe theirmanagement, to develop and validate a schema for evaluating the student'slevel of concept understanding, to develop and validate a model of cognitiveprocesses in concept learning, to generate knowledge concerning the semanticcomponents of concept learning, and to identify conditions associated withmotivation for school learning and to describe their management.

iii

iv

ACKNOWLEDGMENT

The writers wish to thank the principal, teachers, and pupils ofBlack Earth Elementary School who aided in Experiment III.

U.S DEPARTMENT OF HEALTH, EDUCATION& WELFARE

OFFICE OF EDUCATIONTHIS DOCUMENT HAS BEEN REPRODUCEDEXACTLY AS RECEIVED FROM THE PERSON ORORGANIZATION ORIGINATING IT POINTS OFVIEW OR OPINIONS STATED DO NOT NECESSARILY REPRESENT OFFICIA I., OFFICE OF EDI.)CATION POSITION OR 90LICY,

CONTENTS

List of Tables

Abstract

Page

vii

ix

I Introduction 1Deep Structure Variations: Free-Recall Learning 1Deep Structure Variations: Paired-Associate Learning 1

II The Sentence-Learning ExperimentsExperiment I

MethodSubjectsMaterialsProcedureScoring

ResultsExperiment II

MethodSubjectsMaterialsProcedureDesignScoring

ResultsDiscussion

III The Paired-Associate ExperimentExperiment III

MethodSubjectsMaterialsApparatusDesignProcedure

ResultsDiscussion

References

3

3

3

3

3

3

3

4

44

44

4

4

5

5

5

7

7

7

7

7

7

7

7

89

11

LIST OF TABLES

Table

1 Transitional Error Probabilities for Each Kind of Sentenceand Transition

2 Means and Standard Dgviations of Correct Responses forthe Three Independent Groups

Page

8

3 Multivariate Analysis of Variance Summaries and AssociatedUnivariate Statistics 9

vii

ABSTRACT

The deep structure description of a sentence marks the actual gram-matical relationships that exist among the words. It is in the deep struc-ture that "meaning" is rendered. Two sentences that are marked differentlyin deep structure might give rise to the same description in surface struc-ture.

Three experiments examined the "psychological reality" of deep struc-ture variations for sentences of the language. Analyses of a variety of de-pendent variables from a variety of experimental paradigms with two differ-ent populations suggested that deep structure variations produce differentialeffects in learning and recall.

Two experiments with University students using free-recall and cued-recall paradigms indicated that transitional error probabilities, i.e., errorsin word-to-word transitions, could detect different recall patterns reflectingthe deep structure variations.

A third experiment with Second Grade children serving as Ss in a paired -associate paradigm revealed that relatively greater facilitation of learningcharacterized sentences where an integral subject-object relationship couldbe specified between the paired-items in some deep structure form.

INTRODUCT'ON

DEEP STRUCTURE VARIATIONS:FREE-RECALL LEARNING

Recent linguistic theory postulates a "deepstructure" to any sentence in the language(Chomsky, 1965). The deep structure descrip-tion marks the actual grammatical relationshipsthat exist in the sentence. Two sentenceswhich are marked differently in their underlyingstructure might give rise to sentences whichhave the same constituent descriptions in "sur-face structures," Thus, while two sentencesmay be the same in surface structure,

(i) Delicate lace was producedby tailors.

(ii) Delicate lace wasiroducedby hand.

the grammatical relations that exist in deepstructure are markedly different. In (i) tailorsactually serves as the logical subject, and thepassive sentence can be made active: Tailorsproduced delicate lace. In (ii) hand is actuallya part of a manner adverbial modifier and thesubject of the sentence is not stated.

Psycholinguistic research has attempted todemonstrate that if Ss are responsive to deepstructure when they process sentences, thenthey should show behavioral differences in somekind of appropriate task. For example, Fodorand Garrett (1967) found that relative pronouncues in surface structure made sentences easierto understand because such cues signaled im-portant relationships in deep structure. Also,the research of Blumenthal (1967) and Blumenthaland Boakes (1967) indicates that, under certainconditions, the number of sentences recalled ismore a function of deep structure than it is ofsurface structure descriptions. The work ofSavin and Perchonock (1965) suggests that sen-tences may be coded in memory in a way thatparallels their deep structure descriptions.

On the other hand, the surface descriptionof sentences seems to be the more important

determiner in perceptual tasks such as speechperception (Fodor & Bever, 1965) or reading(Mehler, Bever, & Carey, 1967). However,the basically nonperceptual experiments ofsentence learning over trials, in which the num-ber of transitional errors ,between adjacent words

din the sentence was the dependent measure,have consistently shown that error patterns fol-low the constituent structure at the surface(Johnson, 1965). Certainly a sentence learningtask requires the processing of sentences inmemory and the question that arises is, willtransitional error probability (TEP) patterns beresponsive to the deep structure differences insentences that have the same surface structure?Experiments I and II address themselves to thisquestion in a free-recall task and a cued-recalltask respectively.

DEEP STRUCTURE VARIATIONS:PAIRED-ASSOCIATE LEARNING

The syntactic facilitation of paired-associate(1)1) learning is well documented (Jensen & Rohwer,1963; Davidson, 1964; Rohwer, 1966). Recentattempts to specify the conditions for syntacticfacilitation have examined, inter alia, semanticconstraint (Rohwer (Sc Lynch, 1966), intralistsimilarity (Rohwer & Lynch, 1967), and impliedaction differences in transitive verbs (Rohwer& Levin, in press). Recently, in a study whichexamined sentence contexts in PA learning(Rohwer, Shuell, & Levin, 1967), it was ob-served that the entire sentence, its structuralconfiguration, may be the functional stimulusfor the facilitating effect. This observationcan be related to current linguistic theory.

Chomsky (1965, 1968) has drawn a distinc-tion between the "surface" and "deep" structuredescription of a sentence. According to thisview, the structure of a sentence in its mani-fest or surface form might not account for allthe complexity perceived by the hearer. That

1

is, while the surface structure of two sentencesmay be alike, the deep structure can be verydifferent. It is the deep structure that signalsthe actual grammatical relationships that existin the sentence. To understand or interpret asentence is to recover the grammatical rela-tionships that exist in deep structure.

Recent experiments suggest that surfacestructure descriptions are important to percep-tual tasks such as speech perception (Fodor& Bever, 1965) and reading (Mehler, Bever,& Carey, 1967), while deep structure seemsto be the important determiner in memory (Savin& Perchonock, 1965), and sentence interpreta-tion (Fodor & Garrett, 1967).

The following examples from Blumenthal (1967)m',ght make the linguistic distinction clear:

The child was warmed bythe stove.

(ii) The child was warm bythe stove.

By the mere omission of the suffix "--ed" fromthe verb, the word "stove" changes its functionfrom logical subject, to part of an adverbialmodifier. Thus, the two sentences have thesame constituent phrase structure at the sur-face but they differ markedly in deep structure.

Another way of putting this is to say thatthe transformational history of the two sen-tence strings are different. In (a) , neither the

2

underlying structure nor the transformationsleading t the manifest passive form precludesome active representation, i.e., The stovewarmed the child. This active paraphrasehighlights an integral relationship betweenthe (logical] subject and object of the sen-tence. In (12), a similar relationship cannotbe specified, and the manifest passive formhas no intimate link to some active paraphrase.

Experiment III examines the effects of deepstructure variations in PA learning. Paired-associate items were embedded in sentencesalike in their surface configuration but differ-ent in their deep structure. With the to-be-learned associates functioning as subject,object, or adverbial, relatively greater facili-tation might be expected for those sentencesin which the first and second nouns (i.e., thepaired items) bear an integral subject-objectrelationship. A similar relationship does notexist between the nouns of the adverbial kind.

For each sentence kind both the first andsecond nouns were used as stimulus in thePA task. In view of the results of Blumenthal(1967), there should be little difference infacilitation of PA learning when the first nounacts as the stimulus because it fulfills thesame function in both kinds of sentences.However, when the second noun is used asstimulus, it should favor those sentenceswhere it functions as subject .

II

THE SENTENCE LEARNING EXPERIMENTS

EXPERIMENT I

METHOD

Subjects

Forty students at the University of Wiscon-sin served as Ss fulfilling a course requirement.Twenty Ss were assigned at random to the treat-ments.

Materials

Ten sentences of the manner-adverbial kind(M) and ten of the full-passive kind V) wereconstructed. Examples:

MA Important messages weredispatched by wire.

FP Important messages weredispatched by governments.

Notice that the constituent phrase structure atthe surface is the same for both types of sen-tence. The structure was always of the form:

adjective - noun -was /were -verb -by -noun.

The sentences were recorded on tape. Thereading time for each was approximately fiveseconds, with a slight pause between sentences.

Procedure

Small groups of Ss (4 to 8 in each group)listened to either MA or FP sentences over fivelearning-test trials. For each trial, the S hearda different random arrangement of the 10 sen-tences. After the last sentence was read forthe learning portion of the trial, the Ss weretested for their recall. They were given 3 min-utes to write down the words they could remem-ber. Booklets with five pages (trials) were pro-

vided, and each page contained 10 separatelines with six dashes to the line. The Ss wereencouraged to guess words and their locationsin sentences.

Scoring

Three different kinds of scores were com-puted for each S. The first of these was thenumber of words correct at the constituentword level, i.e., adjective. .noun...mainverb .noun (only the main verb was scoredbecause of the constant (...was/were...by...)pattern). The constituent word scores weresummed over the 10 sentences for each trial.Thus, four scores for the number of adjectives,main verbs, etc., were computed for each sub-ject at each trial.

A second kind of score, the TEP, was com-puted for each S. The TEP for each left-to-right, word-to-word (again ignoring the (...was/were...by)) transition within each sen-tence was determined by dividing the frequencythat a word following a transition was incorrectby the frequency that the word before the transi-tion was correct. That is, for the first transi-tion (adjective...noun) in the example sen-tences, the TEP would be the frequency thatmessages was incorrect, given that importantwas correct, divided by the frequency thatimportant was correct. There are three transi-tions in the test sentences. They have thefollowing pattern: adjective to noun (TEP1),noun to main verb (TEP2), main verb to noun(TEP3). This last noun is, in fact, the logicalsubject in FP but is part of the adverbial modi-fier in MA. A TEP for each transition was ob-tained by summing over the sentences and overtrials for each S.

A third kind of score computed for each Swas the deep-structure transitional error proba-bility (DTEP). It was computed by the TEP pro-cedure, but here the transitions were ordereddifferently. A DTEP1 score was for the transition

3

between the logical-subject noun (or the manner-adverbial noun in MA sentences) and the mainverb. A DTEP2 transition existed between themain verb and adjective, and a DTEP3 transi-tion was between adjective and noun (e.g.,Governments (DTEP1) dipatched (DTEP2) im-portant (DTEP3) messages). Note that the TEP1DTEP3 is the same adjective-to-noun transition.

The justification for the DTEP score is thatit should be sensitive to the passivization trans-formation that intimately links FP passive sen-tences and their active representation to somedeeper structure. Slmilar links do not exist bl-tween MA passives and some active representa-tion, or between these MA sentences and somedeep structure which the DTEP score would re-flect.

RESULTS

A repeated measures analysis of variance forthe words-correct measure showed differencesin recall for the various constituent words (F24.24, df = 3/114, 2

FP) and Cues (First noun, Second noun) . De-pending upon his group assignment, the S re--ceived a booklet which provided cue words foreach sentence at every trial. Either the firstnoun of the sentence or the second noun of thesentence was printed. The Ss were encouragedto use the cue words to aid them in recall. Thisdesign is, in gross outline, that of Blumenthal(1967). Predictions were similar. That is, theSs serving under the FP condition and who re-ceived the second noun (logical subject) as acu, were expected to be facilitated in theirrecall.

Scoring

With some necessary changes, the threekinds of scores computed in Experiment I werecalculated here. Because different cue wordswere provided, word constituent scores werenot meaningful. Instead, the number-of-words-correct score was a summation of the three re-maining words in each of the 10 sentences foreach trial; these were cast in a 2 x 2 x 5 re-peated measures analysis of variance. TheTEP score and DTEP score were determinedas before. It can be noted that either largeror non-existent error probabilities for somecells would occur because cue words mightconstitute one side or the other of a transi-tion.

RESULTS

Analysis of the words-correct measurerevealed a significant cue effect = 4.59,

df = 1/56, a < .05) with the first noun pro-yicling greater facilitation of recall. Neitherthe Sentence effect (F = 2.28; df = 1/56) northe predicted interaction (F < 1.00) was foundLo be significant.

In some conditions, the cues that were pro-vided made up the right hand side of the transi-tion; therefore, certain transitional errors didnot exist. But despite this general limitation,the critical transitions could be tested. Againthe tests of interest contrast the MA and FPgroups. On Mann-Whitney tests, only theDTEP1 transition scores were significantlydifferent (Z = 2.57; a < .005) with MA =.3474 and FP = .2869.

DISCUSSION

Transitional error patterns have consistentlyshown themselves to be sensitive to sentencesurface structure differences. The results ofthese experiments suggest that the metric maybe very sensitive to deep structure differencesas well. Of course the addition or elaborationof any good operational test to the study oflanguage is especially needed at this time.The psycholinguistic researcher has reachedthe point where he no longer is content merelyto demonstrate the verities of a linguistic com-petence model; instead, he is ready to lookat language behavior differently than he hasin the past. He is ready to postulate psycho-logical competencies with respect to languagethat are likely to be, at the same time, atvariance with any fully elaborated system oflinguistic operations and with any traditionalpsychological theory as well.

5

III

THE PAIRED-ASSOCIATE EXPERIMENT

EXPERIMENT III

METHOD

Subjects

Fifty-four Second-Grade children of mixedability levels were recruited from one schoolin a semi-rural community.

Materials

Forty pictures were made into transparenciesand set into 35mm mounts for slides. The pic-tures were simple line drawings of familiar ob-jects found in primer workbooks.

The pictures were paired in a way to avoidobvious associations, and sentences of an in-tegral passive (L12) or manner adverbial (MI)kind were constructed for each pair. The sur-face structure of the sentences was always ofthe form:

adjective-noun-was/were-verb-by-noun

Between the IP and MA versions, the verbs andplural suffixes were changed to signal the ap-propriate interpretation of the sentence.Examples:

MA Soft bread was delivered by pony.

IP Soft bread was tasted by ponies.Two response booklets were provided for each

S, one for each trial. These consisted of a facesheet and 20 identical response pages. Each re-sponse page depicted all 20 response items.Ten of these responses had appeared as the firstnoun in the sentence, and ten had appeared asthe second noun. Three practice pages showinggeometrical figures were included in the book-lets used for the first trial.

Apparatus

The apparatus allowed a fully automated ex-periment. Pairs of slides were shown on ascreen in a semi-darkened room from two KodakCarousel 800 projectors operated from a singleremote control unit. Sentences were recordedon one track of stereo recording tape. On thesecond track of the tape a timed signal wasrecorded by means of a Kodak sound synchro-nizer. Thus, the slide projectors were pro-grammed to change slides automatically and inconjunction with the playback of the appropriateverbalizations. The timing signal was notaudible to the Ss. Tapes were played on aSony TC-200 stereo tape recorder.

Design

A three-group multivariate design was em-ployed using a three-level independent factorof mediation (IP, MA, Control), and two bi-level repeated factors of trials and nouns.Thus, each S was assigned to one of threegroups and four observations were obtained,two at each trial for the first and second noun.

Procedure

A recognition PA method with small groupsof children was employed (Davidson, 1964).The general nature of the task was explainedand the Ss were familiarized with the locationof the pictures on the response pages by askingthem to point to each picture as it was namedby E This familiarization was included to avoidthe loss of correct responses because of diffi-culty in locating items.

Specific instructions were then stated. Sub-jects in the mediated conditions were told to106k at the pictures, listen to the sentences,

6/7

and remember which two pictures belonged to-gether. It was emphasized that the sentenceswould help them remember the pairs. Theywere also informed that the response trialswould proceed at a rapid pace and that theymust work quickly, turn to the appropriatelynumbered page as instructed by the tape-recorded voice, and follow the page numberscarefully. Guessing was encouraged.

The instructions for the control conditionwere identical except that in place of anyreference to the sentences Ss were told thatthe pictures would be named as they were shown.

Three practice items were then performedusing geometric figures on actual slides. Thepractice items illustrated the fact that the teststimuli would be in a different order than thepairs and that either the first or second itemcould be the stimulus, although E did not statethese facts explicitly.

The actual experimental trials were then be-gun. For the study portion of each trial all 20pairs of slides were shown at 4-second expo-sures, allowing an additional half-second forthe projectors to change slides. In the experi-mental groups the pairs were accompanied bythe spoken sentences. All sentences wereeither of the IP or MA construction. For thecontrol group a more traditional paired-associatesprocedure was followed in that the pictures weremerely named as they were shown.

Timing for the test or recognition portion was8 seconds. As each of the 20 stimuli was shownand named, Ss were to circle the other (missing)member of the pair on their response sheets.

The procedure for the second trial was, identi-cal to the first except for the use of briefer in-structions and omission of the practice items.

There was an approximate 2-minute i4ntervalbetween trials as new booklets were di(Aributed,slide-holders were changed, and portions of theinstructions were repeated.

The entire experimental session ran lessthan 30 minutes per group.

RESULTS

The dependent measure was the number ofcorrect responses. A response was scored ascorrect if the appropriate item on the appropri-ate page and only that item Ares circled by S.

The results presented here are based on atotal of 18 Ss in each experimental group and16 Ss in the control group. Two of the original18 Ss were lost from the control group becauseof their inability to keep up with the pace ofthe responses and resultant cessation of per-formance.

Means and standard deviations for noun andtrial totals are presented in Table 2.

Initially a Dunnet statistic for total scoreswas performed which indicated that both ex-perimental conditions differed significantlyfrom the control group (Integral Passive vs.Control, t = 6.5, p < .001; Manner Adverbialvs. Control, t = 4.7, p < .001).

The magnitude of the effect was analyzed bymeans of the w2 / statistic, which indicated thatapproximately 62% of the variance could beaccounted for by the three main conditions.(This is comparable to the results of other in-vestigators in similar mediation studies, es-pecially Davidson (1964), .u.12/ = .73).

The Dunnet test statistic revealed that, aswas expected, the mediiAing materials used in

Table 2

Means and Standard Deviations of Correct Responsesfor the Three Independent Groups

MediationType

Totals

Trial 1 Trial 2 Noun 1 Noun 2

Integral Passive M 11.06 15.76 12.06 14.78(N = 18) SD 3.57 3.51 4.04 3.10

Manner Adverbial M 7.94 13.50 9.72 11.72(N = 18) SD 3.26 4.05 3.68 3.39

Control M 2.63 5.38 4.00 4.00(N = 16) SD 1.59 3.50 2.19 2.22

8

the experimental groups produced a very strongand distinct facilitation of learning over the con-trol group, which used the traditional, nonmedi-ated procedure of PA learning. Therefore, thecontrol group was omitted from further analysisin order to permit a more precise examination ofthe effects of major interest.

A multivariate analysis of variance was per-formed for the two mediation groups. The over-all joint multivariate test of nouns within trials(all variates) was of sufficient magnitude toallow the associated set of contrasts for trials,nouns, and total score to be formed and tested(F = 2.07, df = 4/31, p < .11). The totalscore contrast was a univariate test indicatingthat the two mediation groups differ significantly(F = 6.03, df = 1/34, p < .02). The subse-quent multivariate tests and their associatedunivariate statistics are presented in Table 3.The joint tests for trials and nouns and an in-spection of the related means show that the IPcondition had the greater number of correct re-sponses in all cases. The associated univari-ate statistics were used to examine the locusof these effects. The several tests indicatedthat the two groups differed significantly forthe Trial 1 and Noun 2 response scores.

Table 3

Multivariate Analysis of Variance Summariesand Associated Univariate Statistics

Joint Multivariate Analysis of Variance of Trials(2 Variates)

F

3.61

DFHYP

2.00

DFERR

33.00 < .04

Univariate Tests of Trials:

Total-Trial 1Total-Trial 2

F(df 1/34)

7.443.26

MS

87.11 < .0146.69 < .08

Joint Multivariate Analysis of Variance of Nouns(2 Variates)

F DFHYP DFERR

3.91 2.00 33. PO

Univariate Tests of Nouns:F (df 1/34) MS

Total-Noun 1Total-Noun 2

3.29 49.007.96 84.03

< .03

P

< .08< .008

GPO 816-250-3

tr. Vastleica.illarditarnu

DISCUSSION

The finding that syntactic mediation is su-perior to a control condition was hardly unex-pected. The phenomenon is so well documentedthat any discussion is unnecessary. In thepresent experiment, once the control group hadprovided an estimate of some minimum base-line performance, it had served its purpose andwas eliminated from further consideration. Fu-ture experiments that aim at assessing the effectsof various kinds of syntactic mediation mightdeem it unnecessary, except in special cases,to include a traditional PA control. Such agroup seems to add little to the assessment ofsyntactic mediators.

The major finding of this study does lendsupport to the assertion that deep structurevariations may be expected to affect differenti-ally PA learning. A full explanation as to thenature of the effect must await further study;however, the overall superiority of the IP con-dition suggests that the integral subject-objectrelationship between the paired items in someunderlying deep structure form has a general,facilitating effect. This result is reminiscentof the hypothesis of Miller (1962) which sug-gestei that it is the "kernel" (read, underlyingstructure of the sentence) plus some transforma-tional footnote that is stored in memory. Lin-guistic theory has changed somewhat since theMiller writing, but the essence of the hypothesisseems cogent with respect to these experimentalresults.

The more analytic explanation was not sup-ported, i.e., the explanation that would attrib-ute the facilitation specifically to the mediatingsentences in which the stimulus functioned assubject. Indeed, the univariate results suggestthe opposite. The use of the first noun as stim-ulus produced better recognition than did theuse of the second noun. An explanation for thediscrepancy is not readily available, especiallysince the findings here seem counter-intuitive.As the study of Blumenthal (1967) suggests, itwould seem logical that when nouns differ infunction, their effectiveness as stimuli shouldalso differ.

Of additional interest in this study was therelatively stronger effect between kinds of sen-tences found at Trial 1. This is consistent withthe findings of Rohwer (1966) that syntacticfacilitation seems to have its greatest effect onthe initial trial. However, it must be rememberedthat syntactic mediators are extremely powerfulfacilitators of PA learning, even with 20 pairs tobe learned, and many children are at asymptoteat the second trial.

9

REFERENCES

Blumenthal, A. Promoted recall of sentences.Journal of Verbal Learning and Verbal Be-havior, 1967, 6, 203-206.

Blumenthal, A., and Boakes, R. Promptedrecall of sentences, a further study.Journal of Verbal Learning and Verbal Be-havior, 1967, 6, 674-767.

Chomsky, N. Aspects of the theory of syntax,Cambridge: M.I.T. Press, 1965.

Chomsky, N. Language and the mind. NewYork: Harcourt, Brace & World Inc., 1968.

Davidson, R. E. Mediation and ability inpaired-associate learning. Journal ofEducational Psychology, 1964, 55, 352-356.

Fodor, J. A., and Bever, T. G. The psycho-logical reality of linguistic segments.Journal of Verbal Learning and Verbal Be-havior, 1965, 4, 414-420.

Fodor, J. A., and Garrett, M. Some syntacticdeterminants of sentential complexity. Per-ception and Psychophysics, 1967, 2, 289-295.

Jensen, A. R., and Rohwer, W. D., Jr. Verbalmediation 11 paired-associate and seriallearning. Journal of Verbal Learning andVerbal Behavior, 1963, 1, 346-352.

Johnson, N. F. The psychological reality ofphrase-structure rules. Journal of VerbalLearning and Verbal Behavior, 1965, 4,469-475.

GPO 816-250-2

Mehler, J., Bever, T., & Carey, P. What welook at when we read. Perception & Psy-chophysics, 1967, 2, 213-218.

Miller, G. A. Some psychological studies ofgrammar. American Psychologist, 1962, 17748-762.

Rohwer, W. D., Jr. Constraint, syntax andmeaning in paired-associate learning.Journal of Verbal Learning and Verbal Be-havior, 1966, 5, 541-547.

Rohwer, W. D., Jr. & Lynch, S. Semanticconstraint in paired-associate learning.Journal of Educational Psychology, 1966,57, 271-278.

Rohwer, W. ID., Jr. & Lynch, S. Form classand intralist similarity in paired-associatelearning. journal of Verbal Learning andVerbal Behavior, 1967, 6, 551-554.

Rohwer, W. D., Jr., Shuell, T. J., & Levin,J. R. Context effects in the initial storageand retrieval of noun pairs. Journal of VerbalLearning and Verbal Behavior, 1967, 6, 796-801.

Rohwer, W. D., Jr. & Levin, J. R. Action,meaning and stimulus selection in paired-associate learning. Journal of Verbal Learn-ing and Verbal Behavior (in press).

Savin, H. B., & Perchonock, E. Grammaticalstructure and the immediate recall of Englishsentences. Journal of Verbal Learning andVerbal Behavior, 1965, 4, 348-353.