R EPOR T R ESUMESED 015 111 24

THE EFFECT OF MEMORY SPAN ON CUE PATTERNS IN WORDRECOGNITION.

BY- WEISSGLASS, ROBERTA

REPORT NUMBER TR-16REPORT NUMBER OR- 5- 0216-TR -16CONTRACT OEC-5-10-154EDRS PRICE MF.-$0.25 HC-$1.04 24P.

RE 000 999

PUB DATE DEC 66

DESCRIPTORS- *WORD RECOGNITION, *READING RESEARCH,*KINDERGARTEN CHILDREN, LEARNING PROCESSES, VISUAL STIMULI,LINGUISTIC PATTERNS, *ATTENTION SPAN, UNIVERSITY OFWISCONSIN, MADISON,

KINDERGARTEN CHILDREN WERE USED AS SUBJECTS IN ANATTEMPT-TO DISCOVER WHETHER CHILDREN WITH SHORT AND LONGMEMORY SPANS USE DIFFERENT CUES TO RECOGNIZE WORDS ANDWHETHER MEMORY SPAN TESTS ARE EFFECTIVE PREDICTORS OFDIFFICULTY IN LEARNING TO READ. IT WAS HOPED THAT CHILDRENWOULD EXHIBIT A CONSISTENT PATTERN OF CHOICES OVER TRIALS,WHICH MIGHT HAVE IMPLICATIONS FOR TEACHING WORD RECOGNITIONAND OTHER READING SKILLS. THE DIGIT SPAN SUBTEST OF THEWECHSLER INTELLIGENCE SCALE FOR CHILDREN, AN AUDITORY TEST,MEASURED MEMORY SPAN AND WAS USED AS THE SCREENING DEVICE.FIVE-LETTER.NONSMNSE WORDS COMPRISED THE WORD RECOGNITIONTASK. CHILDREN WERE REQUIRED TO SELECT FROM A GROUP OFNONSENSE WORDS THE ONE SIMILAR TO THE WORD THAT HAD JUST BEENSHOWN TO THEM. EACH WORD IN THE RESPONSE GROUP CONTAINED ONECUE WHICH APPEARED IN THE SAME POSITION AS IN THE STIMULUSWORD WITH THE OTHER CUES HELD CONSTANT. FIVE CUES WEREEXAMINED -- POSITIONS 1, 2, 3, 4, AND 5, AND EACH SUBJECT HADAN EQUAL-OfFORTUNITY TO RESPOND TO EACH CUE. ALL GROUPSSHOWED A PREFERENCE FOR CUE 1, AND THERE WAS A TENDENCY TORESPOND TO CUE 3 AND CUE 5. AN AVOIDANCE EFFECT SEEMED TO BEOPERATING WITH CUE 2 AND CUE 4. THE RESULTS ON FRIEDMANSANALYSIS OF VAUIANCE OF RANKS INDICATED THAT RANDOM SELECTIONMAY HAVE BEEN OPERATING FOR BOYS AND GIRLS IN THE LOW MEMORYSPAN GROUP. FOR THE MIDDLE AND HIGH MEMORY SPAN BOYS, APREFERENCE PATTERN WAS NOT LIKELY TO ARISE BY CHANCE TABLES.24 REFERENCES, AND AN EXAMPLE OF THE FORM FOR RECORDiNGRESPONSE CHOICES ARE INCLUM. (AUTHOR)

"71,1111PIRIPIMPPRIIR,PIRI" : ,'

CI{

, ., , t ,,

I a ' ' 1 ,

`MILs

".

f,

s a 4

0.&-!:.!

`,..21''',75

,

"4II' "

.;

... ,..., ..

, ,:.;i' '.

. ; :".4''' A ., e ' f ., ..

16 ' 1. , .,:' ''',' '`'4 2. ';'.4 ''.' *,.` 4.':411' 4; 1 14. '` ',

'.4. " 4' ,,.' 4, 'a

,d a

4u

s a, i;,:i..,.*,...,

'"

.%1 r

, ' , ;pd 4

'4

f

:4 ;4.t ,f 4.

; 14- I :T. t. "

S.' s,

' "ify

...a OCOC CO

..... IA......a CO3; O. tar

1.6..

...-ar ...,

....at 04... ::::. ca. 4.-= ,67.:C% fli 1.6..Ira OK ss.. a,Ira= ....9 CO Ili1. = ,...,

=sr. 0S ..... s=o MC e,,,.., 4.a) 4.7.... a2C :IC :Ile erg= ...... co ...I

b.i./II 1. ...,Og CO .4 WI

VI 0.4._...g is5IC 1,17e .:=s.

s=o

V% = 0= C.g.., 0=.

4,=. :_

V.5 0.....1

Om O. tol CI.

41,411 I

Technical Report No. 16

THE zer uF MEMORY SPAN ON

CUE PATTERNS IN WORD RECOGNITION

Roberta R. Weissglass

Based on a master's paperunder the direction of

Wayne OttoAssociate Professor of Curriculum and Instruction

Research and Development Centerfor Learning and Re - educationThe University of Wisconsin

Madison, Wisconsin

December 1966

The research reported herein was performed pursuant to a contract with the United StatesOffice of Education, Department of Health, Education, and Welfare, under the provisions of theCooperative Research Program.

Center No. C-03 / Contract 0E5-10-154

OTHER REPORTS OF THE R & D CENTER FOR LEARNING AND RE-EDUCATION

TECHNICAL REPORTS

No. 1 Klausmeier, H. J. , Davis, J. K. , Ramsay, J. G. , Fredrick, W. C., &Davies, Mary H. Concept learning and problem solving: A bibliography,1950-1964. October 1965.

No. 2 Goodwin, W. L. The effects on achievement test results of varying condi-tions of experimentalatmosphere, notice of test, test administration, andtest scoring. November 1965.No. 3 Fredrick, W. C. The effects of instructions, concept complexity, methodof presentation, and order of concepts upon a concept attainment task.

November 1965.No. 4 Ramsay, J. G. The attainment of concepts from figural and verbal in-

stances, by individuals and pairs. January 1966.No. 5 Van Engen, H., &Steffe, L. P. First grade children's concept of additionof natural numbers. February 1966.No. 6 Lynch, D. 0. Concept identification as a function of instructions, labels,

sequence, concept type, and test item type. June 1966.No. 7 Biaggio, Angela M. B. Relative predictability of freshman grade-pointaverages from SAT scores in Negro and white Southern colleges. Sep-tember 1966.

No. 8 Kalish, Patricia W. Concept attainment as a function of monetary incen-tives, competition, and instructions. September 1966.No. 9 Baldwin, Thelma L. , & J_thnson, T. J. Teacher behaviors and effective"-ness of reinforcement. September 1966.

No. 10 Fang, M. C. S. Effect of incentive and complexity on performance of stu-dents from two social class backgrounds on a concept identification task.September 1966.

No. 11 Lemke, E. A., Klausmeier, H. J. & Harris, C. W. The relationship ofselected cognitive abilities to concept attainment and information pro-cessing. October 1966.No. 12 Burmeister, Lou E. An evaluation of the inductive and deductive group

approaches to teaching selected word analysis generalizations to disabledreaders in eighth and ninth grade. November 1966.

No. 13 Boe, Barbara L. The ability of secondary school pupils to perceive theplane sections of selected solid figures. December 1966.No. 14 Steffe, L. P. The performance of first grade children in four levels ofconservation of numerousness and three IQ groups when solving arith-metic addition problems. December 1966.

No. 15 Sweet, R. C. Educational attainment and attitudes toward school as a func-tion of feedback in the form of teachers' written comments. December1966.

OCCASIONAL PAPERS

No. 1 Staats, A. W. Emotions and images in language: A learning analysis oftheir acquisition and function. June 1966.

No. 2 Davis, G. A. The current status of research and theory in human problemsolving. June 1966.No. 3 Klausmeier, H. J., Goodwin, W. L. Prasch, J., & Goodson, M. R.

With an introduction by W. G. Findley. Project MODELS: Maximizingopportunities for development a nd experimentation in learning in theschools. July 1966.

No. 4 Otto, W. The relationship of reactive inhibition and school achievement;Theory, research, and implications. September 1966.

No. 5 Baker, F. B. The development of a computer model of the concept-attain-ment process: A preliminary report. October 1966.

PREFACE

This technical report is based on Roberta R. Weissglass' seminar report,prepared under the direction of Wayne Otto, Associate Professor of Curricu-lum and Instruction.

The R & D Center for Learning and Re-education has as its primary goalthe improvement of cognitive learning in children and youth, commensuratewith good personality development. Through synthesizing present knowledgeand concla,.-.ting research to generate new knowledge, we are extending the under-standing of human learning and the variables associated with h efficiency ofschool learning. Knowledge is being focused upon the three main problemareas of the Center: developing exemplary instructional systems, refining thescience of human behavior and learning as well as the technology of instruc-tion, and inventing new models for school experimentation, development activ-ities, and so on.

Within the broad Center goal of improving cognitive learning in children,the goal of the reading group is to understand causes for reading problems inorder to strengthen the reading skills needed for efficient cognitive develop-ment in subject matter areas. Reading is conceived as a tool, not as a contentarea. Mrs. Weissglass designed this study to determine the effect, if any, ofmemory span upon the cue patterns of young children as they begin to acquireskill in word recognition. If the role of memory span in readiness for wordrecognition skill can be clarified, then there will be one more basis for makingprognostic assessments of children's capacity for learning to read. The pre-sent study is a step toward that goal: relevancy of memory span is demonstratedand limitations of the task used in the experiment have implications for improvingthe task.

Herbert J. KlausmeierCo-Director for Research

ACKNOWLEDGMENTS

I wish to express my gratitude to Professor Wayne Otto for his guidance

throughout all phases of this study. I also wish to thank Dr. David H. Ford

for many comments and suggestions, and Tom Houston and Larry Toothaker

for their patient assistance during the many hours spent designing the study

and analyzing the data.

Appreciation is expressed, also, to the following personnel in theSun

Prairie, Wisconsin, school system: Mr. Oliver Berge, Superintendent; Mr.

Bernard Dunn, Supervising Principal; Mr. Lawrence Danielson, Principal,

Northside School; Mr. Martin French, Principal, Southside School; and the

kindergarten teachers who so graciously permitted me entrance into their

classrooms.

R. R. W.

v

CONTENTS

Page

List of Tables and Figure ix

Abstract xi

I. Rationale for the Study 1

IL Method 4Selection of Subjects 4Materials 4Procedure 5

Design and Analysis 5

III. Results 7

Chi Square Analysis of Cue Frequencies 7

Long Memory Span Groups 7

Middle Memory Span Groups 7

Short Memory Span Groups 8Sex Across Memory Span 8Forward Memory Span Groups 9

Analysis of Cue Preference Rankings 9

IV. Discussion 11

V. Summary, Limitations, and Implications 13

References 14

Appendix 16

vii

LIST OF TABLES AND FIGURES

Table Page

1 Sample Groups as Divided on the Bask of WISC DigitSpan Subtest 6

2 Cue Preference Frequencies 7

3 Chi Square Values for Cue Preference Frequencies 8

4 Friedman's Analysis of Variance on Cue PreferenceRanks

5 Sign Test of Cue Strengths

10

10

Figure

1 Mean Cue Frequencies Across Memory Span Groups 9

ix

ABSTRACT

Kindergarten children were used as subjects in an attempt to discoverwhether children with short and long memory spans use different cues to recog-nize words, and whether memory span tests are effective predictors of diffi-culty in learning to read. It was further hoped that children would exhibit aconsistent pattern of choices over trials, which might have implications forteaching word recognition and other reading skills.

The WISC Digit Span Subtest, which measured memory span, was used asthe screening device, and 5-letter nonsense words comprised the word recog-nition task. Children were required to select from a group of nonsense wordsthe one similar to the word that had just been shown to them. Each word inthe response group contained one cue which appeared in the same position asin the stimulus word, with the other cues held constant_ Five cues were ex-aminedpositions 1, 2,3,4, and 5and each subject had an equal opportunityto respond to each cue.

All groups showed a preference for Cue 1 and there was a tendency to re-spond to Cue 3 and Cue 5. An avoidanceeffect seemed to be operating withCue 2 and Cue 4. Results on Friedman's Analysis of Variance of Ranks indi-cated that random selection may have been operating for boys and girls in thelow memory span group, while for the middle and high memory span boys apreference pattern was not likely to arise by chance.

xi

RATIONALE FOR THE STUDY

The general purpose of this study was todiscover whether children with short and longmemory snans use different cues to recognizewords, and whether or not their pattern ofchoices is consistent over trials. Further-more, it was hoped that insights might begained regarding the effectiveness of usingmemory span tests to predict difficulty inlearning to read.

The study was based upon a recent study byMarchbanks and Levin (1965), which was con-cerned with cuing patterns of children. Intheir report, Marchbanks and Levin revieweda number of studies concerned with the cuesby which words are recognized: "It has beensuggested that children recognize words aswholes (Cattell, 1896; Smith, 1928); by geo-metric shape, outline, or configuration (Bell,1939; Tinker and Paterson, 1940); by famil-iarity with graphe me-phoneme correspondence(Gibson, 1962; Gibson, Gibson, Danielson,Osser, and Hammond, 1962); by dominant let-ters such as ascending or descending letters(Wilson and Fleming, 1938); by initial andterminal letters (Wiley, 1928; Levin and Wat-son, 1963; Woodward, 1962; Levin, Watson,and Feldman, 1964); or by some combinationof these cues (Davidson, 1931; Gates, andBoekker, 1923; Shotty, 1912). " However,many o f these studies used adult subjects;therefore, extrapolation to children who lackfamiliarity with the written language is notclear.

The effect of memory span upon the cuepatterns of children was also considered, forDominowski (1965) has stated short-termmemory plays an important role in conceptattainment. Jensen (1965) agrees that manyof the processes basic to learning may be in-volved in conceptual learning and that some ofthese processes may be the same as those in-volved in short-term memory. Earlier, Burks&Bruce (1965) hypothesized that "poor readersapproach learning situations in a more con-crete manner a s a result of an inability tohandle abstractions. Since the reading pro-

cess... consists of abstractions strongly de-pending on memory functions, these childrenare handicapped" and Anderson (1939) con-cluded that verbal ability is partly dependenton memory span.

The Digit Span s ub test of the WechslerIntelligence Scale f o r Children (WISC), anauditory test, was the instrument chosen toexamine memory span. Other workers, con-cerned with children already disabled in read-ing, have examined the relationship of digitspan and reading ability. Hawkins (1956)found, as did Baumeister, Smith, and Rose(1955), that Digit Span had validity for pre-dicting short-term memory. Graham (1952),Burks and Bruce (1955), and Robeck (1960)found that unsuccessful readers were signifi-cantly low on Digit Span, and on the other sub-tests requiring memory. On the other hand,Altus' (1956) study of WISC patterns showedDigit Span to be significantly higher than othersubtest scores.

According to Rizzo's (1939, p. 208) survey,the concept of memory span can be traced tothe work of Jacobs (1887) who devised a "mem-ory span test to determine the maximal numberof related or unrelated elements which a te s teecan reproduce immediately after a single pre-sentation. " Warren 934) used "logical mem-ory" to designate the capacity to retain mean-ingful material as compared with the term"memory span" which indicated memory formeaningless material. English (1934), Cros-land (1929), and Tinker (1929) used the terms"memory s pa n," "range of attention," and"visual arprehension, " re spec tively, to denotethe same behavior, showing that in generalthere was agreement on what memory span is.

Yet, despite this agreement, the resultsof many of the studies are not comparable be-cause researchers used different kinds of ma-terials, diverse methods of testing and scoring,different groups of subjects, and because stud-ies tended to lack careful controls. In the pre-sent investigation, then, memory span is de-fined specifically as the number of digits which

1

can be repeated orally immediately after asingle presentation.

Recent investigations into the area of mem-ory have been concerned with memory as itrelates to personality and learning theories,but very few researchers have attempted torelate memory span i;o reading ability. Per-haps one of the earliest cti.:>: us sions of memoryspan and reading was a n article (Saunders,1925) in which the author concluded that chil-dren with s hort auditory memory spans showedthe following traits: 1) slowness in acquiringfacility with language; 2) difficulty in learningto read; 3) difficulty using phonics; 4) diffi-culty learning rote material orally. Saundersalso felt that while all reading disabilities arenot allied withpoor memory spans, "yet it canbe stated with certainty that all poor memoryspans are allied with difficulty in reading andspelling." According to Rose (1 958), Bondlikewise concluded that auditory memory abili-ty" . . . is probably of importance as a factorin learning to read by the look-and-say method,but is apparently of little advantage in the pho-netic type of instruction," and Rizzo (1939)found that the relationship between memoryspan and reading ability was more marked inthe lower grades (Grades 2, 3, and 4) than inthe upper grades.

Others who have since investigated the re-lationship of auditory memory span to readinghave come to similar conclusions, that is,that auditory memory span correlates signifi-cantly with various silent reading test scores(Reynolds, 1953) that tests for auditory mem-ory span are difficult for children who havereading difficulties (Rose, 1958) and that therelationship between auditory memory spanand the development of adequate word recog-nition is statistically significant (Poling 1953).Yet, it should be clear that poor auditorymemory span is not necessarily a cause forreading difficulty. It may be that poor audi-tory memory is merely a symptom of someother difficultyand, as such, is .t one factorcontributing to reading difficulty.

By grouping c hildr e n according co W1SCmemory span scores, the intent was to examinepossible relationships between cue patternsand memory span. This procedure requiredthat some assumptions be made regarding thevalidity o f comparing a number task with aletter task. Blankenship (1938) and Brener(1940) reviewed experiments which indicatedthat the type of material does affect memoryspan scores, but there was no agreement onthe order of difficulty of various materials.In general, however, nonsense syllables were

2

found to be harder to remember than digits.This is not entirely relevant here for the wordidentification task employed did not requirechildren to remember the nonsense words

Pertinent here a r e areas which hay.. Gotbeen fully investigated: the reliability of testsof memory span and the relationship betweenauditory and visual span. Attempts made inthese directions cannot adequately be evaluated,not only because of the varying c onditions underwhich experiments have been conducted, butalso because methods of pre sentation and scor-ing have lacked constancy. Rose (1958) foundthat the visual memory test on the Stanford-Binet was not as difficult for children withreading difficulties a s the auditory memoryspan tests. She speculated that this may havebeen due to the design of the tests. The digittest did not presenta "pattern or unified whole"as did the visual memory test. The questionraised in this regard was whether the child isbetter able to remember a unified whole thandetails.

The word-recognition task used in the pres-ent study did not require words or letters tobe repeated or reproduced; rather it requiredrecognition. Whether or not this more nearlyequalized the two tasks is difficult to deter-mine, particularly since a delay interval inthe word task added a variable not present inthe digit task. Here, the relevant literatureis again inconclusive: Baumeister, Smith, andRose (1965), using objects in a picture, andBorkowski (1965), using trigrams, found thatresponses can be delayed for 12 seconds and15 seconds, respectively, but that beyondthese intervals performance begins to deteri-orate. This is inconsistent with Griffith'sfindings, as pointed out by Baumeister et al.Griffith found "no reliable effect due to thedelay interval" even when retardates' respon-ses were delayed for 120 seconds.

Kindergarten children served in the studybecause they were considered reasonably naivesubjects, having had no formal instruction inreading. This was felt to offer two advantages.First, because the children had not been taughtth=. left-to-right sequence required in reading,it couldbe assumed they would cue to the poc-1-tion most natural for them. Thus, a frequentlyrecurring pattern for either group might haveimplications for teaching word recognition andother reading skills. Second, as nonreaders,theywould not be able to utilize letter blendingand other similar knowleege to facilitate theirremembering the word; .-other, the y wouldeither concentrate on the individual letters or

develop their own strategy, and the effects oflearning would be minimized.

The present study, although not specificallyconcerned with the relationship between mem-ory span and intelligence, nevertheless tookinto account a possible correlation between thetwo. This was done because: 1) when manychildren, during the initial screening task,were unable to understand what was expectedof them in the digits backward task, it seemedthat an intelligence factor might be causing theproblem, and 2) previous studies have indi-cateda relationship between memory span andintelligence. Because the results of the exist-ing studies are so varied, making it difficultto determine the true correlation betweenmemory span and intelligence, two groups,whose scores on the WISC did not include thedigits backward task, were added to this study.

It is obvious at this point that measuringmemory span is an elusive task, not only be-

cause previous research is largely inconclu-sive, but also because of: the many factors thatmay affect memory span. In addition, the re-lationship of memory span to reading abilityand to intelligence is not clear. Yet, threeconclusions seem warranted. First, becausememory play s an important role in certaintypes of concept attainment and because read-ing r e q u i r e s utilization of many concepts,success inreading depends upon certain mem-ory functions. Second, limited auditory mem-ory is a factor contributing to reading diffi-culty. This conclusion is supported by evi-dence which indicates that good readers tendto have longer auditory memory sp-tns thanpoor readers, although the relationship is moremarked :.n the lower grades than in the uppergrades. This d, Digit Span appears to providea valid basis for predicting short-term mem-ory even though memory span varies with thematerials used.

II

METHOD

SELECTION OF SUBJECTS

Subjects were chosen from among 168 kin-de rgarten children in Sun Prairie, Wisconsin,on the basis of an individual screening pro-cedure designed to identify those childrenhaving the shortest and longest memory spans.The Digit Span subtest of the Wechsler Intel-ligence Scale for Children was used for thescreening. This subtest, which is given oral-ly, yields three scoresa forward score, abackward score, and a total score which is acombination of the forward and backwardscores.

Each child was asked first to listen to theexaminer, who read digits at one-second in-tervals, and then to repeat what he had heard.The examiner began with three digits, in-creasing the number of digits by one eachtime the child repeated the digits perfectly.The maximum number of digits, as well asthe maximum possible score on the digits-forward section was 9. After the child hadreached his maximum, he was given the digits-backward test. This time, instead of repeat-ing what he had heard, the child was asked tosay the numbers backwards. The highe st pos-sible score for digits backward was 8; a per-fect total score was 17.

Children whose total scores contained ascore of zero in either the digits forward orbackward taskwere eliminated after the ran-dom selection of children with the shortestmemory spans had been made, because toomany children had scores of zero. Had allof them remained as subjects the five groups,into which the subjects were divided, wouldnot have been comparable in size. Thus, ofthe original 168 children screened, 96 sub-jects remained and were subsequently pre-sented with the word-recognition test.

MATERIALS

The task comprised 26 groups of five-letternonsense words; each word was typed, using

4

a primary typewriter, on a separate 4 x 6card in lower case letters. In order to ex-amine all of the positions, and to give eachsubject an equal opportunity to respond toeach cue, it was necessary that each cue ap-pear in combination with every other cue anequal number of times over the 26 groups ofcards (hereafter called "trials"). In otherwords, the positions were combined as fol-lows: 1 and 2, 1 and 3, 1 and 4, 1 and 5, 2and 3, 2 and 4, etc. This grouping of cuesresulted in some trials having four responsecards and others having five or six. Eachtrial, therefore, contained one stimulus cardand four to six response cards, the latterbeing numbe red on the back to facilitatescoring.

The response cards were designed to con-tain all possible responses the subject mightmake. That is, each card contained a cue(letter) which appeared in the same positionas in the original word on the stimulus card.Five cues were examinedPositions 1,2,3,4, and 5. Tor example, the stimulus cardI'm o g e d" was followed by the response cards"mythi" (Cue 1 present), "borsu" (Cue 2 pres-ent), "eagle" (Cue 3 present), "fimet" (Cue4 present), and "jukad" (Cue 5 present) ran-domly arranged.

In addition to the single cues, every cuewas presented in combination with every othercue, as explained above. For example, thestimulus card "s eg of" was followed by theresponse cards "skgaf," "stgor," "segay,""segip, " "slgom, " and "shguf. " Each of theseresponse cards contains Cue 1 and Cue 3 ofthe original word. In addition, each word con-tains one further cue: "skgaf" and "shguf"contain Cue 5; "stgor" and "slgom" containCue 4; "segay" and "segip" contain Cue 2.

The purpose of the duplication of cues wasto provide for more of a choice than wouldhave been available had the subjectbeen shownjust three response cards. It is possible thatmany children may have refused to chooseamong only three cards, since it may have

been obvious that none looked like the stimu-lus word. The task was constructed to giveeach subject an equal opportunity to recognizethe word on the basis of any of the five cues.

Both the stimulus and response words weretaken from those used ed Marchbanks andLevin (1965) in their study with some altera-tions being made to accommodate the new sit-uation. First, shape was omitted as a cue,since Marchbanks an d Levin reported thatshape had been the least used cue in both theirtrigram and pentagram series. This neces-sitated a second alteration: increasing thenumber of choices available where the deletionof shape as a cue left too few response words.A third variation was the omission of the tri-gram task. Marchbanks and Levin had foundthat strong competition existed between thefirst an d last letters in word recognition.Owing to the possibility that this effect mayhave beendue either to the open-endedness ofa short word (i.e. , the children might havebeen cuing to the first and last letters of thetrigrams because the complete word was eas-ily seen and there was a great deal of spacenext to those letters) or to children's abilityto remember the complete word, the trigramtask was eliminated in favor of a more inten-sive study of the pentagram task.

Four identical sets of cards were preparedfor the four examiners. 'A form containingthe numbers 1-125 (to correspond with the125 response cards ) was prepared for the re-cording of each child's choices. The stimu-lus cards appeared in the same order in eachof the four sets, but the response cards wererandomly arranged within trials.

PROCEDURE

Individual subjects were presented with theword-recognition task by ad illt examiners.The task was explained as follows: We aregoing to play a word game. I will show youa word. After I take it away I will show youseveral other words. You are to point to theword you just saw or to the one most like it."The subject was then given two sample tri-gram recognition trials to orient him to thetask. The examiner was free to re-explainthe task to the subjectduring this time if nec-essary. The trigrams were immediately fol-lowed by the 26-trial pentagram task withoutthe child's being made aware that the trigramswere to be discarded as practice.

Within each trial the subject was shown aword on a stimulus card and was allowed to

look at it for a maximum of 30 seconds. Thecard was then withdrawn from sight and theresponse, cards for that trial were randomlyarranged on the table. The child was againtold to point to the word he had just seen orto the one most like it. His response waschecked on the scoring form, and the taskcontinued through the 26 trials.

DESIGN AND ANALYSIS

Scores on the Digit Span subtest were ar-rived at on the basis of the forward and back-ward tasks as described earlier. During theinitial screening task, many children wereunable to understand what wa s expected ofthem in the digits-backward task. In addi-tion, there were numerous scores of zero,particularly on the digits-backward task. Be-cause level of intelligence might be the causeof both phenomena, it seemed reasonable toselect groups in such a way that some wouldcontain no zero scores. There was reason tobelieve that such grouping might yield usefulinformatiori. Therefore, three groups wereset up according to Total Score, which in-cludedboth tasks, and two groups were addedon the basis of just the Forward Score.

After the completion of the word-recogni-tion task, 18 of the 96 subjects were elimi-nated owing to mistakes in scoring. There-fore, a total of 78 subjects-41 boys and 37girlswere included in the analyses. Thefive groups contained 22 subjects each, 12boys and 10 girls, with the two Forward Scoregroups containing several subjects who alsoappeared in either the short, middle, or longmemory span groups.

After the testing was finished, the num-bered responses were converted to the cueposition which was represented by each word.A tally of responses by cue was made withinand between the groups and sexes. The cueswere subdivided into primes; that is, insteadof Cues 1,2, 3, 4, and 5 being tallied, Cues1, 1', 2, 2', 3,3', etc., were tallied. Thiswas done for three reasons. First, becausemany response cards within individual trialsduplicate a particular cue (as explained ear-lier), it was reasoned that the primes wouldassist in differentiating between the two pos-sible choices and hopefully add to the infor-mation yield. Using the example previouslymentioned, if the stimulus card "s e g o f" hasas two of its six response choices "skgaf" and"shguf," both o f which contain Cue 5, theimmediate question is which of the two words

5

Table i

Sample Groups As Divided On TheOf WISC Digit Span Subtest

Basis

Group Girls Boys

Total Score GroupsI - Short Memory Span

(0-4 digits) N=10 N=12II - Middle Memory Spana

(5-6 digits) N=10 N=12IiI - Long Memory Spana

(8-12 digits) N=10 N=12

Forward Scores GroupsbIV - Short Memory Span

(0-3 digits) N=10 N=12V - Long Memory Span

(5-7 digits) N=10 N=12

allo zeroes are included in either the Forwardor Backward score.

bSeveral subjects in each group appear againin one of each of the Total Score groups.

6

will the subject choose if he is cuing to posi-tion five. Second, it seemed desirable tocompare the total times each of the two cueswas chosen. Third, it was felt that the datamight contain the answer to why one was cho-sen more often than the other.

The primes were assigned to every secondappearance of a cue within a trial. In Trial5, for example, Cards 17, 18, 19, 20, 21,and 22 contained Cues 5,4,3,3,4,5, respec-tively. Since each cue appeared twice, primeswere assigned to Cards 20, 21, and 22 (whichmarked the second appearance of the cues)so that the chart then read: 5, 4, 3, 3', 4',V.

With the cues subdivided in this manner,comparisons were made across cues 1) be-tween all of the boys and all of the girls, 2)among boys in Groups I, II, and III, 3) amonggirls in Groups I, II, and III, 4) between boysin Groups IV and V, 5)between girls in GroupsIV and V, and 6) between boys and girls ineach of the five groups taken individually.

III

RESULTS

The expectation was that the long memoryspan group would respond to Cue 1 or Cue 5more often than to any other cue because itwas assumed that in the s canning proce s s be-ginning a nd ending letters would dominate.It was further speculated that because none ofthe response cards replicated the stimuluscard the long memory span subjects wouldstart either at the far left or far right whenlooking for a familiar letter; thus, if no cardexactly like the stimulus card were found thetendencywouldbe to choose the card that con-tained Cue 1 or Cue 5.

For the short memory span group, it wasexpected that the last letter the subjects sawmight be remembered; thus, if they had beenscanning from left to right, they would tend tolook for Cue 5. If during the task they real-ized that they could not retain the stimulus in-formation, it was anticipated that they mighttry to keep just a single letter in mind. Thusthe feeling was that these subjects, more thaneither of the other two groups, would cuerandomly.

CHI SQUARE ANALYSIS OF CUE FREQUENCIES

To test the null hypothesis that cues wouldbe chosen with equal frequency, chi squaretests were used. Single and combined (prime)cues have been combined for discussionpurposes.

Long Memory Span Groups

For the boys, Cue 5 was the second mostutilized cue (Figure 1, Table 2)and Cue 4 wasthe weakest cue. The girls responded some-what differently, their weakest cue being Cue2 and the second strongest cue being Cue 4(Figure 1; Table 2). Both boys and girls withlong memory spans responded to Cue 1 signif-icantly more often than to any other cue (Table3). Listing the cues in the order of frequencyof choice, starting with the most frequently

chosen cue, we have these patterns: Boys-1,5,3,2,4; Girls-1,4,3,5,2:

Table 2

Cue Preference Frequencies

SexMemor TS zsl_.._..._.___:1_

Cue Hjgh Middle Low Total

1 103 95 82 2802 51 49 60 160

Boys 3 54 60 60 1744 44 57 52 1535 60 51 58 169

1 85 72 67 2242 39 47 45 131

Girls 3 45 42 61 1484 49 40 37 1265 42 59 50 151

1 188 167 149 5042 90 96 105 291

Total 3 99 102 121 3224 93 97 89 2795 102 110 108 320

Middle Memory Span Groups

The subjects in the middle memory spangroup were considered comparable to thosewith long memory spans, since the digit spanscores of both of these groups contained nozeroes. In other words, their Total Scoresdid not include a zero in either the digits for-ward or backward task. Considering that thegroups were similar, finding that Cue 1 wasthe most utilized cue was not unexpected. Forthe boys, significance beyond the .05 levelwas calculated (Table 3). The order of cuepreference for boys was 1,3,4,5,2. For thegirls, there was no statistically significant

7

Table 3

Chi Square Values for Cue Preference Frequencies

SexMemory Span

Cue High Middle Low Total

Boys

Girls

Total

1

23

45

1

23

45

1

23

45

26.416*2.082

. 5655. 425

. 092

20.942 **3. 250

. 942

. 1711.923

47. 351**5.2042.0734.0031.344

17.0315.755

. 092

.4672.082

7.692. 480

1.9232.769

. 942

24.1842.9591.3442.646

. 154

6.669.092.092

1.733.310

5.769.942

1.5575.769

.076

10.464*.772.380

5.639.358

46. 003**3. 952

. 9306.2482.566

29.641**4. 006.410

5.769. 160

75. 339**7.9391.309

12. 009*1.568

* = p > . 05** = p > . 01Degrees of freedom = 4Critical values were: p > .05, 9.488; p > .01, 13.2777

p> . '1, 7.779; p .001, 18.46

preference for any single cue; order of pref-erence was 1,5,2,3,4.

Combining sexes for this group, preferenceof Cue 1 was highly significant (Table 3), andorder of cues was 1, 5,3,4,2.

Short Memory Span Groups

As stated earlier, it was thought that thepattern of results for the long and short ?mem-ory span groups would be dissimilar in termsof the strongest cue. However, a visual sur-vey of the cue choices (see Figure 1) for thelatter group reveals that the differences arenot large.

For the boys, the chi square test showedno statistically significant preference for anycue (Table 3). The observed order of pref-erence was 1,2-3 (tie), 5,4. For the girls,no cue frequency reached significance, butobserved order of cue preference was 1,3,5,2,4.

For the low group as a whi-le, Cue 1 waschosen significantly more frequently thanother rues, and order of preference was 1,3,5,2,4. (Tables 1 and 2)

8

Sex Across Memory Span

The three groups of boys were combinedand the total choices per cue were tested usingchi square. Cue 1, of course, had a frequencysignificantly different from chance; and Cue 4was utilized with infrequency (Table 2). Theorder of cue preference was 1,3,5, 2, 4.

The girls' choice: were similarly com-pared. They, too, responded significantly toCue 1, andagainCue 4was utilized least fre-quently. Girls' order of cue preference was1,5,3,2,4.

Comparing across sexes and memory span,the pattern of cue preferences was 1, 3, 5, 2, 4,Cue 1 being utilized significantly more thanchance (p .001), and Cue 4 being used sig-nificantly less than chance (p > . 05, Table 3).The competition between Cue 3 and Cue 5 forsecond most utilized cue which was evidentwith the low memory groups (Tables 2 and 3)was again operating here.

Since Cue 1 was definitely preferred by allgroups, a comparison was made of the othercue choices for those trials where Cue 1 was

not available as a possible choice. In makingthis comparison for all boys, the followingdominance pattern appeared: 1> 2 > 3 > 5 >4;and for the girls: 1 > 2 > 3 > 5 > 4. Fromthese results, we can conclude that both boysand girls exhibited the same dominance patternof cuing. This does not mean that boys andgirls showed the same frequency of choices,but that, when theywere forced to choose be-tween certain cues, the same ones dominatedfor both sexes.

Note that since the possible range of fre-quencies for a given cue was 0 to 15n, ratherthan 0 to 26n (where n is the number of sub-jects), the theoretical variance ender the nullhypothesis is smaller than that assumed forthe chi square distribution; thus, the criticalvalues required for significance are somewhatconservative.

cuU

....O

.00o)00oitG

9.0

8.5

8.0

7.5

7.0

6.5

6.0

5.5

5.0

4,5

4.0

3.5

3.0

Forward Memory Span Groups

Of the boys in Forward 0-3 and 5-7, sevenin eachgroupwere also in Groups 0-4 or 8-12.Of the ten girls in each of the forward groups,five appeared again in Group 0-4, and anotherseven appeared in Group 8-12. Because therewas considerable overlap of subjects, com-parisons of the forward groups with othergroups would have been difficult to interpret,and so shall not be pursued further.

ANALYSIS OF CUE PREFERENCE RANKINGS

Because of the conservative bias of the chisquare testwith regard to these data, frequen-cies of cue preferences were obtained for eachindividual, summed, and converted to ranks.

pm Girls(- -)8 - 12 ( ) 8 - 12

(-13--a-0--0)5 - 6 5 - 6 6-o--oo--o--)0- 4 0 - 4 ..-io.-----)(O 0 0 b)

1 2Cue

3 4

Fig. 1. Mean cue frequencies across memory span groups.

5

9

Friedman's Analysis of Variance of Ranks(Ferguson, 1959) was performed on the result-ing ordinal data in order to determine whethera cue preference pattern could arise by chance.If the null hypothesis that conditions did notdiffe,.. among the population were true, thenthe distribution of ranks would be expected toappear in all columns with about equal fre-quency. If the subjects' scores were dependenton some condition (that is, if the null hypoth-esis were false), then the rank totals wouldvary.

Results were significant for the high andmiddle memory span boys (Table 4), indicat-ing the improbability of a pattern's arising bychance. Results for bothboys and girls in thelow memory span group were not significant,indicating that random selection may have beenoperating.

Table 4

Friedman's Analysis of Varianceon Cue Preference Ranks

Memory SpanSex High Middle Low Total

Boys 10.45* 11.82* 2.65 24.57 **

Girls 6.74 4.16 6.08 11.32*

Total 14. 63** 21. 61** 6. 991 29. 45**

* = p < . 05** = p < 01Degrees of freedom = 4Chi square values were: p < .05, 9. 188

p <.01, 13.277p < .10, 7. 779p <.001, 18.46

Using the chi square test for the two groupscombined, that is, vexes and memory span,significance was reached for the high and mid-dle groups. This would indicate that somethingother than random selection was operating.

As stated earlier, a comparison acrosssexes and memory span revealed a pattern ofcue preferences 1,3,5,2,4. The W-shapedtrends (Figure 1) in the totals of the groups'choices seem to imply the existence of signif-icant differences between cues. To test thestrengthof individual cues and groups of cuesthe sign test (Freund, 1960) was performedon the raw data for individual subjects' totalchoices per cue. Since there was a tendencyfor the most frequent choices to be Cues 1,3, and 5 and for the least preferred choices

10

to be Cues 2 and 4, they were combined intotwo groups for the analysisCues 1, 3, 5 andCues 2, 4. Results were significant for allsubjects (p< .001, Table 5) indicating agreater preference for the extreme and medialcues versus Cues 2 and 4 than can be assignedto chance.

Both short and long memory span groupschose Cues 1, 3, and 5 significantly (p < . 01)more often than Cues 2 and 4, while the sig-nificance level reached by the average mem-ory span group and the Forward 5-7 groupwas p< .05. Results were not significant forthe Forward 0-3 group.

Individual subjects' total choices for Cue 1versus Cue 5 were compared next. Resultswere significantfor all subjects (p.4. 001) andwithin the three memory span groups in favorof Cue 1.

Since Cue 1 was preferred by all groups,the above results, which included Cue 1 in theanalysis, were not unexpected. It seemedlogical, therefore, to compare Cues 3 and 5with Cues 2 and 4 to see if either of them, inaddition to Cue 1, was contributing signifi-cantly to the dUferences. Results were justshort of significance at the .01 level, indicat-ing that Cue 1 was the only cue chosen withany great significance. Here the low powerof the sign test may be unfortunate.

Table 5

Sign Test of Cue Strengths

Boys & GirlsCues Cues1,3,5 vs. 2,4

Cues Cues3, 5 vs. 2,41

Long MemorySpan 1 =2.56* 1 = 1.34

Average MemorySpan 1 = 2.13** 1= .89

Short MemorySpan 1 ..: 2.56* 1 = .447

rorward 5-7 1 = 2. 13 **

Forward 0-3 1 = .852

All Subjects 1 = 4. 19* ** 1 = 1.544

* p < .01** p< .05

*** p < . 001

aResults do not include the two Forwardgroups.

IV

DISCUSSION

Observations of the data and subsequentanalyses have indicated some emerging pat-terns among the memory span groups. Thestrength of these patterns, however, cannotbe determined from this study because of un-fortunate limitations in the task. A particu-larly pertinent area to pursue concerns thevarious conditions which may have beenop-erating during the task to a ff e c t responsechoices.

The letters appearing in the duplicate-cueresponse cards, for example, might have in-fluenced subjects' choices. In fact, individualletters may have be en a major influencethroughout the task, since many children chosea card after saying to the examiner, "Thatletter is in my name. " The factor of familiar-ity or unfamiliarity with different letters forindividual subjects may have influenced theirchoices in a variety of ways. They may havechos en a card not on the basis of the five stim-ulus cues, as assumed in the tabulations madeof responses, but rather on the basis of famil-iaritywith a particular letter which they hap-pened to recognize imr-..tdiately.

If this possibility were to be ignored, sinceit was not a lw ay s detectable and certainlycould not be assess ed objectively, it stillwould not be possible to evaluate the choicemade between duplicate-cue response cards.Such an evaluation would involve a compari-son of the two words which contained duplicatecues to see if, over trials, a pattern evolvedshowing that children tended to choose theword of eachpair which contained certain let-ters. This, unfortunately, could not be doilesince in Marchbanks and Levin's pentagramlist letters did ry-4. appear equally over frials.

There was, It; addition, reason to expectthat incompkte learning, and considerableforgetting, of the stimulus pentagrams wouldoccur. Since subjects were nonreaders and,for the most part analphabetic, phonetic en-coding of stimuli into nonsense sounds wa sunlikely. Instead, subjects had to retain fivevisual "bits" of information, whose correct

sequencing (critical to the study) probablywe nt b eyo nd the limit of the subjects' info r ma-tion-p roc e s s ing capabilities. Furthermore,since response pentagrams necessarily re-sembled the stimuli, considerable proactiveinterference would be expected, so that sub-jects might cue on similarities of response toalternative response, rather than response tothe stimulus. Since on items for which cueswere held constant (96% of the items) suchresponse preferences would tend to be basedon unkeyed cues, the recorded cues would beuncorrelated with the "true" cues in thesecases. In other words, where responsechoices began with the sa.ne letter, for exam-ple, subjects might forget about the stimuluscard and choose between those two responsecards, thus making a choice on the basis ofsomething other than one of the five stimuluscues. The result of responding in this marinerwould be an invalidation of the data. Fortu-nately, giving irrelevant cues in random fash-ion would not bias the study by favoring one cueover the others, but would raise the error ofmeasurement.

Kendall's coefficient of concordance (sym-bolized W) is a measure of agreement withina group of raters. In this case, the raterswould be the memory span groups. Becausethere is no reason to suppose that the differentsexes or digit span groups have different pro-portions of the various s canning patternsamong their members, they should all havethe same value of W if all things were equal.The results of the computation follow:

0-4 boys W = .055 0-4 (both sexes)0-4 girls W = .152 W = .079

5-6 boys W = .246 5-6 (both sexes)5-6 girls W = .104 W = .127

8-12 boys W = .218 8-12 (both sexes)8-12 girls W = .169 W = .163

These values of W are quite low, representingpoor agreement. The suggestion is that either

11

different strategies are being employed orthere is much nonretention of stimuli. Thevalues do reflect more agreement in patternsof cue preference for the long memory spangroups than for the short memory span groups,suggesting that all things relating to retentionare roan; not equal and that the low memoryspan group may be responding to many itemsfor reasons other than cues, since they can-not recall thr stimulus anyway, These lowvalues corroborate the results of the Fried-manAnalysis of Variance (Table 4) where onlythe short memory span group differed fromthe other groups.

In testing for significance of cue prefer-ence, no significant differences among the non-initial cues were found, perhaps as a resultof the long-retention intervals which involveda great deal of interference of additional ma-terials. The W-shaped trends (Figure 1) inthe totals of the groups' choices (the tendencyfor the most frequent choices to be 1, 3, and5 and for the least preferred choices to be 2and 4), seem to imply the existence of signif-icant differences between cues; however, thesignificance level does not bear this out, pos-sibly because of the small numbers of subjects.

The findings of the present study comparedfavorably with Marchbanks and Le sin's find-

12

ings. It was generally agreed that Cue 1 wasthe most utilized cue for all groups, and thatCue, 4 was one of the least preferred cues.Competitionbetween Cue land Cue 5 for theirkindergarten boys, however, seemed to relateclosely, in the present study, only to the highmemory span boys and to the girls when fre-quency of choice was combined across mem-ory span. Further differences appeared inin the original study's report that kindergartenboys often based their judgments on Cue 5,utilizing Cue 1 as their second choice. Thiswas not the case in the present study even forthe low memory span boys (as we had antici-pated), although a few of the boys did preferCue 5.

Both studies exhibited W-shaped trends,suggesting an avoidance effect for Cue 2 andCue 4, both of which are imbedded in the otherletters. The present study, unlike the pre-vious one, demorstrated strong competitionbetween Cue 3 and Cue 5. This finding, al-though it has not been adequately measured,is regarded as potentially significant for wordrecognition. More subjects, however, andtighter controls would be required if a com-parison of cue preference were to be pursuedin a meaningful way.

V

SUMMARY, LIMITATIONS, AND IMPLICATIONS

The data indicate that all groups showed apreference for Cue 1; however, the strengthof the preference differed from group to group.There was a tendency, too, across memoryspan groups, for boys and girls to respond toCue 3 and Cue 5, while there seemed to be anavoidance effect for Cue 2 and Cue 4 (althoughsome individual groups did not show this cuepreference). Thus, the W-shaped trends.Why these tendencies occurred is a matter forspeculation; but it seems that cuing to 1,3, and5 may be the most natural pattern of scanningfor untrained children. The ends are open andmay be more eye-catching for that reason.The middle cue might have been a frequentchoice because from that point subjects couldscan in either direction. Cue 2 and Cue 4 arecouched within the words, so would tend to beoverlooked in favor of other positions.

It appears that scanning itself may have af-fected subjects' performance and the datamore thanwas anticipated. This scanning ap-plies not only to that done with the stimuluscards, but with the response L-rds as well.It is difficult to determine why a subject chosea particular card, since there was no instru-ment set up to record eye movements, for ex-ample. Since the examiner could only makesubjective judgments regarding the subjects'strategies, this problem cannot be pursued..Although it seemed at times that subjects werepicking the first card they saw, there was noway of determining, in this study, whether ornot they actually did look at all of the cards.To assess this particular aspect of the prob-lem, it mightbe better to present the responsecards in some controlled pattern, rather thanrandomly. For example, if the cards werearranged in a row, knowing where in the roweach card appeared would throw more lighton the subjects' patterns of choice. By vary-ing cue positions within the rows, the experi-menter would be able to decide if the child

were cuing to the position of the card in therow or to the actual cue positions of the letters.

As mentioned in Chapter IV, it is also dif-ficult to determine to what extent variablessuch as familiaritywith individual letters, thelong-retention intervals involving interferenceof additional materials, or responding to un-keyed cues influenced the entire task. Inaddi-tion, the taskwas not set up to distinguish be-tween the effects of primacy and recency.

With the individual memory span groups,randomness of choice seemed to be operating.This was particularly true for the low mem-ory span groups. This randomness may havebeen the result of subjects' poor retention a-bility and, possibly, low intelligence. If thesegroups' problems were just an inability to re-tain the stimuli, a tendency to make randomselections would be expected, for it would notbe unreasonable to expect responses to bemade for reasons other than cue positions orstimulus letter cues. The Friedman Analysisof Variance of Ranks showed that for the lowboys a preference pattern could arise bychance, while for the other boys it was 95%likely not to arise by chance.

It was expected that a comparison could bemade of cue preferences between the duplicate-cue response cards. This would have requiredthat the letters appear equally over trials byposition. Unfortunately, Marchbanks andLevin's pentagram list was not controlled inthis way, so the comparison could not bemade. In order to see now individual lettersoperate in word-recognition an experimentcould be designed controlling letter difficultyand frequency of appearance over trials; orthe number of letters used in the total taskmight be limited. Any of these designs mightbe expected to show which letters contain themost information for children. This could bea valuable aid in the teaching of new words tobeginning readers.

13

REFERENCES

Altus, Grace T. A WISC profile for retardedreaders . Journal of Consulting Psychology,1956, 20, 155-156.

Anderson, V. A. Auditory memory span test-ed by speech sounds. American Journalof Psychology, 1939, 52, 95-99.

Baumeister, . A., Smith, T. E., and Rose,J. D. The effects of stimulus complexityand retention interval upon short - termmemory. American Journal of Mental De-ficiency, 1965, 70, 129-134.

Blankenship, A. B. Memory span: A reviewof the literature. Psychological Bulletin,1938, 35, 1-25.

Borkowski, J. G. Interference effects inshort-term memory as a fur:ction of levelof intelligence. American Journalof Men-tal Deficiency, 1965, 70, 458-465.

Brener, R. An experimental investigation ofmemory span. Journal of Exm..rimentalPsychplogy, 1940, 26, 467-482.

Burks, H. F. and Bruce, P. The character-istics of poor and good readers as disclosedby the Wechsler Intelligence Scale forChildren. Journal of Educational Psychol-og_y., 1955, 462 488-493.

Crosland, H. R. The influence of letter posi-tion on range of apprehensionA reply toDr. Tinker. Psychological. Bulletin, 1929,26, 375-377.

Dominowski, R. L. Role of memory in con-cept 1 e a. r ning. Psychological Bulletin,1965, 63, 271-280.

English, H. B. Student's dictionaetiological terms. New York: Harper andBrothers, 1934.

14

Ferguson, G. A. Statistical analysis inysy-chology and education. New York: McGraw-Hill, 1959.

Freund, J. E. Modern elementary statistics.(2nd ed.) Englewood Cliffs, N. J. : Prentice-Hall, 1960.

Graham, E. E. Wechsler-Bellevue and WISCscattergrarns of unsuccessful readers.Journal of Consultin Psychology, 1952,16, 268-271.

Hawkins., W. F. Trace a s a predictor ofshort-term memory of organic and familialretardates. American Journal of MentalDeficiency, 1956, 70, 576-579.

Jacobs, Experiments on prehension. Mind,1887, 12, 75-79.

Jensen, A. R. Individual differences in con-cept learning. Paperpresentedat the Con-ference on Analyses of Conceptual Learn-ing. The Research and Development Centerfor Learning and Re-education, Universityof Wisconsin, Madison, Wisconsin.October, 1965.

Marchbanks, G. and Levin, H. Cue s bywhich children recognise words. Journalof Educational Psychology, 1965, 56, 57-61.

Poling, Dorothy L. Auditory deficiencies ofpoor readers. Clinical Studies in Reading,II: 107-111. Supplementary EducationalMonographs No. 77, University of ChicagoPress, 1953.

Reynolds, M. C. A stady of the relationshipsbetween auditory characteristics and spe-cific silent reading abilities. Journal ofEducational Research, 1953, 46 439-449.

Robeck, Mildred C. Subtest patterning ofproblem r ea de r s on WISC. CaliforniaJournal of EducationalResearch, 1960, 11,110-115.

Rose, Florence C. The occurrence of shortauditory memory span among school child-ren referred for diagnosis of reading diffi-culties. Journal of Educational Research,1958, 51, 459-464.

Saunders, Mary J. The short auditory spandisability. Childhood Education, 1931, 8,59-65.

Tinker, M. A. Visual apprehension and per-ception in reading. Psychological Bulletin,1929, 26, 223-240.

Warren, H. C. Dictionary of psychology.Boston: Houghton Mifflin, 1934.

15

NAME:

AGE:

APPENDIX

FORM FOR RECORDING RESPONSE CHOICES

DATE:

1. 26. 51. 76. 101.

2. 27. 52. 77. 102.

3. 28. 53. 78. 103.

4. 29. 54. 79. 104.

5. 30. 55. 80. 105.

6. 31. 56. 81. 106.

7. 32. 57. 82. 107.

8. 33. 58. 83. 1.08.

9. 34. 59. 84. 109.

10. 35. 60. 85. 110.

11. 36. 61. 86. 11.1.ill

12. 37. 62. 87. 112.

13. 38. 63. 88. 113.

14. 39. 64. 89. 114.

15. 40. 65. 90. 115.

16. 41. 66. 91. 116.

17. 42. 67. 92. 117.

18. 43. 68. 93. 118.

19. 14. 69. 94. 119.

20. 45. 70. 95. 120.

21. 46. 71. 95. 121.

22. 47. 72. 96. 122.

23. 48. 73. 97. 123.

24. 49. 74. 98. 124.

25. 50. 75. 99. 125.

16