achievement tests consist of batteries of subtests each of ... · test. performance determines...

20
ED 064 655 AUTHOR TITLE INSTITUTION SPONS AGENCY REPORT NO BUREAU NO PUB DATE CONTRACT NOTE EDRS PRICE DESCRIPTORS DOCUMENT RESUME 24 CG 007 503 Calfee, Robert C.; Venezky, Richard L. Component Skills in Beginning Reading. Wisconsin Univ., Madison. Research and Development Center for Cognitive Learning. Office of Education (rHEW)0 Washington, D.C. TR-60 BR-5-0216 Jul 68 OEC-5-10-154 19p. MF-$0.65 HC-$3.29 *Cognitive Development; Intelligence; *Language Development; Performance Factors; *Reading; Reading Readiness; Reading Skills; *Reading Tests; *Skill Development ABSTRACT Reading research in which different methods or materials have Peen compared has proven inconclusive. TLIs paper is restricted to beginning reading, defined as the acquisition of letter-sound decoding ability, and raises the question: what skills are required by current tests? Available reading readiness and achievement tests consist of batteries of subtests each of which is designed to measure a component skill necessary in reading. However, high intercorrelations between the subtests indicate either that separable skills are not being measured, or that skills develop at the same rate in most children. However, the makeup of the items in the tests is such that ability to follow instructions and general language competence are common factors which enter significantly into performance on all subtests. The experience of psychologists in constructing tests to identify separable skills in language and intelligence indicates that this task is possible but difficult. Current tests are suitable for prediction of reading performance, but tests that evaluate separable skills are urgently needed for further research on the development of the reading process, as well as diagnosis. Examples are presented for articulation and phonetic discrimination. (Author)

Upload: others

Post on 22-May-2020

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: achievement tests consist of batteries of subtests each of ... · Test. performance determines choice of curriculum. program for a child, the vocabulary to which he is exposed, and

ED 064 655

AUTHORTITLEINSTITUTION

SPONS AGENCYREPORT NOBUREAU NOPUB DATECONTRACTNOTE

EDRS PRICEDESCRIPTORS

DOCUMENT RESUME

24 CG 007 503

Calfee, Robert C.; Venezky, Richard L.Component Skills in Beginning Reading.Wisconsin Univ., Madison. Research and DevelopmentCenter for Cognitive Learning.Office of Education (rHEW)0 Washington, D.C.TR-60BR-5-0216Jul 68OEC-5-10-15419p.

MF-$0.65 HC-$3.29*Cognitive Development; Intelligence; *LanguageDevelopment; Performance Factors; *Reading; ReadingReadiness; Reading Skills; *Reading Tests; *SkillDevelopment

ABSTRACTReading research in which different methods or

materials have Peen compared has proven inconclusive. TLIs paper isrestricted to beginning reading, defined as the acquisition ofletter-sound decoding ability, and raises the question: what skillsare required by current tests? Available reading readiness andachievement tests consist of batteries of subtests each of which isdesigned to measure a component skill necessary in reading. However,high intercorrelations between the subtests indicate either thatseparable skills are not being measured, or that skills develop atthe same rate in most children. However, the makeup of the items inthe tests is such that ability to follow instructions and generallanguage competence are common factors which enter significantly intoperformance on all subtests. The experience of psychologists inconstructing tests to identify separable skills in language andintelligence indicates that this task is possible but difficult.Current tests are suitable for prediction of reading performance, buttests that evaluate separable skills are urgently needed for furtherresearch on the development of the reading process, as well asdiagnosis. Examples are presented for articulation and phoneticdiscrimination. (Author)

Page 2: achievement tests consist of batteries of subtests each of ... · Test. performance determines choice of curriculum. program for a child, the vocabulary to which he is exposed, and

U.S. DEPARTMENT OF HEALTH.EDUCATION & WELFAREOFFICE OF EDUCATION

THIS DOCUMENT HAS BEEN REPRO-DUCED EXACTLY AS RECEIVED FROM

THE PERSON OR ORGANIZATION ORIG-INATING IT POINTS OF VIEW OR OPIN-IONS STATED DO NOT NECESSARILYREPRESENT OFFICIAL OFPCE OF EDU-CATION POSITION OR POLICY

W.JSCONSIN RESEARC:i AND DEVELOPMFNT

Page 3: achievement tests consist of batteries of subtests each of ... · Test. performance determines choice of curriculum. program for a child, the vocabulary to which he is exposed, and

Technical Report No. 60

COMPONENT SKILLS IN BEGINNING READING

Robert C. Calfee and Richard L. Venezky

Report from the Project on Language Concepts and Cognitive SkillsRelated to the Acquisition of Literacy

R. C. Calfee and R. L. Venezky, Principal Investigators

Wisconsin Research and DevelopmentCenter for Cognitive LearningThe University of Wisconsin

Madison, Wisconsin

July 1968

The research reported herein was performed pursuant to a contract with the United States Office ofEducation, Department of Health, Education, and Welfare, under the provisions of the Cooperative

Research Program.

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

Page 4: achievement tests consist of batteries of subtests each of ... · Test. performance determines choice of curriculum. program for a child, the vocabulary to which he is exposed, and

PREFACE

A goal of this Center is to create knowledge and theory which can beeffectively utilized in the construction of instructional systems for the schoolsof tomliorow. Some researchers prefer to begin by constructing new instructionalmaterials immediately, while others prefer to begin by studying the fundamentalprocesses presumed to be required for the mastery of such instructional materials.Regardless of the particular approach, a Research and Development Center shouldideally provide an atmosphere within which scholars with different techniques andareas of competence but with common interests can form effective research teams.Such a team is represented in this project, with Professor Calfee from Psychology,and Professor Venezky from English and Computer Sciences.

While their ultimate goal is the construction of reading materials which willoptimize reading acquisition, these researchers are presently attempting to gaina better understanding of the fundamental independent cognitive skills related tothe reading process. This report contains the rationale for their approach and theresults of their analyses of existing tests of component reading skills. Signifi-cantly, the authors conclude that existing diagnostic tests do not measure inde-pendent skills. However, the authors express confidence that sensitive testscan be developed for measuring the independent cognitive skills related to reading,and, in the process for prescribing remedial treatment for those children lackingthese prerequisite skills.

Harold J. FletcherDirector of Program 1

Page 5: achievement tests consist of batteries of subtests each of ... · Test. performance determines choice of curriculum. program for a child, the vocabulary to which he is exposed, and

CONTENTS

List of Tables

Page

vii

Abstract ix

Component Skills in Beginning ReadingIntroduction 1

Components of the Reading Process 1

Tests and Reading 2Language and IQ Testing 7The Trouble with Tests 8Test-Taking and Language Skills 8Measurement of Component Skills 1 0Conclusion 11

References 13

Page 6: achievement tests consist of batteries of subtests each of ... · Test. performance determines choice of curriculum. program for a child, the vocabulary to which he is exposed, and

LIST OF TABLES

Table page

1 Intercorrelations Among Metropolitan ReadinessSubtest Scores 3

2 Correlations of Subtest Scores of MetropolitanReadiness with Subtest Scores on Murphy-Durrelland Pintner-Cunningham Primary Tests 4

3 Predictive Validity of Experimental Edition ofMetropolitan Readiness Test 5

4 Intercorrelations Among Murphy-Durrell Subtests 5

5 Predictive Validity Coefficients for Murphy-DurrellReading Readiness Analysis with StanfordAchievement Test 5

6 Subtest Intercorrelations for.MacMillan ReadinessTest 6

7 Intercorrelations Among Subtests Administered Beforeand After First-Grade Reading Instruction UsingBasal Programs 6

Page 7: achievement tests consist of batteries of subtests each of ... · Test. performance determines choice of curriculum. program for a child, the vocabulary to which he is exposed, and

ABSTRACT

Reading research in which different methods or materials have been com-pared has proven inconclusive. This paper is restricted to beginning reading,defined as the acquisition of letter-sound decoding ability, and raises thequestion: what skills are required current tests? Available reading readi-ness and achievement tests consist of batteries of subtests , each of which isdesigned to measure a component skill necessary in reading. However, highintercorrelations between the subtests indicate either that separable skills arenot being measured, or that skills develop at the same rate in most children.However, the makeup of the items in the tests is such that ability to followinstructions ane general language competence are common factors which entersignificantly into performance on all subtests. The experience of psychologistsin constructing tests to identify separable skills in language and intelligenceindicates that this task is possible but difficult. Current tests are suitable forprediction of reading performance, but tests that evaluate separable skills areurgently needed for further research on the development of the reading process,as well as diagnosis. Examples are presented for articulation and phoneticdiscrimination.

r ix

Page 8: achievement tests consist of batteries of subtests each of ... · Test. performance determines choice of curriculum. program for a child, the vocabulary to which he is exposed, and

COMPONENT SKILLS IN BEGINNING READING

INTRODUCTION

Suppose that the time and money availablefor further research on improvements in readinginst.uction were limited. Given but a year ortwo of support and tight limitations on thebudget, what research would have highestpriority? Russell and Fea (1965) point out intheir review of reading research that no otherarea of the curriculum has garnered such a hugepile of reports. Nevertheless, despite thou-sands of studies over the past 50 years, thereis no clear evidence of improvements in readinginstruction or significant changes in instruc-tional technique.

The majority of experiments on reading haveexplored the relative efficiency of variousmethods or materials. One finds comparisonsbetween phonic programs and whole-word ap-proaches, between ITA and TO, studies of theeffects of different grouping practices, of stress-ing comprehension or drill, and argumentsabout the effectiveness of visual or auditorypresentation. Type font size, the kind of pic-tures accompanying the reading text, the styleand content of the vocabulary, and the lengthand placement of sentences have been examined.Of the many techniques that have been tried,(a) most seem to work with most children, butall fail with many children; (b) there appearsto be no best method; and (c) the efforts of theteacher appear to override in importance theeffects of variation in methods or materialsorso goes the folklore.

Bond and Dykstra (1967) in the report of theCoordinating Center for the Cooperative ResearchProgram in First-Grade Reading Instruction pre-sent data from 27 research projects. From thisextensive report, the conclusions most pertinentto the effectiveness of different methods were(a) various innovative methods, whether phonic,linguistic, orthographical, language experience,or what have you, produced reading achievementscores at the end of the first grade that were

slightly higher than basal reader methods; (b)these differences were generally small and werenot consistently observed by all researchers inall school systems; and (e) there was no evi-dence of differential effectiveness (i.e., it wasnot true that some methods worked better withlow IQ students and others with high IQ stu-dents). It was further concluded that readingachievement must be determined by many fac-tors of equal or greater importance than thoseexamined in the report (i.e., other than readi-ness, IQ, method/material variation, teacherexperience, and community background, etc.).Although attention was directed to the need formore adequate teacher training, none of theteacher variables measured (sex, age, educa-tion, certification, experience, attitude towardteaching, and rated effectiveness) bore anysubstantial relation to reading performance.

A Hawthorne or novelty effect may have ledto the slight superiority of the several innova-tive methods. Chall (1967) has pointed to sev-eral sources of noveltyfresh books and sup-plementary materials, special training for theteacher, and the knowledge by students andparents that they were being treated differently.

COMPONENTS OF THE READING PROCESS

In reviews and research reports, one fre-quently finds reference to the reading process.For example, Russell and Fea (1965) speak of"the reading act as consisting of two compo-nents, (a) identifying the symbol, and (b) ob-taining meaning from the identified symbol."Levin (1966) refers to one skill as decoding thewritten language into its spoken form, and asecond skill as the use of this decoding abilityfor comprehension. Other authors have expressedthe distinction most succinctly as learning toread and reading to learn.

If he is to become literate, the child mustsomehow acquire the ability to decode or trans-late written material to that form of the spoken

1

Page 9: achievement tests consist of batteries of subtests each of ... · Test. performance determines choice of curriculum. program for a child, the vocabulary to which he is exposed, and

language with which he is already familiar.This skill may assume different forms overtime. A beginniny reader, haltingly translatingsingle words or phrases, almost certainly usesdifferent psychological operations than thoseavailable to the accomplished reader who canskim a paragraph or a page in a matter of mo-ments.

In this paper, we will be concerned primarilywith the acquisition of a rudimentary decodingability. If the ability to translate from lettersto sounds is considered a complex skill, thenthe individual must have at his disposal certainmore basic skills which are augmented andintegrated during the acquisition of the new,more complex skill. It is natural to ask, whatare the component skills for reading?

Improvements in the effectiveness of readinginstruction have not come about by variationsin method per se. These variations have toooften bren based on guesses about the readingprocess. More definitive knowledge about theprocess and its component skills might lead toimprovements that have to date eluded us.Testsreadiness, achievement, and diagnos-tic should suggest directions for research,since they are designed to measure componentskills.

Accordingly, it is the purpose of this paperto ask; what skills are required to perform wellon current reading tests? An answer to thisquestion calls for a critical evaluation of ex-isting tests, many of which do not seem toexamine reading ability by any definition ofreading. Instead, both readiness and achieve-ment tests appear to measure general languagecompetence appropriate to middle-class cauca-sian families, and the effects of other kinds ofpreschool training.

Tests play an important role in beginningreading instruction, and necessarily so. Testperformance determines choice of curriculumprogram for a child, the vocabulary to which heis exposed, and the attitudes and expectationsof the teacher toward him.1 Ideally, test per-formance should present information to theteacher about specific disabilities, information

1 Goslin (1968) points out some sociologicalproblems related to standardized al-. tests."One of the most important criticisms of testsis that they contribute to their own validity byfunctioning as self-fulfilling prophecies. . .The likelihood of the optimistic prediction madeon the basis of a high test score is . . . in-creased because the person who scores highreceives special advantages, whereas the in-dividual who does poorly is often denied oppor-tunities."

2

that can dictate the most efficient correctiveaction. The design and suggested use ofreadiness and achievement tests fits naturallyinto the analysis of beginning reading as anintegration of component skills.

Unfortunately, there is little understandingof the reading process to which reference isfrequently made. There are not adequate dataon the stimulus cues used by readers at variouslevels of competence and stages of develop-ment. It is not known how these cues areselected and integrated during oral reading. 2

We don't really understand how basic skills(speaking, seeing and hearing) or more com-plex abilities (the child's linguistic flur_mcyas measured by productive or recognition vo-cabulary, or the length and complexity of sen-tences) enter into the development of readingcompetence, however defined.

TESTS AND READING

Kindergarten children are a motley crew.They differ tremendously in height, weight,physical features, and intellecwal capacityand potential. Some children will have alreadylearned much of what is supposed to be taughtin kindergarten and the first grade, while otherswill not have this advantage.3 The ideal edu-cational system meets each child at his levelof competence and leads him as far as possiblein the direction of the desired instructionalgoals. In this ideal system, tests serve anessential role in initial evaluation of a childand in continuing appraisal of the results ofinstruction. As Stott and Ball (1965) so nicelyphrase it, "The assessment and equitable so-cial management of individual differences inmental ability [are] matters of great practicalimportance [p. 4]." There is a special need toprovide more effective assistance to childrenfrom culturally-disadvantaged backgrounds forwhom present programs of testing and teachingseem especially inappropriate.

2Goodman's (1968) work on oral reading errorsrepresents an important step in this direction.Interesting possibilities are also implicit inthe research on visual search (Neisser, 1968).

3Durkin (1966) studied the progress of earlyreaders, children who already read at the first-grade level when they entered first grade.Although they generally had high IQ's, whenmatched on IQ they still maintained the one-grade advantage as late as the fifth and eighthgrades. Durkin stated that attitude and thehome environment were as important as instruc-tion per se.

Page 10: achievement tests consist of batteries of subtests each of ... · Test. performance determines choice of curriculum. program for a child, the vocabulary to which he is exposed, and

Readiness and achievement tests typicallyconsist of a collection of subtests, each ofwhich is equated, in name at least, with aunique subskill. For example, in the Metro-politan Readiness Test (Hildreth, Griffiths, &McGauvran, 1965), one finds the following listof subskills:a. comprehension and use of oral language,b. visual perception and discrimination,c. auditory discrimination,d. richness of verbal concepts,e. general mental ability; capacity to infer

and to reason,f. knowledge of numerical and quantitive

relationships,9. sensory-motor abilities of the kind required

in handwriting, writing of numerals anddrawing,

h. a:luquate attentiveness; the ability to sitquietly, to listen and to follow directions.

Diagnostic reading tests usually stress thattheir purpose is not evaluation of overall read-ing performance, but determination of thosespecific skills in which a child has deficiencies.According to the Doren Diagnostic Reading Test(1956), "In an achievement test, the number ofcorrect responses is the measure of the degreeof success. In a diagnostic test, it is themistakes which an individual makes that willindicate his areas of need, and an exact identi-fication of the types of error will direct theexaminer to specific remedial work [p. 17]."In the Durrell Analysis of Reading Difficulty(1955) a similar rationale is expressed. "Someof the common difficulties in learning are: (1)lack of adequate background abilities to per-form the task, (2) failure to master the earlyelements on which later abilities are based,and (3) confusions resulting from instructionnot correctly adjusted to the level of ability andthe learning rate of the child, etc."

In fact, differences in the format of diagnos-tic, readiness, and achievement tests are mini-mal. All are comprised of three or more subtests,each designed to evaluate a different subskillassumed to he important in reading. The teacheris usually advised to consider not only the over-all score in readiness or achievement tests, butto look at subtest performance for specificweaknesses, Given present teaching loads,such advice seems impractical.

Furthermore, closer examination reveals thatthe intercorrelations between subtests are sohigh that doubts are raised about whether inde-pendent skills are being tested, or (as an alter-native hypothesis) whether the various skillsrelated to reading develop at significantly dif-ferent rates within the typical individual.

The Metropolitan Readiness Test (METRO)(Hildreth et al., 1965) consists of six subtests.Word Meaning, "a measure of the child's storeof verbal concepts," is a picture vocabularytest with words chosen from kindergarten andprimary word lists. Listening "strives to tapthe child's ability to comprehend syllables andsentences." It is also a picture test. Match-ing requires the child to discriminate and per-ceive correspondences between word forms.Alphabet requires the child to recognize lettersof the alphabet spoken by the examiner. Num-bers tests familiarity with various azithmeticand number concepts. Copying is a test of thechild's ability to copy letter-like forms.

In Table 1 are the subtest correlations fromthe standarization of the METRO.4 The inter-correlations are all substantial. Factor analysisof the data in Table 1 indicated that two ortho-gonal components accounted for about 70% ofthe variance. Tests 1, 2, and 5 loaded on onefactor; Tests 3, 4, 5, and 6 on the other. Itis not obvious how one would interpret thesefacwrs, but they are not basic skills in anyobvious sense.

TABLE 1

lntercorrelations Among Metropolitan ReadinessSubtest Scores, N = 12,225

Subtest 2 3 4 5 6

1. WordMeaning .83 .56 .59 .72 .49

2. Listening .65 .60 .76 .523. Matching .61 .71 .554. Alphabet .74 .50S. Numbers .596. Copying

Note. Reproduced from Metropolitan ReadinessTests. Copyright 1965 by Harcourt,Brace & World, Inc. Reproduced byspecial permission of the publisher.

Another way of determining whether independ-ent skills are being measured is to look at cor-relations between the METRO subtests and othercriterion measures. During the standardizationof the METRO. the PintnerCunningham PrimaryTest (PC; Pintner, Bess, & Durost, 1946) andMurphyDurrell Reading Readiness Analysis(MD; Murphy & Durrell, 1964) were also

4The correlations in Tables 1 to 7 have beencorrected for attenuation using reliability coef-ficients in the test manuals where possible.

3

Page 11: achievement tests consist of batteries of subtests each of ... · Test. performance determines choice of curriculum. program for a child, the vocabulary to which he is exposed, and

administered. The intercorrelations are pre-sented in Table 2. It can be seen that (a) theP-C is highly correlated with all METRO sub-tests , (b) the M-D Learning Rate subtest ap-pear :. measure something different from anyof the METRO subtests, and (c) the other twoM-D subtests correlate highly with the METRO.Except for the METRO Alphabet and M-D LetterNames subtests, which are identical, there isno evidence that subtests of the two readinesstests allow differential evaluation of basicabilities. For example, the correlation betweenthe Listening and Phonemes subtests, both pre-sumably sensitive to auditory discrimination,is about the same as Alphabet and Phonemes.

Data are also provided in the METRO manualon predictive validity with the MetropolitanAchievement and Stanford Achievement Tests(Table 3). None of the intercorrelations differsignificantly from one another. The most reliablepredictors of performance on any of the achieve-ment subtests were Alphabet and Numbers sub-tests .

The Murphy-Durrell Reading ReadinessAnalysis (Murphy & Durrell, 1964) consists ofthree subtests. In the Letter Names test, thechild must identify upper or lower case lettersas the teacher gives their names. The Phonemestest is unique, in that the child is first taughtto segment initial and final consonant sounds,and then is tested on segmentation ability. Forexample, the child hears the words, salt, sand,and soft as examples of the initial /s/ and thenmust mark those pictures whose names beginwith /s,/, e.g. , sun, pillow, ,soap, and basket.The teacher reads the names of the picturedobjects, so the effects of familiarity with thepictured objects should be negligible. In the

TABLE 2

Learning Rate test, thc- Is first taught toassociate the r.-Arnc;1 :`: .7: a ..tnon objects withtheir writte,1 equivalents. For example, theteacher wri...as on the boz-:ri tong:le, hair.. andeyes and then names each word. After thispreliminary session, the children are retested.he teacher pronounces one of the words andthe child must pick out the spelled word froma list. Except for the Letter Names test, theM-D would appear to be tapping differentabilities than the METRO and yet, as notedabove, the various subtests of the METRO cor-relate highly with all but the Learning Rate sub-test. In Table 4 are subtest correlations fromthe M-D standardization. In Table 5 are cor-relations with the Stanford Achievement Test.The data sdeak for themselves. The LearningRate subtest has the lowest predictive validity,a strange result since this subtest involvesprocedures quite similar to those used in read-ing instruction.

Finally, consider the MacMillan ReadingReadiness Test (Harris, 196), which has foursubtests. The Rating Scale consists of a sub-jective evaluation of the pupils readiness bythe kindergarten teacher. Visual Perceptionrequires matching of single letters or words.Auditory Perception measures ability to hearsimilarities and differences in initial consonantsounds and rhyming endings. This is also amatching test, based on pronunciations of keywords by the teacher. Vocabulary and Conceptsis a picture vocabulary test. In Table 6 arepresmted intercorrelations for two standard'.za-tion groups, middle-class first graders andlower socioeconomic Negro children. Again,the intercorrelations are reasonably high forboth populations. It might be noticed that the

Correlations of Subtest Scores of Metropolitan Readiness with S best Scoreson Murphy-Durrell and Pintner-Cunningham Primary Tests,

N = 12,225, inter-test interval 2-3 weeks

Metropolitan Readiness Tests

TestWord

Meaning Listening Matching Alphabet Numbers Copying

Pintner-Cunningham Primary .72 .82 .67 .60 .75 .61

Murphy-Durrell AnalysisPhonemes .60 .61 .54 .60 .64 .50

Letter Names .58 .58 .57 .91 .70 .49

Learning Rate .33 .35 .34 .37 .37 .30

Note. Reproduced from Metropolitan Readiness Tests. Copyright 1965 by Harcourt, Brace &World, Inc. Reptoduced by special permission of the publisher.

4

Page 12: achievement tests consist of batteries of subtests each of ... · Test. performance determines choice of curriculum. program for a child, the vocabulary to which he is exposed, and

TABLE

Predictive N!dity of Experira:I.ntof Edition of Metropolitan Readiness Tests

MetropolitanReadiness Subtest

WordKnowledge

I ')olitan Achievement Test: Primary 1a

WordDiscrimmation

ArithmeticReading Concepts & Skills

1. Word Meaning .53 .45 .48 .522. Listening .56 .50 .54 .573. Matching .55 .50 .54 .524. Alphabet .69 .66 .62 .525. Numbers .65 .59 .63 .686. Copying .45 .42 .45 .44

Stanford Achievement Test: Primary, Form lb

ParagraphMeaning

WordMeaning

ArithmeticSpelling Reasoning

ArithmeticComputation

MetropolitanTotal Test .58 .64 .74 .69 .64

Note.-Reproduced from Metropolitan Readiness Tests. Copyright 1965 by Harcourt, Brace& World, Inc. Reproduced by special permission of the publisher.

aCorrelations based on medians from six groups of students, N per group ranging from 1q1 to 246.b Correlations based on N 96.

TABLE 4

Intercorrelations Among Murphy-DurrellSubtests, N = 12,231

Subtest 2 3

1. Phonemes .62 , S82. Letter Names .403. Learning Rate

Note.-Reproduced from Murphy-DurrellReading Readiness Analysis.Copyright 1965 by Harcourt, Brace& World. Inc. Reproduced by specialpermission of the publisher.

rating by the kindergarten teacher is the bestsingle predictor of test performance. One cando just about as well by asking the teacher torate a child's readiness as by administeringthe entire test.

A comprehensive set of reading subtest inter-correlations is in the Bond and Dykstra (1967)report. All students were given the METRO,M-D, P-C, and Stanford Achievement tests.The Thurstcne Pattern Copying Test and theThurstone-jeffrey Identical Forms Test were

TABLE 5

Predictive Validity Coefficients for Murphy-Durrell Reading Readiness Analysis with

Stanford Achievement Test

Murphy-Durrell

Stanford Achievement: Primary I

Word Paragraph WordMeaning Meaning Study Skills

Phonemes

LetterNames

LearningRate

.67 .64 .70

.60 .61 .60

.52 .54 .43

Note.- Reproduced from Murphy-Durrel1Reading Readiness Analysis. Copy-righi 1965 by Harcourt, Brace & World,Inc. Reproduced by special permissionof the publisher.

also administered to test copying ability andvisual perception. In Table 7 are the subtest

5

Page 13: achievement tests consist of batteries of subtests each of ... · Test. performance determines choice of curriculum. program for a child, the vocabulary to which he is exposed, and

TABLE 6

Subtest Intercorrelations for MacMillan Readinczs Test

Subtest Test I Test II Test III Test IV Total Score

I Rating Scale .50 .43 .56 .96II Visual Perception .57 .43 .56 .76

HI Auditory Perception .67 .68 .48 .64IV Vocabulary and Concepts .54 .60 .69 .79

Total Score .98 .76 .90 .69

Note.- Upper set of r's, Disadvantaged Group, N = 142. Lower set of r's, Middle-classGroup, N = 165.

TABLE 7

Intercorrelations Among Subtests Administered Before and After First-GradeReading Instruction using Basal Programs, Bond and Dykstra (1967), N = 4,266

Subtest 2 3 4 5 6 7 8 9 10 11 12 13

1. M-D Phonemes .52 .42 .35 .29 .43 .38 .50 .54 .50 .51 .40 .532. M-D Letter Names .49 .31 .30 .41 .30 .46 .60 .56 .46 .51 .553. M-D Learning Rate .28 .27 .33 .37 .38 .44 .45 .34 .35 .404. Thurstone Copying .32 .26 .25 .49 .34 .34 .32 .30 .365. Thurstone-Jeffrey .24 .46 .29 .29 .32 .26 .31

Identical Forms.28

6. METRO .77 .51 .42 .38 .61 .33 .41Word Meaning

7. METRO .51 .33 .34 .49 .24 .38Listening

8. P-C Raw Score .50 .47 .59 .35 .49

STANFORD ACHIEVEMENT:

9. Word Reading .87 .62 .71 .8110. Paragraph Meaning .58 .73 .7711. Vocabulary .47 .6712. Spelling .7113. Word Study Skills

Note.- This table is based on an N of 4,266 from 187 classes in 17 projects. Reliabilitycoefficients for Tests 4 and 5 were not available, and hence correlations associatedwith those tests are not corrected for attenuation.

intercorrelations.5 Subtest correlations be-tween and within tests were relatively high,

5The data in Table 7 are based on studentstaught by some form of basal reading program.Correlation matrices for students taught by dif-ferent reading programs, such as ITA, languageexperience, linguistic/phonics, etc., werequite similar. Table 7 is representative, andthe original report may be consulted for details.Dykstra (1967) has reported that data from thesame children at the end of the second gradeyield a similar pattern or results.

6

except for the METRO Listening and Thurstone-Jeffrey Identical Forms Tests, which for thesestudents were also unrelated to the criterionperformance on the Stanford. The P-C vocabu-lary test correlated to the same extent with allof the readiness and achievement subtests.Factor analysis showed that two factors ac-counted for 55% of the variance in Table 7. Thefirst factor loaded most heavily on Tests 2, 9,10, 12, and 13. It appears that (a) ability toidentify the letters of the alphabet and readingachievement at the end of first grade are closelyrelated, and (b) four of the five subtests on theStanford yield similar achievement scores. The

Page 14: achievement tests consist of batteries of subtests each of ... · Test. performance determines choice of curriculum. program for a child, the vocabulary to which he is exposed, and

second factor loaded on Tests 6, 7, and 8,which are all vocabulary tests. Interestingly,knowledge of the letters of the alphabet at thebeginning of first grade predicted readingachievement at the end of first grade as wellor better than vocabulary at the beginning offirst grade, even though these children weretaught and tested by procedures which wouldstress comprehension.

We have not chosen these particular readi-ness tests with any malicious intent. To thecontrary, they appear to constitute the mostadequately constructed and standardized readi-ness tests available. One might conclude thatit is difficult to identify separable skills since,on the face of it, different testing proceduresand materials are represented in the collectionof subtests. An alternative interpretation isthat perceptual and cognitive development aresuch that an individual child is not likely todiffer much in the degree to which he has mas-tered the requisite perceptual and language

LANGUAGE AND ICI TESTING

Psychologists have for some time facedproblems analogous to the measurement of in-dependent reading skills in the assessmen:. ofintelligence and language ability. The firstIQ test was developed by Sir Frances Galton totest his theory of inherited intellect. Galtondevised a battery of tests measuring sensory-motor performance, immediate memory, andother primary skills, but was discouraged tofind that none of these measures bore any sub-stantial relationship to other criteria of intel-ligence. Alfred Binet in France was more suc-cessful in devising tests with immediate prac-tical implications; they predicted school per-formance. Binet's approach proved viable inapplied settings, whereas the research tradi-tion begun by Galton finds its current niche inthe experimental psychology laboratory.

Intelligence tests, like reading readinessand achievement tests, generally consist ofsubtests designed to test presumably independ-ent cognitive functions. The question of therelative independence of the cognitive abilitiestapped by the various subtests has been impor-tant both practically and theoretically. Forexample, the Wechsler Intelligence Scale forChildren (1949) consists of two scales, Verbaland Performance. The Verbal scale is designedto measure language fluency, and the Perform-ance scale, sensory-motor and perceptualability. The subtests for the rwo -Ales arequite different. A typical Verbal item is, "Whatis the population of the United States?", whilea typical Performance subtest requires a child

to put together a simple jigsaw puzzle. Cor-relation between the two scales is about .80(corrected for attenuation, about .86). Thehigh correlation is useful for diagnosis; largedifferences between Verbal and Performancescores are presumed indicative of abnormalintellectual functioning, In the typical child,however, the Verbal and Performance subtestsproduce very nearly the same score.

Construction of tests sensitive to identifiableskills has been important for investigation ofthe "differentiation hypothesis" (cf. Stott &Ball, 1965). The supposition is that early indevelopment, prior to age three, cognitiveabilities are not differentiated to any extent.As a child matures , cognitive abilities maydevelop at unequal rates and hence appear asdifferentiated. The usual test of the hypothesishas relied on correlations among subtests de-signed to measure different skills. A report byMeyers, Dinzman, Orpet, Sitkei, & Watts(1964) is representative. These investigatorsconstructed tests for children between two andsix years of age which were designed to meas-ure four types of basic cognitive abilities;psychomotor, perceptual speed, linguisticability, and figural reasoning. They were in-terested in two questions. First, did the sub-tests measure independent identifiable abili-ties, or could the data be adequately describedby a general intelligence factor? Second, didthe degree of skill differentiation increase withage? Meyers et al, were reasonably success-ful in constructing subtests sensitive to sepa-rate cognitive skills. While the intercorrela-tions were not as low as one might desire(range .04 to .57, median .34), factor analysisshowed that the data were adequately describedby four factors. There was no support for thedifferentiation hypothesis.

A study by Lesser, Fifer, and Clark (1965)provides another example. These investigatorssought to determine whether or not childrenfrom different social classes and cultural groupsin New York City exhibited unique patterns ofmental abilities. They constructed subtests tomeasure Verbal, Reasoning, Number, and Spaceability. Moderate subtest intercorrelationswithin social and ethnic groups were observed(range .12 to .72, median .35). The Reasoningsubtest correlated most highly with the othersubtests, especially Number and Space.Lower-class children performed more poorlythan middle-class children on all subtests inall ethnic groups (percentile difference ofabout 10 points on the average). Chinese andJewish children performed better on the Reason-ing, Number, and Space scales than Negroesand Puerto Ricans, but performance on thesesubtests was not substantially different within

7

Page 15: achievement tests consist of batteries of subtests each of ... · Test. performance determines choice of curriculum. program for a child, the vocabulary to which he is exposed, and

those subgroups. The Jewish children showedbetter and the Puerto Ricans poorer Verbal abilitythan did the Chinese and Negro children, whosescores were similar. Performance on the Verbalscale was thus different from performance onthe other three, but there was less convincingevidence that the nonverbal subtests were meas-uring subtantially different abilities.

As a !Thal example, consider the Illinois Testof Psycho linguistic Ability (ITPA) (McCarthy &Kirk , )963) which is based upon a rather elabo-rate model of language functioning. As childrenmature, presumably new language skills areadded and old skills become further refined.The original test was composed of nine sub-tests , selected to measure skills at severalpoints within the language system. It wasstandardized on children ranging in age from2.5 to 9 years. Subtest intercorrelations weregenerally high for all age groups, and factoranalysis showed that most of the systematicvariance in the tables of intercorrelations couldbe accounted for by a single variable, bestdescribed as general linguistic ability. Therewas no consistent evidence of a systematicdevelopment of specific skills with age.

Further work on the same test by Quereshi(1967), using a different factor analytic tech-nique, led to a more optimistic outcome. Therelative importance of the general factor ap-peared to decrease with age (41% of the vari-ance at age 2.5 to 23% at age 9), and threegroup factors were found, each accounting for10-15% of the variance. The analysis wasbased on a rational division of the subtests onthe ITPA into subsets, and hence made moresense than the unconstrained analysis ofMcCarthy and Kirk. The procedure did not yieldorthogonal factors, and correlations among thefour factors (the general factor and the threegroup factors) were about .45. Quereshi con-cluded that, because of the importance of thegeneral factor and the high factor correlations ,test constructors should concentrate on thegeneral factor. However, again there is evi-dence that tests can be constructed that meas-ure component skills that are to some degreeseparable.

THE TROUBLE WITH TESTS

Tests can be used for several purposesprediction , diagnosis, measurement of aptitude,interest, performance or achievement. In thearea of beginning reading, there is little troublein finding tests to predict performance or read-ing achievement at the end of first grade. Asmentioned previously, a child's ability to namethe letters of the alphabet and the kindergartenteacher's ratings are both reliable predictors.

8

.4.

Correlation continues to resist any efforts tobe equated with causality, however. By theend of first grade, most children have learnedto identify the letters of the alphabet, but manyhave not become satisfactory readers (e.g. ,Olson, 1958). Children who are not able tohandle phonetic discrimination or segmentationare also likely to be poor readers (Durrell &Murphy, 1953). The conclusion has beendrawn that such children must be taught to lis-ten more carefully to what they hear and say.Yet pilot studies in our laboratory and the ex-perience of teachers with whom we have spokensuggest that it is difficult to explain phoneticsegmentation to a child until he learns to read.As soon as the child learns the reading game,i.e. the correspondence between letters andsounds, he acquires a vocabulary which allowshim to talk about phonetic segmentation. Thisis not meant to imply that a nonreader cannotbe taught to segment. The question is whethersegmentation ability is a prerequisite to read-ing, or vice versa.

Performance on readiness subtests has been_put forward as a source of diagnostic informa-tion, yet there is no clear evidence of thevalidity of these measures as diagnostic indi-cators nor is it apparent what remedial actionshould be taken when a child performs poorlyon a readiness subtest. The trouble with read-ing readiness tests is that they do not providemeasures of component skills that are relatedto reading performance in any well definedmanner.

There has been relatively little effort toestablish the validity of diagnostic test pro-cedures in remedial reading. The causal rela-tion between a particular deficiency and readingis established either by fiat or through correla-tional evidence. For example, it is consideredobvious that if a child ci..anot articulate cor-rectly, he will therefore have problems in learn-ing to read. Accordingly, speech therapy isrecommended. To the best of our knowledge,there is little evidence of high correlation be-tween articulation and reading achievement,nor has it been shown that correction of articu-lation .2 el se has any positive effect on readinr;performance.

TEST-TAKING AND LANGUAGE SKILLS

The inclusion of different types of subtestsin readiness tests would seem defensible tothe extent that the subtests are sensitive todifferent skills and insensitive to general abili-ties. Yet there is reason to suspect that cur-rent tests are so constructed that two generalability factors determine whether a child canperform well on any subtest. The first of these

Page 16: achievement tests consist of batteries of subtests each of ... · Test. performance determines choice of curriculum. program for a child, the vocabulary to which he is exposed, and

factors is the ability of the child to follow in-structions, and the second (and related) factoris general language competence. These char-acteristics of the test may be appropriate anduseful in prediction. The problem is that theycompromise the diagnostic value of the test.

Consider some specific examples from theMetropolitan Readiness Test. The first sub-test, Word Meaning, is a picture vocabularytest in which the pupil selects from three pic-tures the one corresponding to a word spokenby the teacher. Presumably, the subtest isdesigned to measure extent of recognition vo-cabulary. The words were selected from stand-ard kindergarten and primary word lists. Yetfrom the construction of the subtest, the selec-tion of target items and alternatives, it is hardtc ascribe performance to extent of recognitionvocabulary alone. Of the sixteen target items,eight (windmill, moose, yarn, knitting, tobog-gan, spectacles (not glasses), blueberry, andmoccasin) are either archaic, specialized, orunfamiliar.6 What remedy is prescribed whena child does poorly on this subtest? The se-lection of alternatives is likewise curious (thetarget item is underlined): walnut, chestnut,acorn; shingled house, brick house, stonehouse; knitting or tatting a bootee, knitting(a larger item), embroidery; hoop, horseshoe,hoof. The ability to select the correct itemdepends on visual discrimination and logicalinference as much as vocabulary. The choiceof vocabulary items appears singularly inappro-priate for urban children, especially those fromlower socioeconomic backgrounds.

Similar comments hold for the Listening sub-test, which is largely a test of inferentialability, attention to visual detail and memory.For example, "in the fall, father rakes theleaves and burns them;" the child is to dis-tinguish between a man lighting a fire in abrick barbecue, a man tossing leaves into abasketful of burning leaves, and a man rakingleaves onto a burning pile. Or again, "It is abig animal. It has four legs like other animals.It has a tail. It has many things other animalshave, but it has one thing they do not have."Pictured are a bear, a horse and an elephant.The test is designed to measure ability to com-prehend sentences. If a child performs poorlyon this test, what should be done?

6Except for windmill and knitting, these wordsare relatively rare. According to the Thorndike-Lorge (1944) count, they are not among the5,000 most common words in English. Thesame comment holds for the items from theMetropolitan Achievement Test mentioned laterin the paper, where only bonnet is among the5,000 most common words.

Investigation of other readiness tests turnsup similar examples. In the Lee-Clark ReadingReadiness Test (1962), the kindergartener isasked to identify a "short-haired dog"; aDoberman, a Saint Bernard, and a cockerspaniel are pictured, but small and with littledetail. l'or another item, the instructions are,"Fut a mark on the two little chickens." Thealternatives are a hen and a pair of chicks,two medium-sized chickens with combs andwattles, and a slightly larger pair (a hen anda rooster, judging from the tail feathers onone). The two middle-sized birds are the cor-rect choice. For another item, the child mustindicate which vehicle carries the most peoplea horse, a jet airplane, a car, or a boat.

The Lee-Clark predicts first-grade readingachievement reasonably well. Data presentedin the manual show that at the end of firstgrade, those children who did most poorly onthe readiness test can be expected to be halfa grade behind their classmates. One canfurther predict that these children will be at arelatively greater disadvantage in later grades.The sad fact seems to be that readiness testinformation can be used only to delay the be-ginning of reading instruction by interventionof "readiness" activities.

Achievement tests, frequently used as cri-terion devices , are also inadequate. Bormuth(1968) has argued that "achievement tests con-structed by current methods have no logicaland objectively demonstrable relation to theinstruction . . . a scot:: on an achievementtest made by (current) procedures must be in-terpreted as a student's response to the testwriter's responses to the instruction."

There are three reading subtests on theMetropolitan Primary Achievement Test. TheWord Knowledge test is designed to measurethe child's sight vocabulary or word recog-nition ability. It is a picture vc.::abulary test.In the Word Discrimination test, the teacherpronounces a word and the child must thenmark the corresponding word from a list offour. In order to perform well on this test,the child must be able either (a) to associatethe pronounced word with its printed repre-sentation, and choose from a set of words thatare visually similar, or (b) remember the wordwhile pronouncing each of the test items andcomparing with the test item. The third test,Reading, requires the child to look at a pic-ture, decide what the picture portrays andthen select the sentence which best describeswhat is happening. (One of the items re-quires the child to infer that a man in a bluesuit who helps children and tells them tostop and go, is a policeman.)

All of these subtests require of the child afair degree of inferential ability, an extensive

9

Page 17: achievement tests consist of batteries of subtests each of ... · Test. performance determines choice of curriculum. program for a child, the vocabulary to which he is exposed, and

reading vocabulary (e.g. , muss, mane, wringer,bonnet, clothespin), and the ability to discrimi-nate very sharply between conceptually similarand conceptually ambiguous items (e.g. , apicture of a turtle going down a road past asleeping rabbit "the turtle is afraid the rabbitwill get ahead of him" or "the rabbit sleepswhile the turtle crawls down the road"). Thereis no question that a bright child who readswell can perform well on all of these tests, orthat a dull child who can't read will do poorly.On the other hand, none of the tests constitutethe most straightforward test of the child'sreading ability, whether one chooses to stressthe decoding or comprehension aspects of read-ing The Word Discrimination test is as mucha test of spelling ability and the clarity of theteacher's articulation as it is the child's abilityto read. A child might be able to read aloudevery word in the test and still perform verypoorly. To be sure, this is an achievementtest, not a diagnostic test. The question re-mains, is this the best approach to the designof an achievement test, and must the design ofachievement tests be such thar they provide nouseful diagnostic information?

MEASUREMENT OF COMPONENT SKILLS

At the beginning of this paper, the questionof research priorities was raised. In answer,it has been suggested that substantial improve-ments in reading instruction will require moredetailed knowledge of the reading process andthe component skills which relate to the devel-opment of this ability. There is an obviousneed for more adequate measures of basic skills.

For the past two years in our research pro-gram at trie Wisconsin Research and Develop-ment Center, we have been investigating articu-lation and phoneme discrimination skills inyoung children preschoolers, kindergartenersand first graders. We quickly discovered thata major hurdle was development of testing pro-cedures that made minimal demands on the child,apart from the skill being measured.

The usual approach in constructing an articu-lation test (e.g. , Templin, 1957) has been toselect pictures of familiar objects until all themajor phonetic contrasts in English are includedin the set. A child is shown each picture andasked to name the object. To do well on thetest, a child must (a) be able to interpret anabstract representation of an object, (b) befamiliar with the object in question (i.e. , rec-ognize it and have an appropriate name for itin the speaking vocabulary), and (c) be able togive the pronunciation correctly. Since theobjects are presumed to be familiar, this typeof testing procedure is sensitive to dialect

10

variations. Problems of recognition, familiar-ity, and dialect should be minimized if thechild repeats a word spoken by the experimenteror recorded on tape. In fact, Snow and Milisen(1954) showed that articulation performance ofchildren with speech defects was significantlybetter on imitation than on picture-naming testsusing the same words.

We have just completed testing the articula-tion ability of over 600 kindergarten and first-grade children using an imitation procedure(Venezky and Calfee, 1968). The details arebeyond the scope of this paper, but certainfindings are germane to the discussion.

In the testing, a child repeated each wordas it was read by a trained experimenter orplayed from a tape recorder. As might be ex-pected, the recorded presentation produced asomewhat higher error rate, but the variabilitybetween schools was twice as great for thelive presentation. It appears that even a trainedtester may contribute to a child's articulationperformance.

There-were marked word-context effects eventhough an imitation procedure was employed.For example, there were 116 errors on the /britcluster in broil, but only 21 in breathe. InitialAD/was mispronounced 3 times in birth, but 40times in beige. These differences might be at-tributed to familiarity or woLd frequency, butnot the /k/ errors in coins (3) vs. cage (32).Context effects are especially noteworthy inlight of the finding that children did not pro-duce uniform patterns of errors. Most of theerrors involved semivowels, A./ and /1/, orfricatives, /s/, /zit, /0/ and /. A childmight make a substitution such as /w/ for /r/in one context but not in others. In less than5% of the children was there evidence that achild was totally unable to produce one or morephonemes.

Substitution or deletion of initial /s/ isfairly common in first graders, especially whenthe sibilant is part of a cluster. To get aclearer picture of the consistency of /s/ errors,a 30-item test was prepared consisting of con-sonant clusters such as /sp/, /sk/, /sw/,etc. Typical items were span, speak, spright,spray, sprawl and sprflg, Of 57 children, 18made at least one //s/ error; of those 19, only2 made more than 4 errors, and none missed allof the 30 items. Thus, even with a difficultphoneme, many children make occasional errors,but very few children are entirely incapable ofproducing the phoneme.

Our data suggest that the phonetic environ-ment may be as important a determinant of per-formance as individual differences. For ex-ample, in clusters such as /br/, /pr/, and/fr/ where a front (bilabial or labio-dental)

GPO 009-394-1

Page 18: achievement tests consist of batteries of subtests each of ... · Test. performance determines choice of curriculum. program for a child, the vocabulary to which he is exposed, and

consonant was followed by a central semivowel,the semivowel was replaced more frequently(usually by ,e/w/) than where the consonantwas central as in /tr/and /dr/. The transitionfrom front to central, which covered a relativelylong articulatory distance, was difficult forthese children. That /r/ was correctly pro-nounced in a less difficult context (from amotoric standpoint) suggests that a child maybe able to discriminate between /r/ and /w/quite well, even though occasional replace-ments occur.

This result is noteworthy because of thefrequent assertion that speech problems reflect(phonetic) discrimination failures. Terriplin's(1957) finding of substantial correlations (.4to .7) between articulation and discriminationhas been taken as support for the assertion.Once again, correlation is not causality. Therelation which remains to be established is theexistence of articulation and discriminationproblems common to specific phonemes or articu-latory features.

Phonetic discrimination has been shown tobe related to reading performance. The usualtesting procedure with children has been asame-different task (e.g., Robbins & Robbins ,

1948). The child is presented with a pair suchas /fa/- fOal and asked whether the two itemswere the same. Unfortunately, the concept ofidentity is not well developed in all kindergar-ten and first grade children. Although manychildren use the words same and different oralike and unalike, their interpretation of theseterms when applied to speech may be differentfrom the experimenter's interpretation. We ranheadlong into this problem early in our testingprogram when one of the children replied "dif-ferent" when shown two cards containing identi-cal geometrical forms. When asked to justifyhis answer, the child pointed out that one ofthe cards had a smudge on it. With older ormore test-sophisticated children, it is easierto communicate the dimensions with regard towhich identity is to be judged. With youngerchildren, or where the material being testedmay pose a new and difficult test for the child

under the best of circumstances, the relevantdimensions may be extremely difficult to inter-pret for the child.

In another type of phonetic discriminationtest, the child is asked to determine whetheror not a criterion phoneme such a.,; / is pres-ent in a familiar word. For example, the childhears exemplars sue':, as sun and soup and thenis required to point to those pictures in a listwhich contain the same initial sound. Thechild must be able to recognize pictures,identify the objects in them, segment and ab-stract the relevant phoneme, and discriminatebetween phonemes. There is no reliable infor-mation about how a child who substitutes /e/for /s/ performs when he is asked to mark wordsbeginning with /s/. In any event, errors onthis type of task may be traced to many sources.

CONCLUSION

Like others, we would like to find moreeffective ways to teach reading. It seemsfutile to introduce more new methods untilnecessary insights into the nature of the read-ing process are established by appropriate re-search. Dissection of the process into itscomponents is an impossible venture wheneach measure correlates with every other meas-ure to the same extent; hence our concern withtesting procedures. We are optimistic aboutthe possibility of finding reliable instrumentssensitive to well defined skins. We would bequite pleased to find tests of articulation anddiscrimination ability were only slightly cor-related with one another, and that neither wassignificantly related to performance on currentreadiness or achievement tests. After all, itis hard to believe that the sum total of a child'sintellectual ability can be measured by hisknowledge of the letters of the alphabet priorto first grade. Reading is a vital skill withoutwhich a child cannot succeed in virtually anyother area. Today, it is possible to predictquite reliably those children who are not goingto make it. This damning prediction must bechanged into a prescription for treatment.

11

Page 19: achievement tests consist of batteries of subtests each of ... · Test. performance determines choice of curriculum. program for a child, the vocabulary to which he is exposed, and

REFERENCES

Bond, G. L., & Dykstra, R. The cooperativeresearch program in first-grade reading in-struction. Reading Res. Quart., 1967, 2,Num. 4 (Complete).

Bormuth, J. On the theory of achievement testitems. Mimeographed paper, Univ. ofChicago, 1968.

Chall, G. Learning to read: the great debate.New York: McGraw-Hill, 1967.

Doren, M. Doren Diagnostic Reading Test.Minneapolis: Educational Test Bureau,1956.

Durkin, D. Children who read early. NewYork: Teacher's College Press, ColumbiaUniv., 1966.

Durrell, D. D. Durrell analysis of readingdifficulty. New York: Harcourt, Brace,& World, 1955.

Durrell, D. D. & Murphy, H. A. The auditorydiscrimination factor in reading readinessand reading disability. Education, 1953,73, 556-560.

Dykstra, R. Continuation of CoordinatingCenter for first-grade reading instructionprograms. (Final Report of Project 6-1651,Contract OEC3-7-001651-04R), Minneapolis:Univ. of Minnesota, 1967.

Goodman, K. S. Reading: a psycholinguisticguessing game. J. Reading Specialist,1968, in press.

Goslin, D. A. Standardized ability tests andtesting. Science, 1968, 159, 851-855.

Harris, A. J. MacMillan reading readinesstest. New York: MacMillan, 1963,

Hildreth, G., Griffiths, N. & McGauvran, M.Metropolitan readiness tests. New York:Harcourt, Brace & World, 1965.

Lee, J. M., & Clark, W. W. Lee-Clarkreading readi less test, manual. NewYork: McGraw-Hill, 1962.

Lesser, G. S., Fifer, G., & Clark, D. 11.Mental apilities of children from differentsocial-class and cultural groups. Monog.Soc. Res. Child Dev., 1965, 30, WholeNo. 4.

Levin, Harry. "Reading research: What, why,and for whom?" Elementary English, 1966,43 138-147.

McCarthy, J. J. & Kirk, S. A. The construc-tion, standardization and statistical char-acteristics of the Illinois test of psycho-linguistic abilities. Madison, Wisc.:Photo Press, Inc., 1963.

Meyers, C. E., Dinzman, . F., Orpet, R.E., Sitkei, E. G., & Watts, C. A. Fourability-factor hypotheses at three preliteratelevels in normal and retarded children.Monog. Soc. Res. Child Dev., 1964, II,Whole No. 4.

Murphy, Helen, & Durrell, D. D. Murphy-Durrell reading readiness analysis. NewYork: Harcourt, Brace & World, 1964.

Neisser, U. Cognitive psychology. NewYork: Appleton-Century-Crofts, 1968.

A. ili-ties as it relates to success in beginning

Pintner, R., Bess, C., & Durost, W. Pintner-Cunningham primary test, form A. (generalability tests, revised). New York: Harcourt,Brace & World, 1946.

Quereshi, M. Y. Patterns of psycholinguisticdevelopment during early and middle child-hood. Educ. psychol. Meas., 1967, 27,353-365.

Robbins, S. D., & Robbins, R. S. Robbinsspeech sound discrimination test. Mag-nolia, Mass.: Expression Co., 1948.

Russell, D. H., & Fea, H. Research onteaching reading. In N. L. Gage (Ed.),Handbook of research on teaching. Chicago:Rand-McNally, 1965, pp. 865-928.

Snow, K., & Milisen R. The influence of oralpictoral representation upon articulationtesting results. J. Speech Hearing Dis.,Monog. Suppl. 4, 1954, 30-36.

Stott, L. H., & Ball, R. S. Infant and pre-school mental tests: review and evaluation.Monog. Soc. Res. Child. Dev., 1965, 30,Whole No. 3.

13

Page 20: achievement tests consist of batteries of subtests each of ... · Test. performance determines choice of curriculum. program for a child, the vocabulary to which he is exposed, and

in children. Minneapolis:sota Press, 1957.

Thorndike, E. L., & Lorge, I.word book of 30,000 wordsColumbia Univ. Press, 194

Venezky, R. L., & Calfee, R.

Univ. of Minne-Temp lin, Mildred C. Certain language skills

The teacher'sNew York:

4.C. The pattern-

ing of articulation errors in kindergartenand first grade. Paper presented at Psycho-linguistics andReading Assn.,

Wechsler, David.

Reading Presession, Intern.Boston, April, 1968.Wechsler intelligence scale

for children, manual. New York: Psycho-logical Corporation, 1949.