Michigan

Embed Size (px)

DESCRIPTION

michigan test

Citation preview

  • MELABTECHNICAL

    MANUAL

    English Language InstituteTesting and Certification Division

    3020 North University BuildingUniversity of Michigan

    Ann Arbor MI 48109-1057

  • Please address all correspondence to:

    The English Language InstituteTesting and Certification Division3020 North University Building

    The University of MichiganAnn Arbor MI 48109-1057

    telephone: (313) 764-2416fax: (313) 763-0369

    email: [email protected]

    Second Printing, August 1996.1994 by the English Language Institute, The University of Michigan. This document may bereproduced or reprinted, in whole or in part, without permission as long as the source is clearlyacknowledged. This document or any reproductions may not be sold.

    The Regents of the University of Michigan: Deane Baker, Ann Arbor; Paul W. Brown, Petoskey; Laurence B. Deitch,Bloomfield Hills; Shirley M. McFee, Battle Creek; Rebecca McGowan, Ann Arbor; Philip H. Power, Ann Arbor; Nellie M.Varner, Detroit; James L. Waters, Muskegon; James J. Duderstadt (ex officio)

  • Preface

    This manual is intended to provide comprehensive information to those who use or areconsidering using the Michigan English Language Assessment Battery (MELAB) to makedecisions about the English language proficiency of individuals. It is written specifically forcollege and admission personnel and personnel in professional organizations who need suchinformation to assess the appropriateness and acceptability of the MELAB for particularpurposes.

    Included in the manual is general information about the MELAB (Section 1), MELAB statistics(Section 2) and reliability and validity information (Section 3). We hope that providing thisinformation to test users will help them be more knowledgeable about the MELAB and enablethem to judge the technical adequacy of the MELAB.

    The English Language Institute at the University of Michigan (ELI-UM) has a long history ofinvolvement in English as a second language testing, and the development of this manual alsohas a long history. We appreciate the advice and guidance of Liz Hamp-Lyons, Stan Jones, andPeter Skehan, external consultants who helped us map out what we needed to include in themanual. We were fortunate a few years ago to have as a research assistant Sheila Prochnow-Mathias who conducted several systematic studies of the writing component of the MELAB. Mostrecently, we benefited greatly from the expertise of Michael Persinger in handling various aspectsof the statistical analysis and putting into words the results of the factor analyses that appears inSection 3.

    This manual is also the result of a team effort of many staff members in our testing andcertification division. Mary Spaan, in particular, deserves praise for her consistent efforts toproduce tests of high technical quality, the maintenance of excellent records that were used asthe research base of this manual, offering valuable advice at all stages of the development of thismanual, and editing the manual closely at its final stage. We are grateful to Karyn Pidgeon andBob Sage who handled their routine duties of arranging and scoring MELAB tests so efficientlythat they had time to enter data essential for test analysis. Jennifer Engar also provided usefulassistance with data retrieval and entry. Theresa Rohlck used her excellent organizational andadministrative skills to compile a stratified random sample of MELAB papers that was used in thereliability studies and to oversee several large scale data entry projects and has used her capableword processing skills to transform the text and tables into a publishable format. Finally, weappreciate the strong support of our ELI-UM director John Swales who encouraged ourcompletion of this manual .

    As we have developed this manual, we have tried to follow the guidelines set forth in theStandards for Educational and Psychological Testing (1985)1 and hope that the manual offers theinformation you need. If there is information that is missing or unclear, please contact us so thatwe might address such inconsistencies in the next edition or a supplement.

    Sarah Briggs Barbara DobsonAssociate Director for Testing and Certification Research AssistantEnglish Language Institute English Language Institute

    1Standards for Educational and Psychological Testing were developed thorough the joint efforts of the AmericanEducational Research Association, American Psychological Corporation and the National Council on Measurement inEducation and were published by American Psychological Association in Washington, DC.

  • TABLE OF CONTENTS

    Preface ...................................................................................................................................... i

    SECTION 1 GENERAL INFORMATION ABOUT THE MELAB ................................... 1

    1.1 AN OVERVIEW OF THE MELAB ...................................................................................... 1

    1.2 MELAB ADMINISTRATION............................................................................................... 2

    1.3 DESCRIPTION OF THE MELAB ....................................................................................... 31.3.1 MELAB Parts and their Components1.3.2 How the MELAB relates to the former Michigan Battery

    1.4 SCORING OF THE MELAB ............................................................................................... 71.4.1 Part 1 Composition Score1.4.2 Part 2 Listening Score1.4.3 Part 3 Grammar, Cloze, Vocabulary, and Reading (GCVR) Score1.4.4 Final MELAB Score1.4.5 Speaking Test (optional)

    1.5 USING THE MELAB ........................................................................................................ 121.5.1 Interpreting Scores1.5.2 An Example of MELAB Use for a University Context1.5.2.1 Undergraduates1.5.2.2 Graduates1.5.3 An Example of MELAB Use for a Community College Context1.5.4 An Example of MELAB Use for Professional Contexts

    1.6 PREPARING FOR THE MELAB ...................................................................................... 16

    1.7 TEST SECURITY/INVALIDATIONS................................................................................. 16

    SECTION 2 MELAB STATISTICS ................................................................................... 17

    2.1 GENERAL DESCRIPTIVE STATISTICS.......................................................................... 17

    2.2 FREQUENCY DISTRIBUTION OF MELAB SCORES FOR ALL EXAMINEES ................ 18

    2.3 PERFORMANCE OF REFERENCE GROUPS ON THE MELAB ..................................... 22

    2.4 INTERCORRELATIONS AMONG MELAB SCORES AND MELAB ORAL RATING........ 24

  • SECTION 3 MELAB RELIABILITY AND VALIDITY.......................................................25

    3.1 RELIABILITY ...................................................................................................................253.1.1 Reliability of MELAB Part 1 (Composition)3.1.1.1 MELAB Composition Raters: Who they are; how they are trained3.1.1.2 Interrater Reliability3.1.1.3 Intrarater Reliability3.1.1.4 Alternate Form Reliability3.1.2 Test/Retest Reliability3.1.3 Alternate Forms Reliability (for MELAB Part 2 and MELAB Part 3)3.1.3.1 Developing Alternate Forms3.1.3.2 Distribution of Scores on Alternate Forms of MELAB Part 2 (Listening) and MELAB Part 3 (GCVR)3.1.4 Internal Consistency Reliability (KR21 and Cronbach's Alpha)

    3.2 Validity ............................................................................................................................343.2.1 Content-related Evidence3.2.1.1 Content-related Evidence for Part 1: Composition3.2.1.2 Content-related Evidence for Part 2: Listening3.2.1.3 Content-Related Evidence for Part 3: Grammar, Cloze, Vocabulary, Reading (GCVR)3.2.1.4 Content-related Evidence for Speaking Test3.2.1.5 Content-related Evidence for Final MELAB Score

    3.2.2 Construct-related Evidence3.2.2.1 Language Proficiency Theory and the MELAB3.2.2.2 Factor Analysis of the MELAB3.2.2.3 Native speaker performance on the MELAB

    3.2.3 Criterion-Related Evidence3.2.3.1 MELAB and Tests of "Productive" Language Skills3.2.3.2 MELAB and Another Proficiency Battery, the TOEFL3.2.3.3 MELAB and Teacher Assessments

    APPENDICES..........................................................................................................................63

    Appendix A MELAB Centers (By Country)Appendix B Historical Background Leading to the MELABAppendix C Sample MELAB Score Report FormAppendix D MELAB Speaking Test - Spoken English Descriptors and Salient FeaturesAppendix E Descriptive Statistics (1987-1990)Appendix F Reliability (1987-1990)

  • LIST OF ILLUSTRATIONS

    FIGURES

    Figure 3.1 Box Plots of Scaled Scores of Alternate Forms of MELAB Part 2 (Listening)......... 31Figure 3.2 Box Plots of Scaled Scores of Alternate Forms of MELAB Part 3 (GCVR)............. 32Figure 3.3 Final MELAB Scores for Seven Levels of Written and Spoken English.................. 59

    TABLES

    Table 1.1 Examples of MELAB Part 2 (Listening) Test Items.................................................. 4Table 1.2 Examples of MELAB Part 3 (GCVR) Test Items...................................................... 5Table 1.3 MELAB Part 2 (Listening) Score Converted from Raw Score ................................ 10Table 1.4 MELAB Part 3 (GCVR) Score Converted from Raw Score .................................... 10Table 1.5 Final MELAB Scores and Proficiency Levels in Speaking and Writing................... 12Table 1.6 Examples of MELAB Scores................................................................................. 13

    Table 2.1 Score Descriptives for 4,811 First-Time MELABs Administered 1991-1993 ........... 17Table 2.2 MELAB Scaled Scores Corresponding to Specified Percentiles ............................ 18Table 2.3 Frequency Distribution of Final MELAB Scores..................................................... 19Table 2.4 Frequency Distribution of MELAB Part 1 (Composition) Scores............................. 19Table 2.5 Frequency Distribution of MELAB Part 2 (Listening) Scaled Scores ...................... 20Table 2.6 Frequency Distribution of MELAB Part 3 (GCVR) Scaled Scores .......................... 21Table 2.7 MELAB Scaled Score Mean and Standard Deviation by Examinee Reason for Testing ............................................................................................... 22Table 2.8 MELAB Scaled Score Mean and Standard Deviation by Sex ................................ 22Table 2.9 MELAB Scaled Score Mean and Standard Deviation by Age ................................ 22Table 2.10 Mean and Standard Deviation by Native Language for Parts 1, 2, 3, and Final MELAB Scores ............................................................................................ 23Table 2.11 Intercorrelations of Scaled MELAB Part Scores and Final Scores and MELAB Oral Rating, 1991-1993............................................................................ 24

    Table 3.1 Summary of MELAB Reliability Estimates............................................................. 25Table 3.2 MELAB Part 1 Interrater Reliability ....................................................................... 26Table 3.3 Rater Score Differences for MELAB Part 1 ........................................................... 27Table 3.4 MELAB Test/Retest Results ................................................................................. 28Table 3.5 Correlations Among Alternate Forms of MELAB ................................................... 30Table 3.6 Percentile Rankings of Scaled Scores of Alternate Forms of MELAB Part 2 (Listening).................................................................................................. 30Table 3.7 Percentile Rankings of Scaled Scores of Alternate Forms of MELAB Part 3 (GCVR)...................................................................................................... 31Table 3.8 Kuder-Richardson Formula 21 Internal Consistency Reliability Estimates for MELAB Part 2 (Listening) ..................................................................................... 33Table 3.9 Kuder-Richardson Formula 21 Internal Consistency Reliability Estimates for MELAB Part 3 (GCVR) ......................................................................................... 33Table 3.10 Cronbach's Alpha Reliability Estimates for MELAB Part 2 (Listening).................... 34Table 3.11 Cronbach's Alpha Reliability Estimates for MELAB Part 3 (GCVR)....................... 34Table 3.12 MELAB Part 2 (Listening) Items: Type and Number by Form ................................ 39Table 3.13 MELAB Part 3 (GCVR) Grammar Items: Type and Number by Form.................... 42Table 3.14 MELAB Part 3 Reading Passage Readability Statistics ......................................... 43Table 3.15 MELAB Part 3 Reading Passages (type and length) and Item Difficulty................. 44Table 3.16 MELAB Part 3 (GCVR) Item Difficulty by Sub-section ........................................... 45Table 3.17 MELAB Components and Bachman/Palmer Model of Language Knowledge......... 47

  • Table 3.18 Part 2 (Listening) Form BB Component Means, Standard Deviations, and Correlation Matrix .................................................................................................49Table 3.19 Part 2 (Listening) Form BB Factor Loadings Single Factor Solution.......................49Table 3.20 Part 2 (Listening) Form BB Reproduced Correlation Matrix Single Factor Solution ................................................................................................................49Table 3.21 Part 2 (Listening) Form CC Component Means, Standard Deviations, and Correlation Matrix .................................................................................................50Table 3.22 Part 2 (Listening) Form CC Factor Loadings Single Factor Solution.......................50Table 3.23 Part 2 (Listening) Form CC Reproduced Correlation Matrix Single Factor Solution ................................................................................................................50Table 3.24 Part 3 (GCVR) Form AA Component Means, Standard Deviations, and Correlation Matrix .................................................................................................51Table 3.25 Part 3 (GCVR) Form AA Factor Loadings Single Factor Solution...........................51Table 3.26 Part 3 (GCVR) Form AA Reproduced Correlation Matrix Single Factor Solution ....51Table 3.27 Part 3 (GCVR) Form BB Component Means, Standard Deviations, and Correlation Matrix .................................................................................................51Table 3.28 Part 3 (GCVR) Form BB Factor Loadings Single Factor Solution...........................52Table 3.29 Part 3 (GCVR) Form BB Reproduced Correlation Matrix Single Factor Solution ....52Table 3.30 Part 3 (GCVR) Form CC Component Means, Standard Deviations, and Correlation Matrix .................................................................................................52Table 3.31 Part 3 (GCVR) Form CC Factor Loadings Single Factor Solution ..........................52Table 3.32 Part 3 (GCVR) Form CC Reproduced Correlation Matrix Single Factor Solution ....53Table 3.33 Listening Form BB, GCVR (Forms AA, BB, CC), and Composition Component Means, Standard Deviations, and Correlation Matrix .............................................53Table 3.34A Listening Form BB, GCVR (Forms AA, BB, CC), and Composition Factor Pattern Loadings Two Factor Solution........................................................54Table 3.34B Listening Form BB, GCVR (Forms AA, BB, CC), and Composition Factor Structure Loadings Two Factor Solution .....................................................54Table 3.35 Listening Form BB, GCVR (Forms AA, BB, CC), and Composition Reproduced Correlation Matrix Two Factor Solution..................................................................54Table 3.36 Listening Form CC, GCVR (Forms AA, BB, CC), and Composition Component Means, Standard Deviations, and Correlation Matrix .............................................55Table 3.37A Listening Form CC, GCVR (Forms AA, BB, CC), and Composition Factor Pattern Loadings Two Factor Solution........................................................55Table 3.37B Listening Form CC, GCVR (Forms AA, BB, CC), and Composition Factor Structure Loadings Two Factor Solution .....................................................56Table 3.38 Listening Form CC, GCVR (Forms AA, BB, CC), and Composition Reproduced Correlation Matrix Two Factor Solution..................................................................56Table 3.39 MELAB Scores for Those Claiming English as Their Native Language and MELAB Total Group Scores ..................................................................................57Table 3.40 Brief Proficiency Descriptions for MELAB Writing and Speaking Ratings ...............58Table 3.41 MELAB/TOEFL Descriptive Statistics....................................................................60Table 3.42 Descriptive Statistics for Teacher Assessment Validity Study ................................61Table 3.43 Relationship Between MELAB Scores and Teacher Ranking of Students ..............62

  • 1SECTION 1: GENERAL INFORMATION ABOUT THE MELAB

    1.1 AN OVERVIEW OF THE MELAB

    The Michigan English Language Assessment Battery (MELAB) is an examination whose purposeis to evaluate the advanced level English language competence of adult non-native speakers ofEnglish. The MELAB assesses both spoken and written English:

    Part 1 is an impromptu composition, written on an assigned topic;Part 2 is a listening test, delivered via tape recording;Part 3 is a written test containing grammar, cloze reading1, vocabulary, and reading comprehension problems.A speaking test is optional. The local examiner provides an oral rating based on an oral interview.

    The MELAB was developed to assess the English language proficiency of students who areapplying to U.S. and Canadian universities, colleges, or community colleges where the languageof instruction is English. The MELAB is also used to assess the general English languageproficiency of professionals such as medical personnel, engineers, managers, and governmentofficials who will need to use English in their work or in on-site training in the U.S. Otherindividuals who take the MELAB are non-native speakers interested in obtaining a generalestimate of their English language proficiency to help them make decisions about applying foreducational or employment opportunities.

    Many educational institutions in the U.S. and Canada accept the MELAB as an alternative to theTOEFL2 as evidence of English language proficiency. International organizations such as theWorld Health Organization and the International Monetary Fund use the MELAB when they needevidence of English language proficiency of fellowship and scholarship candidates. Stateprofessional boards such as Boards of Nursing use MELAB scores as an indicator of Englishproficiency when non-native speakers of English apply for certification exams.

    The MELAB is a secure test battery. The test forms included in the battery are not commerciallyavailable.3 The MELAB is administered only by the English Language Institute--University ofMichigan (ELI-UM) and official examiners in the U.S. and around the world who are authorized bythe ELI-UM. A permanent team of testing professionals at the ELI-UM develops the MELAB. Apermanent staff in Ann Arbor oversees all registration for the battery, scores all test papers, andissues all official score reports.

    MELAB score reports include scaled scores for the different parts of the test battery as well as aFinal MELAB score, which is the mean (average) of the scores on Parts 1, 2, and 3. Scores onthe optional speaking test are not averaged with the other part scores. Brief biographicalinformation, the test date, and the test location also appear on score reports.

    Examinees receive one copy of their MELAB scores and may request that the ELI-UM sendofficial score reports directly to particular colleges, universities or professional organizations. Allofficial score reports are embossed and sent out by the ELI-UM. Admissions officers arecautioned not to accept copies of score reports directly from students. Scores are consideredcurrent and valid for two years. No score reports are issued for tests taken more than two yearsin the past.

    1Cloze reading refers to a test method requiring the examinee to identify the words that have been deleted from a text.2Test of English as a Foreign Language (TOEFL). Princeton, NJ: Educational Testing Service.3 Other English language tests developed at ELI-UM are available to educators. Retired forms of MELAB Part 2 and 3are made available as the MELICET (Michigan English Language Institute College English Test) through ELI TestPublications.

  • 21.2 MELAB ADMINISTRATION

    The ELI-UM oversees the administration of all MELABs. MELABs are administered at the ELI-UM in Ann Arbor, Michigan, and by some 300 approved MELAB examiners around the worldfollowing uniform test administration procedures.4 MELABs are administered biweekly to groupsof examinees at the ELI-UM in Ann Arbor and regularly at certain approved group test centers;however, generally MELABs around the U.S. and the world are arranged and administered on anindividual basis with scheduling to accommodate the needs and constraints of the examiner andthe examinee.

    In Ann Arbor, Michigan, MELABs are administered by trained staff of the ELI testing division.Official MELAB examiners who administer the test elsewhere are generally educators who haveapplied to serve as MELAB examiners and have met the following selection criteria:

    permanent affiliation with an educational institution;native or near native proficiency in English;some knowledge of standardized testing and testing procedures;professional background in evaluation, educational measurement, guidance and counseling,

    admissions, or ESL/EFL;personal qualities necessary in a person responsible for the administration of a secure

    examination to individuals and to large groups of examinees.

    Test administration instructions are provided to all MELAB examiners, and examiners regularlyreceive updating on policy or administration changes. Examiners are monitored for compliancewith established MELAB administration procedures. Many MELAB examiners have extensiveexperience in the administration of English language proficiency tests.

    An individual who wants to take the MELAB completes the registration form printed in the MELABInformation Bulletin5 and sends it, along with appropriate test fees, to the ELI-UM. Upon receiptof the form and fees, the ELI-UM staff checks test records to determine whether the individualmeets MELAB eligibility requirements. To be eligible for the MELAB, a person must not havetaken the test more than 3 times in a 12 month period and must wait at least 6 weeks betweenMELABs. Eligible persons who have taken the MELAB before are assigned different forms of thetest than the forms they took earlier.

    Once the staff verifies an individual's eligibility, the ELI-UM sends that person the name andaddress of a local MELAB examiner and an official identification form to present to the examiner.Simultaneously, the examiner is sent the prospective examinee's name and appropriate testingmaterials. It is the examinee's responsibility to contact the examiner to arrange a test date withinsix months of issuance of the identification form.

    Some individuals take the MELAB at the request of a sponsoring agency (such as the WorldHealth Organization). These people are registered by their sponsoring agency, which also paystheir test fees. After verifying their eligibility, the ELI-UM sends sponsored candidates the officialidentification form and the name and address of whom to contact to take the examination.

    Group Testing Centers register groups of examinees and send rosters of these potential MELABexaminees to the ELI-UM, which verifies that these examinees are eligible for testing. EachGroup Testing Center sets its own test date(s).

    Whether taking the MELAB individually or at a group center, every examinee must present properidentification to the examiner on the day of the test. This identification is:

    4 Appendix A provides a listing of MELAB examination centers.5The MELAB Information Bulletin and registration forms are available from the English Language Institute, Testing andCertification Division, free of charge.

  • 3a properly-completed MELAB identification form with two recent photographs of theexaminee attached; and

    two forms of photo identification (including signed passport, alien registration card,or national identity card).

    The identification forms and photographs are sent to the ELI-UM with the examinee's test paperswhere they are kept on file for two years.

    All completed test papers are sent to the ELI-UM for scoring. Score reports are sent by U.S. Mailto examinees and schools seven to fourteen days after test papers are received in Ann Arbor.For examinees who need score reports faster, there is a rush service available. For an additionalfee, test results are sent by courier directly to an admissions office within 48 hours of the time theELI-UM receives the test papers.

    Unless examinees order rush service, the total "turnaround" time, from the time test registrationforms are sent to the ELI-UM until the time score reports are received, is six to eight weeks. Thetime may be shorter for people tested in the U.S. or registered by a sponsor.

    1.3 DESCRIPTION OF THE MELAB

    1.3.1 MELAB Parts and their Components6

    Part 1: Composition. The first part of the battery is a 30-minute writing task. Examinees areinstructed to write an essay on an assigned topic, or prompt. Two prompts are given, andexaminees select the one on which they prefer to write. Examinees are expected to writebetween 200 and 300 words on the topic they choose. Examinees are instructed to ask theMELAB examiner to explain or translate the topics if they do not understand them and areadvised to make a short outline if they wish. They are instructed that extremely short essaysreceive a lower rating and that their handwriting should be readable. They are instructed tochange or correct parts of their essay as they wish but that they should not copy the wholecomposition over. They are informed that their composition is judged on clarity and overalleffectiveness, topic development, organization, and range, accuracy and appropriateness ofgrammar and vocabulary. Examinees are informed that compositions written on topics other thanthe assigned topic are not assigned a rating. The topics typically require the examinee to take aposition on an issue and defend it, to describe something from personal experience, or to explaina problem and offer possible solutions. Some sample composition topics are:

    1. What are the characteristics of a good teacher? Explain and give examples.2. How should students be evaluated: according to their achievements or their effort? Discuss.3. What do you think is your country's greatest problem? Explain in detail and tell what you think can be done about it.4. Would you prefer to live in the city or in the country? Explain the reasons for your choice.

    Part 2: Listening. In this part of the battery, examinees hear 50 test items delivered via an audiotape recording lasting about 25 minutes. Examinees are informed that the purpose of Part 2 is toassess how well they understand spoken English. Test instructions are presented via audio tapeand in writing, and examinees are given the opportunity to ask questions about test procedureprior to beginning the listening test.

    The spoken discourse in the listening test includes short questions and statements and longerdiscourse segments. Questions, statements, short conversations, a mini-lecture on a topic ofgeneral interest followed by questions and a longer conversation followed by questions are

    6For more information about the content of Part 1, see Section 3.2.1.1; of Part 2, see Section 3.2.1.2; of Part 3, seeSection 3.2.1.3; of the Speaking Test, see Section 3.2.1.4

  • 4Table 1.1 Examples of MELAB Part 2 (Listening) Test Items

    Type of item Instructions Aural Cue Answer choicesShort question Select the best

    answer to thequestion.

    Have you been tosee the new movieyet?

    a. Yes, I'm going tomorrow.b. No, it wasn't very good.c. Yes, I went yesterday.

    Short statement Select the answerthat means aboutthe same as whatyou hear.

    Frank never would'vegone to the lecture ifhe'd known howboring it was going tobe.

    a. He didn't want to go.b. He didn't like it.c. He never went.

    Shortconversationalexchange

    Select the answerthat means aboutthe same as whatyou hear.

    M: Let's go to thefootball game.F: Good idea. I don'twant to stay home.

    a. They will go to a game.b. They will stay home.c. They don't like football.

    Statement withemphasizedsegment

    Select what thespeaker will saynext.

    Tom said he wasgoing to drive toChicago nextweek . . .

    a. not last week.b. not next month.c. not fly.

    Question withemphasizedsegment

    Select the bestresponse to thequestion.

    Do you have John'skeys?

    a. No, but Jane does.b. No, I have Jim's.c. No, only his bags.

    Questions aboutcontent of amini-lecture

    Listen to a shortlecture. Takenotes about whatyou hear. Lookat a graph orchart in the testbooklet. Answerquestions aboutthe lecturereferring to notes,chart, and graph.

    Mini-lecture followedby several questions.

    3 printed answer choices foreach question

    Questions aboutcontent of aconversation

    Listen to aconversation,look at a map ordiagramassociated withthe conversation,take notes whilelistening. Answerquestionsreferring to themap or diagramand notes.

    Conversationaldiscourse lasting 4-5minutes followed byseveral questions.

    3 printed answer choices foreach question

  • 5Table 1.2 Examples of MELAB Part 3 (GCVR) Test Items

    Type of item Instructions Problem stem Answer choicesGrammar Choose the word

    or phrase thatbest completesthe conversation.

    "The boys say theywere treatedunfairly.""They got the sametreatment _____everyone else."

    a. thanb. thatc. so asd. as

    Cloze reading Select answersthat areappropriate inboth grammarand meaning.

    A single readingpassage of about250 words with 20words missing--approximately every7th word appearingas a blank in the text.

    Four answer choices arepresented for each missingword

    Vocabularysynonym

    Choose the wordor phrase thatmeans about thesame thing as theunderlined wordor phrase.

    Bill Collins launchedhis restaurant lastJune.

    a. movedb. startedc. soldd. bought

    Vocabularycompletion

    Choose the wordor phrase thatbest fits thecontext.

    I disagree with a fewof his opinions, but_____ we agree.

    a. deliberatelyb. conclusivelyc. essentiallyd. immensely

    Reading Read thepassage andanswer thequestionsfollowing itaccording toinformation givenin the passage.

    A reading selectionof 150-300 wordsfollowed by,generally, fivequestions.

    4 answer choices for eachreading comprehensionquestion

    delivered only once via audio recording. The recording is in standard American English, withmale and female speakers speaking at a normal rate. The mini-lectures are delivered at anaverage rate of about 150 wpm. A pause of 12-15 seconds follows each aurally presented testitem. Examinees are instructed to select answers from three response options presented inmultiple-choice format in a test booklet and to mark their answers on a separate answer sheet.Examinees are advised to take notes as they listen to the lecture and longer conversation. Thenotes may be used when answering questions.

    Part 3: Grammar, Cloze, Vocabulary, Reading (GCVR). The third part of the battery is a 100-item grammar, cloze, vocabulary, and reading comprehension test. Examinees are allotted onehour and 15 minutes (75 minutes) for this part of the MELAB. There are 30 grammar items, 20cloze items, 30 vocabulary items, and 20 reading comprehension questions. Examinees selectresponses from four multiple-choice options. The grammar and vocabulary items are discrete-type items; the cloze items are within the context of a single passage of written text; and thereading items are based on four different passages of text on different topics.

    Speaking Test. The speaking test is an optional part of the MELAB. Interviews lasting 10 to 15minutes are conducted individually by the examiner administering the MELAB. ELI-UM providesinterviewing guidelines to the examiner that suggest a 3-part framework for the interview: anopening warm-up phase, a main part to elicit extended discourse from the examinee, and aclosing phase. The interview might include questions about the examinee's background, future

  • 6plans, and opinions on certain issues. It might elicit discourse about the examinee's field ofspecialization.

    The examiner rates the examinee's general command of spoken English. The examinerconsiders fluency, intelligibility, grammar, vocabulary, comprehension, and functional languageuse.

    1.3.2 How the MELAB relates to the former Michigan Battery

    The ELI-UM has had a language proficiency testing program for many years.7 Through the yearsnew components of the proficiency battery have been developed. In 1985, the old "MichiganBattery" was replaced with the present Michigan English Language Assessment Battery(MELAB).

    MELAB Part 1, the written composition, is essentially the same type of writing task as was in theMichigan Battery, but the essay topics are continually up-dated. The level rating descriptionshave undergone complete revision, and a coding system has been established to providefeedback to examinees about salient features of their written composition.

    MELAB Part 2, the listening test, replaced the Listening Comprehension Test (LCT)8. Comparedto the LCT, MELAB Part 2 was designed to place less emphasis on "aural grammar" andexpanded emphasis on the comprehension of naturally spoken English. Part 2 now includesitems that require understanding the prosodics of spoken English, as well as two segments oflonger discourse, in the form of a lecture and a conversation. In the piloting of the first forms ofMELAB Part 2, examinees were given both an LCT and a MELAB Part 2. Forms BB and CC ofMELAB Part 2 correlated .65 and .67 respectively with LCT.9 Correlation coefficients of the newitem types (emphasis comprehension items and items based on the comprehension of extendeddiscourse) with the LCT items ("aural grammar" items) range from .42 to .52. These coefficientsindicate that there is only a moderate relationship among the skills tested by the new and the olditem types. Typically scores on MELAB Part 2 are significantly lower than on the LCT (by anaverage of 10 points). Therefore, it is crucial that users of MELAB scores do not considerMELAB scores to be equivalent to scores on the old Michigan Battery. Any proficiency testcut scores established prior to 1985 (that is, established for the old Michigan Battery rather thanthe MELAB) must be re-examined.

    MELAB Part 3 replaced the Michigan Test of English Language Proficiency10 (MTELP). Thenumber of grammar and vocabulary items was reduced, and a cloze reading passage (a longerdiscourse segment with items focusing on both grammatical appropriateness and textcomprehension) was added. Correlation of various forms of MELAB Part 3 (Forms AA, BB, CC)with the MTELP (Forms P, Q, R) yield coefficients ranging from .88 to .91 which suggests thatPart 3 of the MELAB may be more similar to its predecessor component, the MTELP, than Part 2of the MELAB is to its predecessor component, the LCT.

    The speaking test, an oral interview, remains an optional component of the battery. The MELABSpoken English Reference Sheet, though, reflects an extensive revision of the level descriptorsused in oral interviews given with the old Michigan Battery.

    7 Information about tests in former Michigan proficiency batteries is available in Appendix B: Historical BackgroundLeading to the MELAB.8Listening Comprehension Test (LCT). (1972). Ann Arbor, MI: English Language Institute, The University of Michigan.9 MELAB piloting is conducted on non-native speakers of English who contact the English Language Institute becausethey need an assessment of their English language proficiency. Pilot MELAB listening tests, Forms AA, BB, and CC wereadministered under normal test conditions with subjects representative of the range of language backgrounds andproficiency levels typical of MELAB candidates; Form AA N=106, Form BB N=215, Form CC N=178. Pilot testingoccurred during 1983-85. In 1986, Form AA, a 40-item listening test, was dropped as it was decided to use a 50-itemoperational format.10Michigan Test of English Language Proficiency (MTELP). (1968, 1971, 1979). Ann Arbor, MI: English LanguageInstitute, The University of Michigan.

  • 71.4 SCORING OF THE MELAB

    MELAB scores are reported on an official score report form. A score for each of the parts isreported, as well as the Final MELAB Score, which is the average of the scores for Part 1, Part 2,and Part 3. Examinees who have had the speaking test have a oral rating reported. The oralrating is not averaged into the Final MELAB Score. Other information about test performancemay be reported in the verbal comments section of the score report. A sample score report formis shown in Appendix C.

    Scores on the various MELAB parts are reported on a numerical scale:

    Section Range NotePart 1 (Composition) 53 - 97 May be supplemented by codesPart 2 (Listening) 30 - 100Part 3 (GCVR) 15 - 100Final MELAB Score 33 - 99 Average of Parts 1, 2 and 3Speaking 1 - 4+ Optional. May be supplemented by interviewer comments

    1.4.1 Part 1 Composition Score

    MELAB compositions are rated on how clearly and effectively ideas are communicated in writtenEnglish. Each MELAB composition score is a numerical score (ranging from 53 to 97) that maybe followed by one or more letter codes. The numerical score represents the general level ofwriting proficiency evident in the composition. Letter codes represent features of the writing thatthe raters found especially strong or weak in relation to the overall level of the writing.

    Compositions are rated or scored by a small group of trained, experienced raters at the ELI-UMwho assess compositions daily. Each composition is scored by at least two raters who readindependently and independent of knowledge about how the examinee performed on other partsof the MELAB. Compositions are typically read in small batches, and their order is altered by thesecond rater to minimize any effect that the position of a composition within a set of compositionsmight have on the score it is given.

    A rater assigns one of ten numerical scores to a composition (53, 57, 63, 67, 73, 77, 83, 87, 93,or 97). If the scores the first two raters assign to a single composition are identical or only onescale point apart, the composition is assigned the average of these two scores. Essays for whichthe two initial ratings differ by more than one scale point are scored by a third rater. When theinitial ratings are two scale points apart and the third rating falls between them, the middle scoreis used. In all other cases, the composition is given the average of the two scores that areclosest to (or equal to) each other. A third rater is also used in cases where there is a largediscrepancy between an examinee's composition score and scores on other parts of the MELAB.Approximately 8% of MELAB compositions are read by a third rater.

    Description of MELAB scores by level:

    97 Topic is richly and fully developed. Flexible use of a wide range of syntactic (sentence level)structures, accurate morphological (word forms) control. Organization is appropriate andeffective, and there is excellent control of connection. There is a wide range of appropriatelyused vocabulary. Spelling and punctuation appear error free.

    93 Topic is fully and complexly developed. Flexible use of a wide range of syntactic structures.Morphological control is nearly always accurate. Organization is well controlled and appropriateto the material, and the writing is well connected. Vocabulary is broad and appropriately used.Spelling and punctuation errors are not distracting.

    87 Topic is well developed, with acknowledgment of its complexity. Varied syntactic structuresare used with some flexibility, and there is good morphological control. Organization is controlledand generally appropriate to the material, and there are few problems with connection.

  • 8Vocabulary is broad and usually used appropriately. Spelling and punctuation errors are notdistracting.

    83 Topic is generally clearly and completely developed, with at least some acknowledgment ofits complexity. Both simple and complex syntactic structures are generally adequately used;there is adequate morphological control. Organization is controlled and shows some appropriacyto the material, and connection is usually adequate. Vocabulary use shows some flexibility, andis usually appropriate. Spelling and punctuation errors are sometimes distracting.

    77 Topic is developed clearly but not completely and without acknowledging its complexity.Both simple and complex syntactic structures are present; in some "77" essays these arecautiously and accurately used while in others there is more fluency and less accuracy.Morphological control is inconsistent. Organization is generally controlled, while connection issometimes absent or unsuccessful. Vocabulary is adequate, but may sometimes beinappropriately used. Spelling and punctuation errors are sometimes distracting.

    73 Topic development is present, although limited by incompleteness, lack of clarity, or lack offocus. The topic may be treated as though it has only one dimension, or only one point of view ispossible. In some "73" essays both simple and complex syntactic structures are present, butwith many errors; others have accurate syntax but are very restricted in the range of languageattempted. Morphological control is inconsistent. Organization is partially controlled, whileconnection is often absent or unsuccessful. Vocabulary is sometimes inadequate, andsometimes inappropriately used. Spelling and punctuation errors are sometimes distracting.

    67 Topic development is present but restricted, and often incomplete or unclear. Simplesyntactic structures dominate, with many errors; complex syntactic structures, if present, are notcontrolled. Lacks morphological control. Organization, when apparent, is poorly controlled, andlittle or no connection is apparent. Narrow and simple vocabulary usually approximates meaningbut is often inappropriately used. Spelling and punctuation errors are often distracting.

    63 Contains little sign of topic development. Simple syntactic structures are present, but withmany errors; lacks morphological control. There is little or no organization, and no connectionapparent. Narrow and simple vocabulary inhibits communication, and spelling and punctuationerrors often cause serious interference.

    57 Often extremely short; contains only fragmentary communication about the topic. There islittle syntactic or morphological control, and no organization or connection are apparent.Vocabulary is highly restricted and inaccurately used. Spelling is often indecipherable andpunctuation is missing or appears random.

    53 Extremely short, usually about 40 words or less; communicates nothing, and is often copieddirectly from the prompt. There is little sign of syntactic or morphological control, and no apparentorganization or connection. Vocabulary is extremely restricted and repetitively used. Spelling isoften indecipherable and punctuation is missing or appears random.

    N.O.T. N.O.T. (Not On Topic) indicates a composition written on a topic completely differentfrom any of those assigned; it does not indicate that a writer has merely digressed from ormisinterpreted a topic. N.O.T. compositions often appear prepared and memorized. They arenot assigned scores or codes.

    Since September, 1989, in addition to giving a numerical score to a composition, raters have hadthe option of assigning letter codes. Any code assigned to the same composition by two or moreraters is reported along with the numerical score to provide additional interpretive information toboth examinees and institutional score users. A letter code means that one feature of the writingwas especially strong or weak for the particular score level, though not strong or weak enough toraise or lower the overall score. For example, the vocabulary used could be especially broad fora score level of 77, but not strong enough to raise the overall score to 83. Codes, like numberscores, are assigned independently by 2 to 3 trained raters.

    Code letters do not raise or lower number scores. The codes do not replace the number score;they add detail to it. There are 20 codes. Each describes one feature of writing. None of thesefeatures works alone, but each one can affect writing quality. A list of the code letters and their

  • 9meanings is given below. The alphabetical order of the codes has no positive or negativemeaning. For example, 'a' does not mean excellent, and 'f' does not mean poor or failure.Letters are used simply as a convenient means for reporting score information.

    Key to composition score codes:

    NOTE: the codes are meant to indicate that a certain feature is ESPECIALLY GOOD OR BAD INCOMPARISON TO THE OVERALL LEVEL OF THE WRITING

    a topic especially poorly or incompletely developedb topic especially well developed

    d organization especially uncontrollede organization especially well controlled

    f connection especially poorg connection especially smooth

    h syntactic (sentence level) structures especially simplei syntactic structures especially complexj syntactic structures especially uncontrolled

    l especially poor morphological (word forms) controlm especially good morphological control

    n vocabulary especially narrowo vocabulary especially broadp vocabulary use especially inappropriate

    r spelling especially inaccurates punctuation especially inaccuratet paragraph divisions missing or apparently random

    v question misinterpreted or not addressedw reduced one score level for unusual shortnessx other (write-in: see score report)

  • 10

    1.4.2 Part 2 Listening Score

    The listening score reported is a converted or scaled score. The raw score, the total number oftest items answered correctly, is converted to a scaled score. There are 50 test items on thelistening test. The scale range in the conversion scale for Part 2 of the MELAB is 30-100. Achance, or guessing, score is about 45 (scaled score). The average score is about 75.

    The specific conversion scale for each form varies from one form to the other. Conversion scalesfor various MELAB forms are based on normative information (primarily percentile rank) onalternate forms of the tests.11 Converted scores are not percentage scores nor the exact numberof problems answered correctly. The following table is extrapolated from the conversion table offour forms of MELAB Part 2 and provides a rough indication of the relationship between the rawand scaled scores of MELAB Part 2.

    Table 1.3 MELAB Part 2 (Listening) Score Converted from Raw Score

    Raw Score(number correct)

    Raw Score(% correct)

    Part 2: MELABScaled Score

    47 - 48 94 - 96 9542 - 43 84 - 86 9037 - 40 74 - 80 8530 - 35 60 - 70 8025 - 29 50 - 58 7523 - 25 46 - 50 7020 - 21 40 - 42 65

    18 33 45

    1.4.3 Part 3 Grammar, Cloze, Vocabulary, and Reading (GCVR) Score

    The MELAB Part 3 (GCVR) score, like the Part 2 (Listening) score, is a scaled score. There are100 test items on Part 3. The total number answered correctly, the raw score, is converted to ascaled score. The score range in the conversion scale for MELAB Part 3 is 15-100. A chance, orguessing score is about 40 (scaled score). The mean score is about 75. The specific conversionscale for each form varies from one form to the other. The following table is a rough indication ofthe relationship between raw and scaled scores of MELAB Part 3.

    Table 1.4 MELAB Part 3 (GCVR) Score Converted from Raw Score

    Raw Score(number correct)

    Raw Score(% correct)

    Part 3: MELABScaled Score

    89 - 93 89 - 93 9578 - 83 78 - 83 9066 - 75 66 - 75 8559 - 66 59 - 66 8052 - 57 52 - 57 7544 - 52 44 - 52 7038 - 46 38 - 46 65

    25 25 40

    Note that the scaled score that is reported for MELAB Part 2 and Part 3 is neither a percentagescore nor the exact number of problems answered correctly. The information in Tables 1.3 and

    11The initial normative group used to establish conversion scales for a new test form is composed of approximately 100MELAB examinees who each take two alternate forms. Various factors such as linguistic background and proficiencylevel are considered in selection of subjects for a test norm group. Conversion scales are re-analyzed later with a largersubject pool, and slight adjustments may be made in the scale.

  • 11

    1.4 is provided only to increase understanding of the relationship between raw scores andMELAB reported scaled scores. Raw scores are not reported on the score report form.

    Generally both Part 2 and Part 3 of the MELAB are at present hand-scored with a scoring stencil.If an examinee questions the accuracy of the scoring, the answer sheets are re-scored. If anydiscrepancies are found, all MELAB reports are immediately corrected, and revised reports aredistributed. There is no fee for this service. Continued monitoring of MELAB scoring procedureshas revealed a consistently high level of accuracy in the scoring and reporting of MELAB scores.

    1.4.4 Final MELAB Score

    The Final MELAB score is the average of the scores of Parts 1, 2, and 3 (e.g. 73 + 87 + 76 = 236divided by 3 = 79).

    Occasionally, instead of a numerical score, the letters "NFS" appear on a score report form."NFS" stands for "No Final Score." "NFS" appears if the English Language Institute TestingDivision can not report a final score because (1) the examinee left the room and did not completethe examination; or (2) the examinee received no Part 1 score, e.g. wrote off topic so that "NOT"(see Section 1.4.1) was assigned in place of a Part 1 score; or (3) some sort of cheating or testcompromise occurred prior to or during the MELAB test administration.

    If the examinee takes the speaking test, the oral rating is never averaged in with the Final MELABscore. It is always reported separately on the score report form.

    1.4.5 Speaking Test (optional)

    The speaking test is available to provide a rating of the examinee's spoken English. The scoreon this part is not averaged in with the three other parts of the MELAB. The MELAB oral scoreprovides information about face-to-face communicative use of English. The oral rating is arrivedat independently of the other MELAB scores and thus provides information that may complementor confirm interpretation of results on other parts of the MELAB.

    MELAB oral scores range from 1 to 4, with 4 being the highest. An examiner who thinks theexaminee's spoken English is between levels may add a plus (+) or minus (-) to the score, forexample, 3+, or 2-. The average oral rating is 2+. In addition to an overall rating, the examinermay comment on:

    features of the examinee's spoken Englishfluency/intelligibilitygrammar/vocabularyfunctional language use/sociolinguistic proficiencylistening comprehension

    The examiner's observations of salient features of the examinee's spoken English are reported onthe score report form. A reference sheet provides examiners with descriptions of each scorelevel and of salient features of spoken English. A copy of this reference sheet appears inAppendix D.

  • 12

    1.5 USING THE MELAB

    1.5.1 Interpreting Scores

    The MELAB is developed to evaluate the English language proficiency of NNSs interested inpursuing academic studies in English at the college and university level. The component itemsand parts are trialed on NNSs of English, primarily in their late teens and twenties, both withinand outside the U.S.

    MELAB scores provide information on the general language proficiency of the examinee at thetime the individual took the test battery. Examinees with higher Final MELAB scores tend to bejudged more proficient in English (as measured by "productive" tests--a written composition and aspeaking test) than those with lower Final MELAB scores. Table 1.5 shows speaking and writingproficiency level information of examinees who scored in various Final MELAB score ranges.This information suggests that examinees with higher Final MELAB scores have higher speakingand writing proficiency than those with lower Final MELAB scores. However, there is someoverlap in the level of the productive skills of examinees in different Final MELAB score ranges.For example, some individuals with Final MELAB scores in the high 70's have speaking andwriting profiles similar to some individuals with Final MELAB scores in the low 80's. Thissuggests that some variability in speaking and writing proficiency is to be expected with referenceto specific Final MELAB scores and that it is inappropriate to make rigid interpretations of whatscores mean with regard to language proficiency.

    Table 1.5 Final MELAB Scores and Proficiency Levels in Speaking and Writing1

    Final MELABScore Level

    Typical RatingSpeaking2

    ProficiencySpeaking3

    Typical RatingWriting2

    ProficiencyWriting4

    Below 60 1+ -- 3 limited to capable 60 -- 67 limited60-69 2 -- 3 modest to capable 65 -- 73 limited to basic70-79 2+ -- 3 modest to capable 70 -- 77 basic80-89 3 -- 3+ capable 77 -- 83 basic to good90+ 3+ -- 4 capable to very good 87 -- 93 good to very good

    1 The subjects were 1705 MELAB examinees of varied linguistic backgrounds who took a MELAB exam with aspeaking test between 1987 and 1990; about 30% of the exams were conducted in the U.S. or Canada and 70%elsewhere around the world. The speaking test they took was the MELAB Speaking Test; the writing test was MELABPart 1. It should be noted that the Final MELAB scores are, therefore, not completely independent of the writing scoresused in this study.2 Typical score: scores of those in the 25th percentile to 75th percentile (the interquartile range)3 Proficiency Speaking: very good/good = speaking at 4 or 4+ level capable = speaking at 3 or 3+ level marginal/modest = speaking at 2 or 2+ level limited = speaking at 1 or 1+ level4 Proficiency Writing: very good = writing at 93 or 97 level good = writing at 83 or 87 level basic = writing at 73 or 77 level limited = writing at 57, 63 or 67 level

    It is important to remember that MELAB scores are only estimates of examinees' true proficiency.MELAB scores, like all test scores, are affected by measurement error. A MELAB score may beinfluenced by factors unrelated to an examinee's language proficiency. Such factors mightinclude temporary characteristics unique to the individual (e.g. fatigue, anxiety, illness). Individualpersonal characteristics with regard to background and personality as well as aspects of testmethod can also affect test performance. Consequently, it is always appropriate to be cautiousabout interpreting scores on language tests.

    A useful statistic to consider when interpreting test scores is the standard error of measurement,or SEM. The SEM estimates the extent to which "observed" test scores (here, MELAB scores)

  • 13

    and "true" test scores (hypothetical scores free of all measurement error) differ. The SEM of theMELAB is estimated to be approximately 3 points. As an example of how to consider the SEMwhen interpreting MELAB scores, consider the case of two examinees, Examinee A, who scores81 on the MELAB, and Examinee B, who scores 79 (see Table 1.6). It is not possible to saywhether these "observed" scores are the same as the examinees' "true" scores. What can beinferred, though, considering the SEM, is that 68 out of 100 times, Examinee A's "true" score willfall between 78 and 84; and Examinee B's will fall between 76 and 82. Because these scoreranges overlap, we cannot be sure that Examinee A is more proficient than Examinee B. To saythat there is a statistically significant (at the .05 level) difference in the proficiency of these twoindividuals, there would have to be at least a 6 point difference in their scores.

    Table 1.6 Examples of MELAB Scores

    ExamineePart I

    CompositionPart 2

    ListeningPart 3GCVR

    FinalMELAB Score

    A 77 84 82 81B 77 81 79 79C 65 86 73 75D 77 62 85 75

    When interpreting MELAB scores, both part scores and the final score should be considered. Forexample, consider two individuals (C and D) who have the same final scores but quite differentpart scores (see Table 1.6).

    Even though C and D have the same final MELAB score, a 75, their scores on the three partssuggest differences in their language proficiency. Such differences may affect their ability to useEnglish effectively in different contexts.

    When interpreting language proficiency test scores, it is also important to consider that factorsother than language also affect how well someone can communicate. Lyle Bachman inFundamentals of Language Testing (1990)12 theorizes that communicative language abilityconsists of both knowledge of language and knowledge of the world. In the general context oflanguage used when pursuing academic studies in English, it follows that the ability to function inthis setting involves not only knowledge of English, but also other knowledge and skills such asintellectual knowledge and study skills. Language is just one of many factors that affect success,or lack of success, in an academic setting. Consequently, MELAB scores should not be used topredict academic success or failure.

    When MELAB scores are used by an institution to provide evidence of English languageproficiency, the institution must determine what level of English language proficiency is desirablefor that institutional context. Relevant factors to consider include the nature of the ESL servicesavailable to non-native speakers of English and the linguistic demands of the instructionalcontext. Examples of how different types of institutions use MELAB scores appear in sections1.5.2 to 1.5.4 below.

    A MELAB score, like a snapshot, captures characteristics of a person at a particular time; and,like a snapshot, it can become outdated. Because language proficiency may change over time,MELAB score users should always consider how recently the test was administered wheninterpreting a MELAB score. The ELI-UM will not issue score reports for tests taken more thantwo years in the past.

    We know that various factors (such as amount and quality of instruction, experience using thelanguage, motivation, language background, and individual differences in language ability) affectthe development of language proficiency, but these factors affect various people differently. It isnot possible to predict the length of time or type of instruction necessary for an individual to

    12Bachman, L. (1990). Fundamentals of language testing. Oxford: Oxford University Press.

  • 14

    demonstrate a particular performance level on the MELAB or to achieve a score gain of aparticular amount on the MELAB.

    1.5.2 An Example of MELAB Use for a University Context

    The University of Michigan is a large public university offering undergraduate and graduateprograms in a variety of disciplines. Generally, international students are admitted for full-timeacademic study rather than part-time academic study. Upon matriculation, an internationalstudent is expected to take a full academic load. Only non-native speakers with good oradvanced proficiency in English are considered for admission, but it is accepted that somestudents may need some ESL work to enhance their written and oral communication skills.

    During the academic year, a range of courses is available to help non-native speakers becomeeffective and fully participating members of the academic community. These ESL courses areprimarily known as English for Academic Purposes (EAP) courses and focus on improving writtenand oral communication. Most of these courses are one credit hour courses that meet 2 or 3hours a week for 14 weeks.

    1.5.2.1 Undergraduates

    At the University of Michigan, the various admissions officers inform applicants who are non-native speakers of English that English proficiency requirements can be met with MELAB orTOEFL scores.

    Applicants are considered for admission to undergraduate study with Final MELAB scores above80 and all part scores at 80 or above, or with TOEFL scores above 560 and all section scores at56 or above. Rigid cut scores are not applied. All relevant information about English languageproficiency is used by admissions staff so the policy is applied in a flexible manner. Applicantswhose Final MELAB or part scores are 85 or lower, or whose TOEFL scores are 600 or lower orwhose section scores are 60 or lower (and also those without a Test of Written English13 score ofat least 5.0) are generally required to have their English language proficiency re-evaluated uponarrival. As a result of this on-campus testing, students may be required to take an EAP (Englishfor Academic Purposes) course. Typically, at UM about half of the entering undergraduates areexempted from English language work. The typical requirement for the other half is one EAPmini-course, usually a writing course that meets two hours a week. The EAP writing course mustbe taken before the student enrolls in a regular university composition course. EAP courses aretaken concurrently with other academic course work.

    1.5.2.2 Graduates

    At the University of Michigan, applicants to graduate programs who already have received adegree from an accredited U.S. institution are not required to provide evidence of languageproficiency with a MELAB or TOEFL score. However, they are evaluated for English proficiencyon campus prior to beginning their first term of admission. Applicants whose previous degree isnot from a U.S. institution or whose degree is from a U.S. institution where the majority ofinstruction is in a language other than English are required to show evidence of Englishproficiency through MELAB or TOEFL scores. Generally, applicants must have Final MELABscores of at least 80 or TOEFL scores of at least 560. Applicants to Biological and PhysicalScience or Engineering generally need a Final MELAB score of at least 80 to be considered foradmission. Those applying to programs in the social sciences and humanities generally need aFinal MELAB score of at least 85 to be considered for admission. As is the case forundergraduate students, there is flexibility when using language proficiency test scores foradmissions decisions, and graduate students who have a Final MELAB score of 85 or below, orpart scores of 85 or below (or TOEFL scores at or below 600/60 or TWE below 5.0) are requiredto be re-evaluated upon arrival on campus. Typically, 65 percent of the graduate students

    13Test of Written English. Princeton, NJ: Educational Testing Service.

  • 15

    reassessed on campus are required to take supplementary EAP instruction of 2 to 4 hours aweek concurrently with their regular academic load.

    We have found at the University of Michigan that MELAB scores need to be interpreted withsome flexibility and that factors such as educational background, first language of the learner,and aural acuity can differentially influence the rate of development of proficiency. Althoughgenerally a Final MELAB score of at least 80 has been the recommended minimum foradmission, there have been special cases in which a graduate student with a score in the low70's has been recommended for admission. An individual may obtain a below average score onPart 3 of the MELAB but obtain significantly above average scores on Part 1 (Composition) andPart 2 (Listening), and the examinee might demonstrate effective communication strategies in thespeaking test. Such an individual might be viewed as nearly adequate and recommended for areduced course load the first term. Prior practical experience in the individual's area of graduatestudy can also compensate for gaps in range and accuracy of English, and English languageimprovement courses can be taken concurrently with graduate academic courses.

    1.5.3 An Example of MELAB Use for a Community College Context

    Washtenaw Community College (WCC) is a two-year community college in Michigan thatrequires international applicants to provide evidence of English proficiency before it issues theauthorization form for a student visa. The minimum MELAB score for admission is 75, and theminimum TOEFL score is 500. Upon arrival, these international students, along with all otherstudents, take an exam of basic skills in reading, writing, and mathematics (the ASSET test)14 .International students with low ASSET scores may be required to take one or more of WCC'sEnglish as a second language classes which are offered for new international students and forresident students for whom English is a second language.

    1.5.4 An Example of MELAB Use for Professional Contexts

    The MELAB was developed to give evidence of English proficiency for academic purposes.However, the test may also be appropriate for assessing the proficiency of individuals who needEnglish for academic examinations certifying their professional competence. Because at leastmodest proficiency in spoken and written English is necessary to succeed on a professionalexamination in English, those with low MELAB scores may not be allowed to take certainprofessional exams. One agency, the Michigan State Board of Nursing, Department of Licensingand Regulation, has specified that the following MELAB scores must be obtained before sitting forthe licensing examination:

    Final MELAB: not less than 75Part scores: none less than 70Oral rating: at least 3

    Various professional agencies may wish to establish certain minimal scores to meet the specificpurposes for which examinees are having their proficiency evaluated. The English LanguageInstitute Testing Staff will work with professional agencies in designing studies to establishappropriate MELAB scores for various academic and professional programs.

    14ACT ASSET Program. Iowa City, IA: The American College Testing Program (ACT).

  • 16

    1.6 PREPARING FOR THE MELAB

    The MELAB is a general language proficiency test that is not linked to any particular book,language study program, or course of study. The best way to develop proficiency in a languageis through active use of the language for communication, combined with study of materials thatwiden exposure to the language. A variety of English language learning materials is available atbookstores and libraries. Some materials are also available that give students practice withmultiple-choice tests. The ELI does not sell any particular test preparation materials; it does,though, recommend A Student's Guide to the MELAB15 by Mary Spaan to examinees who wantextra practice with test questions in the MELAB format. Examinees may also become familiarwith the format of the test by working through the sample problems in the free MELAB InformationBulletin. They may prepare for the impromptu composition portion by writing on a topic for 30minutes. They may prepare for the listening test by giving themselves frequent opportunities tolisten to spoken English.

    1.7 TEST SECURITY/INVALIDATIONS

    Test security is taken extremely seriously in an effort to ensure that MELAB test scores actuallyreflect the proficiency of the examinee. The MELAB is never sold to students, educators,libraries, or the general public. Examiners agree to keep all MELAB materials in a secure, lockedplace when not in use. Examiners are knowledgeable about the testing procedures. A detailedoutline of their duties is given in the Administration Manual for MELAB Examiners16. Examinersare instructed as to what procedures to follow if test compromise is suspected. If testcompromise occurs, the examinee's test will be invalidated and any schools that received scoreswill be notified. The exam center will also be notified, and the examinee will not be allowed to testagain.

    While it is rare, it does occasionally happen that examinees try to cheat on their MELAB. Theymight try to have someone else take the test for them, or they might try to copy from otherexaminees' test papers. Examiners thoroughly check identification before the test. Twophotographs that must look like the person appearing for the test are collected and sent with thetest papers to the ELI-UM. Test forms are alternated in administrations where more than oneexaminee is being tested, which makes copying useless. Test papers and photos are kept on fileat the ELI-UM. Any institute or admissions office suspicious of an applicant's scores may requesta photograph and/or handwriting sample (from the written composition). Official score reports aresent directly from the ELI-UM to an admissions office and bear an embossed seal over the FinalScore. Any report that appears tampered with or any examinee copy or photocopy of a MELABscore report should not be accepted as official, and the ELI-UM should be notified.

    On occasion, when an examinee takes the MELAB a second time, the scores on the two tests areextremely different. We know from our years of experience testing thousands of examinees thatscores do not normally show extreme variation in a short time. For example, an examineescoring a 65 on the GCVR section might after six weeks take the test again and score in the 70's,but it would be very unlikely for the second score to be in the high 80's or 90's. It is that sort ofvariation that makes us take a closer look at tests, to see if there has been any test compromise.Because a composition is an integral part of every MELAB, we have handwriting samples fromboth test administrations. The examiners who administered the MELAB may be contacted wheninvestigating any irregularities. Examiners also inform ELI-UM if anything unusual happens whilethey are administering the MELAB, and this is investigated by the ELI-UM.

    We strongly encourage officials at institutions to contact the English Language Institute MELABtesting office at the University of Michigan in Ann Arbor if there is any doubt about the authenticityor veracity of a MELAB score report. This is a FREE service and all requests are handledpromptly.

    15Spaan, M. (1992). A student's guide to the MELAB. Ann Arbor, MI: University of Michigan Press.16Administration Manual for MELAB Examiners is available from the ELI-UM Testing Division.

  • 17

    SECTION 2: MELAB STATISTICS

    2.1 GENERAL DESCRIPTIVE STATISTICS

    Table 2.1 presents general descriptive test statistics (across all test forms) for MELAB part scoresand the final MELAB score. The statistics are based on the test papers of 4,811 examineestaking the MELAB for the first time between 1991 and 1993.1 Of these examinees, 67.8 percentwere tested in the United States or Canada; 32.2 percent were tested overseas. They spoke 78different native languages. The most common native language was Chinese (20.2 percent).Others of the ten most common languages are, in descending order of frequency, Arabic, Farsi,Spanish, Japanese, Korean, Vietnamese, Russian, German, and Portuguese. Sixty-sevenpercent of the sample spoke one of these ten languages. For mean MELAB scores of this groupclassified by reason for testing, by sex, and by age, see Tables 2.7, 2.8, and 2.9, respectively.

    Table 2.1 Score Descriptives for 4,811 First-Time MELABs Administered 1991-1993

    Part 1(Composition)

    Part 2(Listening)

    Part 3(GCVR)

    FinalMELAB

    Minimum Scaled Score 53 33 25 38Maximum Scaled Score 97 100 100 99Median Scaled Score 75 79 77 77Mean Scaled Score 75.42 77.40 74.69 75.84Standard Deviation 7.90 12.13 14.87 10.40Reliability1 .90 .89 .94 .91SEM2 2.50 4.02 3.64 3.12

    1 Reliability figures were calculated using the mean interrater correlation for Part 1 (see Section 3.1; note that the set ofcompositions used to calculate this coefficient is not identical to the set summarized in Table 2.1) and from KR 21applied to raw scores for Part 2 and Part 3. The reliability estimate for the Final MELAB Score is the mean of theseestimates for Part 1, Part 2, and Part 3.2 Standard error of measurement

    Table 2.2 shows MELAB part scores and final scores that correspond to specified percentiles.Using this table, it is possible to see what MELAB score (final or scaled part score) indicates thata particular examinee scored higher than a certain percentage of all examinees tested.

    For example, to find out what final MELAB score indicates that an examinee scored higher than50 percent of all examinees tested, find 50 in the column labeled Percentile, and look at thecolumn on the far right. A final MELAB score of 77 corresponds to the 50th percentile (for finalMELAB scores). The table can also be used to see what MELAB part scores correspond to agiven percentile. For example, to find out what Listening Test score indicates that an examineegot a higher Listening Test score than 50 percent of all the examinees, look at the column labeledPart 2 (Listening). An examinee must score 79 on the Listening Test to be at the 50th percentile(in terms of Listening Test scores).

    1The statistics are very similar to those calculated for tests taken prior to 1991 (see Appendix E for data on 13,588 first-time MELABs administered between 1987 and 1990).

  • 18

    Table 2.2 MELAB Scaled Scores Corresponding to Specified Percentiles (Based on 4,811 first-time MELABs administered 1991-1993)

    Percentile Part 1Composition

    Part 2Listening

    Part 3GCVR

    FinalMELAB

    99 95 98 99 9695 90 93 95 9190 87 91 93 8985 85 89 91 8780 83 88 88 8575 80 86 86 83

    70 80 85 85 8265 77 84 83 8160 77 82 81 8055 75 81 79 7850 75 79 77 77

    45 75 78 75 7640 73 76 72 7435 73 75 70 7230 70 73 67 7125 70 70 65 69

    20 67 67 62 6715 67 64 58 6410 65 60 53 62 5 63 54 46 57 0 53 33 25 38

    2.2 FREQUENCY DISTRIBUTION OF MELAB SCORES FOR ALL EXAMINEES

    Tables 2.3 through 2.6 give information about score distribution on the MELAB. First, they showthe number and the percent of examinees who obtained a particular score on MELAB (Part 1,Part 2, Part 3, or on the Final MELAB). In the far right column of each table is the cumulativepercent that corresponds to each score. The cumulative percent for a given score point is thepercentage of examinees with that score or a lower score. Thus, if an examinee's scorecorresponds to the cumulative percent of 70, that examinee scored as well as or better than 70percent of all the examinees. These tables are based on MELABs (all forms) administeredbetween 1991 and 1993 (see Appendix E for similar information on MELABs administeredbetween 1987 and 1990).

  • 19

    FinalMELAB Score N

    Percent withthe Score

    CumulativePercent

    99 3 0.1 100.098 7 0.1 99.997 14 0.3 99.896 22 0.5 99.595 22 0.5 99.094 39 0.8 98.693 48 1.0 97.892 76 1.6 96.891 80 1.7 95.290 95 2.0 93.589 99 2.1 91.688 128 2.7 89.587 137 2.8 86.886 154 3.2 84.085 118 2.5 80.884 157 3.3 78.383 196 4.1 75.182 165 3.4 71.081 181 3.8 67.680 207 4.3 63.879 173 3.6 59.578 191 4.0 55.977 176 3.7 51.976 168 3.5 48.375 142 3.0 44.874 166 3.5 41.873 138 2.9 38.472 150 3.1 35.5

    FinalMELAB Score N

    Percent withthe Score

    CumulativePercent

    71 139 2.9 32.470 123 2.6 29.569 126 2.6 27.068 126 2.6 24.367 103 2.1 21.766 85 1.8 19.665 103 2.1 17.864 96 2.0 18.763 102 2.1 13.762 83 1.7 11.661 64 1.3 9.860 46 1.0 8.559 39 0.8 7.558 37 0.8 6.757 62 1.3 6.056 50 1.0 4.755 42 0.9 3.654 21 0.4 2.853 23 0.5 2.352 31 0.6 1.851 16 0.3 1.250 11 0.2 0.949 10 0.2 0.648 7 0.1 0.447 5 0.1 0.346 3 0.1 0.244 5 0.1 0.138 1 0.0 0.0

    Table 2.4 Frequency Distribution of MELAB Part 1 (Composition) Scores (Based on 4,811 first-time MELABs administered 1991-1993)

    MELABPart 1 Score N

    Percent withthe Score

    CumulativePercent

    97 18 0.4 100.095 46 1.0 99.693 101 2.1 98.790 135 2.8 96.687 191 4.0 93.885 254 5.3 89.883 334 6.9 84.580 393 8.2 77.677 586 12.2 69.475 600 12.5 57.273 692 14.4 44.870 482 10.0 30.467 468 9.7 20.365 207 4.3 10.663 199 4.1 6.360 59 1.2 2.257 24 0.5 1.055 14 0.3 0.553 8 0.2 0.2

    Table 2.3 Frequency Distribution of Final MELAB Scores (Based on 4,811 first-time MELABs administered 1991-1993)

  • 20

    Table 2.5 Frequency Distribution of MELAB Part 2 (Listening) Scaled Scores (Based on 4,811 first-time MELABs administered 1991-1993)

    MELABPart 2 Score N

    Percent withthe Score

    CumulativePercent

    100 18 0.4 100.0 98 71 1.5 99.6 96 54 1.1 98.2 95 22 0.5 97.0 94 67 1.4 96.6 93 43 0.9 95.2 92 132 2.7 94.3 91 112 2.3 91.5 90 159 3.3 89.2 89 198 4.1 85.9 88 103 2.1 81.8 87 203 4.2 79.7 86 209 4.3 75.4 85 180 3.7 71.1 84 200 4.2 67.3 83 146 3.0 63.2 82 224 4.7 60.2 81 124 2.6 55.5 80 131 2.7 52.9 79 150 3.1 50.2 78 121 2.5 47.1 77 167 3.5 44.6 76 195 4.1 41.1 75 165 3.4 37.0 74 122 2.5 33.6 73 81 1.7 31.1 72 134 2.8 29.4 71 43 0.9 26.6 70 124 2.6 25.7

    MELABPart 2 Score N

    Percent withthe Score

    CumulativePercent

    69 78 1.6 23.168 61 1.3 21.567 100 2.1 20.266 64 1.3 18.265 61 1.3 16.864 51 1.1 15.663 52 1.1 14.562 51 1.1 13.461 60 1.2 12.460 53 1.1 11.159 63 1.3 10.058 51 1.1 8.757 22 0.5 7.656 70 1.5 7.255 22 0.5 5.754 48 1.0 5.353 28 0.6 4.352 37 0.8 3.750 29 0.6 2.949 15 0.3 2.347 19 0.4 2.046 20 0.4 1.645 4 0.1 1.243 9 0.2 1.142 4 0.1 0.940 19 0.4 0.937 10 0.2 0.535 8 0.2 0.233 4 0.1 0.1

  • 21

    Table 2.6 Frequency Distribution of MELAB Part 3 (GCVR) Scaled Scores (Based on 4,811 first-time MELABs administered 1991-1993)

    MELABPart 3 Score N

    Percent withthe Score

    CumulativePercent

    100 18 0.4 100.0 99 36 0.7 99.6 98 42 0.9 98.9 97 59 1.2 98.0 96 69 1.4 96.8 95 83 1.7 95.3 94 91 1.9 93.6 93 92 1.9 91.7 92 106 2.2 89.8 91 129 2.7 87.6 90 66 1.4 84.9 89 106 2.2 83.6 88 115 2.4 81.4 87 169 3.5 79.0 86 124 2.6 75.5 85 141 2.9 72.9 84 177 3.7 69.9 83 109 2.3 66.3 82 119 2.5 64.0 81 122 2.5 61.5 80 134 2.8 59.0 79 137 2.8 56.2 78 129 2.7 53.4 77 102 2.1 50.7 76 106 2.2 48.6 75 115 2.4 46.4 74 81 1.7 44.0 73 88 1.8 42.3 72 88 1.8 40.4 71 107 2.2 38.6 70 105 2.2 36.4 69 82 1.7 34.2 68 89 1.8 32.5 67 139 2.9 30.7 66 105 2.2 27.8

    MELABPart 3 Score N

    Percent withthe Score

    CumulativePercent

    65 114 2.4 25.664 55 1.1 23.263 85 1.8 22.162 74 1.5 20.361 64 1.3 18.860 37 0.8 17.459 62 1.3 16.758 18 0.4 15.457 53 1.1 15.056 48 1.0 13.955 73 1.5 12.954 42 0.9 11.453 48 1.0 10.552 16 0.3 9.551 43 0.9 9.250 33 0.7 8.349 31 0.6 7.648 51 1.1 7.047 25 0.5 5.946 29 0.6 5.445 34 0.7 4.844 42 0.9 4.143 59 1.2 3.240 25 0.5 2.039 17 0.4 1.538 12 0.2 1.137 5 0.1 0.936 18 0.4 0.735 2 0.0 0.434 6 0.1 0.333 2 0.0 0.232 3 0.1 0.231 3 0.1 0.130 1 0.0 0.025 1 0.0 0.0

  • 22

    2.3 PERFORMANCE OF REFERENCE GROUPS ON THE MELAB

    Tables 2.7 through 2.10 present descriptive information on the performance of various groups ofMELAB examinees. It is important to keep in mind that the statistics describe performance ofexaminees who themselves elected to take a MELAB and that all group classification is based oninformation supplied by the examinees. The group statistics cannot be assumed to berepresentative of the general population, and the data in the tables should not be used to makegeneralizations about differences in the English language proficiency of such groups in thegeneral population.

    Table 2.7 MELAB Scaled Score Mean and Standard Deviation by Examinee Reason for Testing

    (based on information provided by first-time MELAB examinees from 1991 to 1993)

    Reason forTesting

    Number Part 1Mean

    Part 1SD

    Part 2Mean

    Part 2SD

    Part 3Mean

    Part 3SD

    FinalMean

    FinalSD

    To enter a 2-year college

    337 72.16 7.09 75.19 12.07 68.53 14.81 71.96 10.09

    To enter a 4-year college

    1981 74.36 7.59 77.90 11.75 72.28 14.73 74.85 10.26

    To enter a university for graduate work

    1260 76.55 8.00 78.41 11.59 78.30 13.00 77.76 9.60

    For professional certification

    177 77.12 7.85 77.80 11.57 81.54 14.14 78.82 9.94

    Other 721 77.25 8.25 76.84 13.86 77.73 15.78 77.29 11.48

    Table 2.8 MELAB Scaled Score Mean and Standard Deviation by Sex (based on information provided by first-time MELAB examinees from 1991 to 1993)

    Sex Number Part 1Mean

    Part 1SD

    Part 2Mean

    Part 2SD

    Part 3Mean

    Part 3SD

    FinalMean

    FinalSD

    Male 2719 75.12 7.97 76.72 12.24 74.22 15.09 75.35 10.50Female 2089 75.82 7.79 78.28 11.94 75.31 14.57 76.47 10.24

    Table 2.9 MELAB Scaled Score Mean and Standard Deviation by Age (based on information provided by first-time MELAB examinees from 1991 to 1993)

    Age Range Number Part 1Mean

    Part 1SD

    Part 2Mean

    Part 2SD

    Part 3Mean

    Part 3SD

    FinalMean

    FinalSD

    35 521 76.10 8.71 72.09 13.23 74.77 15.38 74.33 11.18

  • 23

    Table 2.10 Mean and Standard Deviation by Native Language1 for Parts 1, 2, 3 and Final MELAB Scores

    (based on 13,588 first-time MELABs administered 1987-1990)

    PART 1 PART 2 PART 3 FINAL

    LANGUAGE2 Cases Mean SD Mean SD Mean SD Mean SD

    Ibo 47 80.85 7.98 72.77 8.33 84.51 8.62 79.38 7.24Yoruba 33 83.33 7.17 78.00 8.60 86.64 10.53 82.76 7.46

    Amharic 106 72.64 8.08 71.40 9.02 67.92 15.00 70.60 9.44Arabic 1914 69.96 6.80 72.31 10.97 63.10 14.97 68.48 9.58Hebrew 122 75.82 6.52 86.19 9.22 78.60 12.02 80.17 7.88Somali 67 72.97 8.67 70.08 11.64 68.10 17.97 70.39 11.48Tigre 36 74.28 8.20 73.67 11.00 70.56 15.93 72.81 10.35

    Malayalam 40 80.85 7.77 78.70 9.05 82.88 11.65 80.85 8.36Tamil 99 80.35 8.77 81.76 9.16 83.78 12.26 81.93 9.19Telegu 68 79.29 7.07 77.15 9.66 80.93 11.79 79.13 8.61

    Chinese 2710 73.32 5.87 75.37 9.97 76.08 12.13 74.91 8.03

    Cambodian 46 71.22 5.01 71.48 8.58 69.46 8.30 70.78 5.81Indonesian 277 72.12 5.44 73.00 10.57 69.00 13.06 71.43 8.16Malay 76 77.20 6.30 80.12 7.65 79.90 11.48 79.07 7.26Tagalog 205 82.21 6.77 82.18 7.52 87.00 9.32 83.83 6.87Vietnamese 482 70.83 5.84 70.03 11.25 66.70 15.76 69.19 9.67

    Burmese 26 77.42 6.98 76.92 9.39 81.00 13.33 78.42 8.71Hmong 87 73.13 6.53 73.59 8.86 70.77 12.26 72.53 8.14Lao 72 70.75 5.86 76.11 7.26 67.32 15.43 71.40 8.54Thai 192 70.33 5.21 70.14 10.04 68.08 11.30 69.50 7.14

    Japanese 1208 68.09 6.72 68.10 12.18 61.46 14.28 65.83 9.76Korean 483 70.83 5.79 75.52 9.28 73.99 13.13 73.45 7.89Turkish 238 70.77 9.36 71.15 14.43 64.12 16.97 68.65 12.81

    Finnish 48 82.48 6.82 86.27 7.31 85.40 9.29 84.73 6.87Hungarian 96 77.50 5.85 78.85 10.68 77.43 12.52 77.90 8.11

    Bengali 94 78.14 8.00 72.88 10.92 76.67 13.94 75.92 9.00Farsi 719 72.59 6.50 76.62 11.36 70.95 13.97 73.38 9.37Gujarati 96 78.17 9.22 78.44 9.82 78.16 15.45 78.20 10.79Hindi 114 82.22 8.27 80.44 11.99 85.29 12.20 82.61 9.51Punjabi 93 79.10 8.07 75.57 12.67 79.60 12.80 78.10 10.20Urdu 164 78.38 8.05 77.62 10.50 76.78 14.54 77.60 10.09

    Greek 161 77.38 6.87 80.73 9.81 76.44 13.61 78.17 9.03

    Polish 232 79.17 6.24 81.00 10.53 78.52 12.25 79.55 8.48Russian 109 76.10 7.09 76.51 12.47 72.31 16.08 75.02 10.73SerboCroatian 76 76.30 6.25 82.09 9.18 76.71 14.06 78.36 9.00

    French 318 77.94 7.29 77.84 12.64 77.44 15.07 77.74 10.53Italian 59 79.83 6.03 81.73 9.04 81.78 11.97 81.07 8.16Portuguese 297 76.67 6.20 78.02 11.79 76.23 12.86 76.92 9.07Romanian 83 80.95 7.50 81.84 10.13 82.81 10.86 81.91 8.48Spanish 1028 74.44 7.15 76.29 11.06 73.52 14.26 74.75 9.72

    Danish 25 82.56 5.23 89.40 6.06 87.64 6.77 86.52 5.36Dutch 67 82.63 7.11 90.08 5.54 90.05 6.51 87.60 5.28German 290 82.84 6.66 87.38 7.03 84.34 11.36 84.87 7.36Norwegian 35 79.89 7.32 86.74 8.91 84.29 10.11 83.69 7.81Swedish 70 80.06 6.54 87.46 7.49 83.46 11.04 83.63 7.56

    1Only those languages that were represented at least 25 times are included in the table.2 Languages are grouped by family and sub-family.

  • 24

    2.4 INTERCORRELATIONS AMONG MELAB SCORES AND MELAB ORAL RATING

    Table 2.11 shows the intercorrelations of scaled MELAB part scores, Final MELAB scores, andscores on the optional MELAB Speaking Test. The correlation coefficients are measures of theextent of the relationships among the various subtests and tests. The coefficients in Table 2.11are based on first time MELABs administered between 1991 and 1993. For correlations involvingthe oral rating, the number of cases is 1,076, a subset of the 4,811 used to calculate the othercorrelation coefficients. The correlation coefficients based on 13,588 MELABs administeredbetween 1987 and 1990 are virtually identical to those shown here (see Appendix E).

    In general, the correlation coefficients are moderate to moderately high. This implies, first of all,that there is overlap in what the various tests assess (which would be expected given thecommonly accepted premise that people who are highly skilled in one area of languageproficiency also tend to be skilled in other areas of language proficiency). Secondly, it suggeststhat although there is this overlap, the various tests do provide some unique information aboutexaminees' s