How Reliable and Valid is the Japanese Language...

25
32 JALT JournAL How Reliable and Valid is the Japanese Version of the Strategy Inventory for Language Learning (SILL)? Gordon Robson Showa Women’s University Hideko Midorikawa Showa Women’s University This study looks at the internal reliability of the Strategy Inventory for Language Learning (Oxford, 990), using the ESL/EFL version in Japanese translation. The results of the Cronbach’s alpha analysis indicate a high degree of reliability for the overall questionnaire, but less so for the six subsections. Moreover, the test-retest correlations for the two administrations are extremely low with an average shared variance of 9.5 percent at the item level and 25.5 percent at the subsection level. In addition, the construct validity of the SILL was examined using exploratory factor analysis. While the SILL claims to be measuring six types of strategies, the two factor analyses include as many as 5 factors. Moreover, an attempt to fit the two administrations into a six-factor solution results in a disorganized scattering of the questionnaire items. Finally, interviews with participating students raised questions about the ability of participants to understand the metalanguage used in the questionnaire as well as the appropriateness of some items for a Japanese and EFL setting. The authors conclude that despite the popularity of the SILL, use and interpretation of its results are problematic. 本研究は、Oxford(1990)の外国語学習ストラテジー・インベントリー (SILL)のEFL/ESL用日本語版の内部信頼性及び構成概念妥当性を実験と統計に よって検証したものである。クロンバック・アルファ検定による内部信頼性 については、インベントリーの全項目は全体としては信頼性が高かったが、 6タイプのサブカテゴリーに分類されたストラテジーについては信頼性が低か った。また、インベントリーを用いたテスト・再テストの相関は低く、全項 目では平均寄与率19.5パーセント、サブカテゴリーでは25.5パーセントであっ た。構成概念妥当性検定のための説明的因子分析の結果は、6タイプのストラ テジーが15因子に細分化されたこと、さらに、全項目を6因子に分けた結果、 それぞれの因子が無秩序に分類される結果となった。最後に、インタビュー によって、この実験に参加した被験者学生にインベントリーの各項目の内容 理解について確認した結果、日本語がわかりにくく判断しいくい記述、日本 のEFLの状況では理解しにくい記述があることが明らかになった。以上のす べてから、SILLの実用的評価にもかかわらず、それを用いること、また、そ こから得た結果の解釈には問題が含まれているというのが、本研究の研究者 が得た結論である。

Transcript of How Reliable and Valid is the Japanese Language...

Page 1: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

3232

JALT JournAL

How Reliable and Valid is the Japanese Version of the Strategy Inventory for Language Learning (SILL)?

Gordon RobsonShowa Women’s UniversityHideko MidorikawaShowa Women’s University

ThisstudylooksattheinternalreliabilityoftheStrategyInventoryforLanguageLearning(Oxford,�990),usingtheESL/EFLversioninJapanesetranslation.Theresultsof theCronbach’salphaanalysis indicateahighdegreeof reliabilityfortheoverallquestionnaire,butlesssoforthesixsubsections.Moreover,thetest-retestcorrelationsforthetwoadministrationsareextremelylowwithanaveragesharedvarianceof�9.5percentattheitemleveland25.5percentatthesubsectionlevel.Inaddition,theconstructvalidityoftheSILLwasexaminedusingexploratoryfactoranalysis.While theSILLclaimstobemeasuringsixtypesof strategies, the two factor analyses include asmany as �5 factors.Moreover,anattempttofit thetwoadministrationsintoasix-factorsolutionresultsinadisorganizedscatteringofthequestionnaireitems.Finally,interviewswithparticipatingstudentsraisedquestionsabouttheabilityofparticipantstounderstand themetalanguageused in thequestionnaire aswell as theappropriatenessof some items fora JapaneseandEFL setting.TheauthorsconcludethatdespitethepopularityoftheSILL,useandinterpretationofitsresultsareproblematic.

本研究は、Oxford(1990)の外国語学習ストラテジー・インベントリー(SILL)のEFL/ESL用日本語版の内部信頼性及び構成概念妥当性を実験と統計によって検証したものである。クロンバック・アルファ検定による内部信頼性については、インベントリーの全項目は全体としては信頼性が高かったが、6タイプのサブカテゴリーに分類されたストラテジーについては信頼性が低かった。また、インベントリーを用いたテスト・再テストの相関は低く、全項目では平均寄与率19.5パーセント、サブカテゴリーでは25.5パーセントであった。構成概念妥当性検定のための説明的因子分析の結果は、6タイプのストラテジーが15因子に細分化されたこと、さらに、全項目を6因子に分けた結果、それぞれの因子が無秩序に分類される結果となった。最後に、インタビューによって、この実験に参加した被験者学生にインベントリーの各項目の内容理解について確認した結果、日本語がわかりにくく判断しいくい記述、日本のEFLの状況では理解しにくい記述があることが明らかになった。以上のすべてから、SILLの実用的評価にもかかわらず、それを用いること、また、そこから得た結果の解釈には問題が含まれているというのが、本研究の研究者が得た結論である。

malcolmswanson
Text Box
Note: Due to file damage, this issue may not be accurate or complete. Please use the print version for referencing purposes
Page 2: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

3333robson & MidorikAwA

Theuseofself-reportinstrumentstoinvestigatevariousaspectsofindividuallearnerdifferencesisacommonandacceptedprac-tice in thefieldofsecondlanguageacquisitionresearch.Asa

consequence,alargenumberofsuchinstrumentshavebeendevelopedandusedover theyears.These include theAttitude/MotivationTestBattery(A/MTB)(GardnerandLambert,�972),theForeignLanguageClassroomAnxietyScale(FLCAS)(Horwitz,Horwitz,andCope,�986),andtheinstrumentunderdiscussionhere,theStrategyInventoryforLanguageLearning(SILL)(Oxford,�990).However,despitethewideacceptanceanduseoftheseinstruments,issuessuchastheirreliabilityandvalidityareoftenlostintheenthusiasmtofindoutwhatstudentsreallyfeelorbelieve.Althoughagiveninstrumentmayhavebeenrigor-ouslydevelopedandevensubjectedtovariousmeasuresofreliabilityandvalidity,whenitistranslatedintoanotherlanguageorusedinaculturalsettingdifferentfromtheoneoriginallyintended,itmustonceagainberigorouslyexamined,assuggestedbyGriffee(�999).

Thisreportwillpresenttheinitialresultsoftheresearchers’attemptstoprovidereliabilityandvaliditydataontheSILLinaJapaneseuniversitysetting.Thisstudyisgroundedintheresearchers’numerousotherat-temptstovalidateotherJapanesetranslationsofmeasuresofindividuallearnerdifferences,suchasmotivation,anxiety,learningstyles,learningbeliefs,andlearningstrategies.ReliabilityistypicallymeasuredthroughstatisticssuchasCronbach’salphaormultipleadministrationsofatestwiththesamesubjects,bothofwhichareusedhere.Regardingvalidity,althoughinthepastothermethodsofvalidatinghavebeenputforward,recentlyChapelle(�994)andMessick(�989)havepersuasivelyarguedforvaliditytobecondensedintoasingle,generalapproachwherethefocusisontheinstrumentasaconstruct.Asthemeasurestypicallyusedinthistypeofresearchhavebeenself-reportquestionnairesinwhichitemsweregroupedintocategoriesorsubscales,researchershavefa-voredfactoranalyticvalidationforthevariousgroupingsorcategoriesassignedtothequestionnaireitems.Theuseoffactoranalysistoconfirmatheorizedgroupingofitemsisalong-establishedpractice(Guilford&Fruchter,�973),especiallyinthefieldofpersonalityresearch,whereithasbeenusedtovalidateself-reportquestionnairesforover50years(seeforexampleAllport,�937;Guilford,�940;McCrae,�989).Therefore,thiswillbetheapproachtakeninvalidatingthesixgroupsofstrategiesmakinguptheSILL.

What is Reliability and Validity?

Therearevariousapproachestotestingandconfirmingthereliability

Page 3: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

3434

JALT JournAL

andvalidityofagivenresearchinstrument.Wecandefinethereliabilityastheproportionofthevariationintestscoresthatistruevariationandnoterror(Bachman,�990;Brown,�988).Typically,whenmeasuringreliability,theitemsonthequestionnairearesubjectedtooneormoretypesofstatisticalmeasurement.ThemostcommonlyemployedstatisticisCronbach’salpha,whichmeasurestheinternalconsistencyofatest.Anotherapproachistoobtainsimplecorrelationsbetweentestitemsofameasurethatisgiventothesamepopulationtwoormoretimes.Thisisreferredtoastest-retestreliability.

Ingeneral,thevalidationofaself-reportinstrumentismuchmoredifficultandinthepastinvolvedseveraldifferenttypesofvaliditysuchasfacevalidity,contentvalidity,constructvalidity,factoranalyticvalid-ity,andcriterion-relatedvalidity.However,inaninsightfularticleMes-sick(�989)pointsoutthatwhileitisimportanttovalidatethemethodofdatacollection,themorecrucialareaistovalidatetheinferences,interpretations,andactionstakenbasedonthescoresderivedfromthedata.Moreover,Chapelle(�994)arguesthat“constructvalidityiscentraltoallfacetsofvalidityinquiry,”andasanongoingprocess,thereisnoonceandforevervalidity(p.�6�).

Fromastatisticalpointofview,thereareseveralwaystoconfirmtheconstructvalidityofaninstrument.Theuseofcorrelationapproachesand factor analysishasbeennotedpreviously. Typically these ap-proachesinvolveusingseveraltestsorquestionnairesthatarebelievedtorepresentaconstruct,suchaslanguage-learningstrategies,andthenconfirmingthevalidityof the itemsthroughhighcorrelations. If thecorrelationsarehighenough,thenwecaninferthattheymeasurethehypothesizedconstruct.Factoranalysiscanbeusedwhenmeasuresforseveraldifferentconstructsarebeingused,suchasformotivation,strategies,andpersonality.Subsequently,theirloadingsondistinctfac-torsconfirmthattheymeasureseparateaspectsoflearnerbehavior.Aseconduseoffactoranalysisistobreakameasureintosubgroupings,suchasthesixhypothesizedpartsoftheSILL,andthenfactorthemtoseeifthesedivisionsarevalid.Evidenceofameasure’svaliditycanalsobeconfirmedexperimentallyorquasi-experimentallythroughrelatedoutcomesusing,forexample,ameasureoflanguagelearningstrategiesandscoresonsomemeasureoflanguagelearningsuchastheTOEFLTest.Thiswouldindicatenotonlythatthemeasurewasvalidlymeasur-ingstrategies,butalsothatsuchstrategieswereuseful.

Page 4: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

3535robson & MidorikAwA

The Strategy Inventory for Language Learning (SILL)

TheStrategy Inventory for LanguageLearning (SILL) is a self-reportquestionnaire for determining the frequencyof language learningstrategyuse.Itconsistsof50itemswithfiveLikert-scaleresponsesofneveroralmostnevertrueofme,generallynottrueofme,somewhattrueofme,generallytrueofme,alwaysoralmostalwaystrueofme.Basedonafactoranalysisofanearlier,largerversion,OxfordorganizedtheSILLintosixstrategysubscales:(a)MemoryStrategies(9items),(b)CognitiveStrategies(�4items),(c)CompensationStrategies(6items),(d)MetacognitiveStrategies(9items),(e)AffectiveStrategies(6items),and(f)SocialStrategies(6items).ThequestionnairewastranslatedintoJapaneseaspartoftheJapaneselanguageversion(Oxford,�990/�994)ofOxford’s(�990)LanguageLearningStrategies:WhatEveryTeacherShouldKnow.AlthoughOxforddoesnotdirectlydiscusstheprocessforestablishingthereliabilityandvalidityoftheSILL,anotetoChapterSixexplainsthatanearlier,�2�-itemversionoftheSILLwasfoundtohaveareliabilityof.96basedona�,200-personsampleand.95witha483-personsample.Shethengoesontostatethatthereliabilityof9of�0factorswasfoundtobemoderatetohighwithfiguresof.60to.86,althoughforthe�0thfactoritwasonly.3�.Thisisnotthetypicalwayofreportingtheresultsoffactoranalysis,andifthe�2�-itemversionwasclaimingtomeasuresixstrategytypes,thena�0-factorsolutionishardlyconfirmation.Thenotegoesontostatethatthefifty-itemversion7.0oftheSILLunderdiscussionherewasstillbeingassessedforreli-abilityandvalidity.Thus,whileitwouldseemthatthevariousversionsoftheSILLhaveaprovenlevelofreliability,thisdoesnotsuggestthatthequestionnaireisvalid.AsBachman(�990)hasstated,エheprimaryconcernintestdevelopmentanduseisdemonstratingnotonlythattestscoresarereliable,butthattheinterpretationsanduseswemakeoftestscoresarevalid(p.237).IfatthispointintheSILL’sconstructionitwerefoundtobeunreliable,therewouldbenoneedtoproceed,asanunreliablemeasureissimilarlynotvalid.

Oxford(�996)discussesthepsychometricqualitiesoftheSILL,andin termsof reliability, shecitesWatanabe (�990),wherea JapaneseversionoftheSILLachievedaCronbach’salphareliabilityof.92,andotherstudieswithsimilarreliabilitiesinthe.90range.Followingtheabove-mentionedMessick(�989)andChapelle(�994)approachtotestvalidity,OxfordexaminedanumberofstudieswheretheSILLcorrelatedsignificantlywithvariousmeasuresof language learning. InOxford,Park-Oh,Ito,andSumrall(�993),amultiple-regressionanalysisfoundlowbutsignificantpredictiverelationshipsbetweenstrategiesandfinal

Page 5: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

3636

JALT JournAL

testgrades(.20).Takeuchi(�993),alsousingmultiple-regressionanalysiswithlanguageachievementasmeasuredbytheComprehensiveEnglishLanguageTest (CELT), found that fourSILL items(�7,2�,22and32)positivelypredictedlanguageachievementwhilefouritems(6,30,43and49)negativelypredictedlanguagesuccess.Finally,Watanabe(�990)foundlowcorrelationsbetweenSILLitemsandstudents’self-ratingsoftheirownproficiency.Althoughtheseresultsprovidesomemeasureofvalidation,onlyafewSILLitemsareinvolved,andthecorrelationsareextremelylow.

InBrown,Robson,andRosenkjar(�996)anindependentlytranslatedversionoftheSILLwasusedinamultiple,individuallearnerdifferencesstudy.Theoverallreliabilityofthattranslatedversionwas.94withthereliability for the six strategy typesbeing .74 forMemoryStrategies,.84 forCognitiveStrategies, .69 forCompensationStrategies, .88 forMetacognitiveStrategies,.63forAffectiveStrategies,and.73forSocialStrategies.ThefactoranalysisintheBrownetal.(�996)studywasonlyusedtodetermineiftheSILLwasmeasuringsomethingdistinctfromtheothermeasuresofsuchvariablesaspersonality,anxiety,andmo-tivation.Thesixstrategytypeswerefoundtoloadonasinglefactor,whichconfirmedthattheSILLwasmeasuringavariabledistinctfromtheother instruments.The researchersknowofnootherpublishedstudythathasattemptedtoestablisheitherreliabilityorvalidityinthismannerusingaJapaneseversionoftheSILL.

However,atTESOL2000inVancouver,Canada,HsiaoandOxford(2000)presentedtheresultsofamulti-groupconfirmatoryfactoranalysisforan80-itemSILL.Thefactoranalysisplacedonly�7itemsintothesixhypothesizedgroupings,leaving63itemswithnorelationtothesixstrategycategorieshypothesized.The�7itemswereMemoryStrategies(4,5,8),CognitiveStrategies(26,27,28),CompensationStrategies(4�,43),MetacognitiveStrategies(49,53,55),AffectiveStrategies(66,68,69),andSocialStrategies(72,73,74).Ofthese,onlyitems5,27,28,68,and72arethesameasorsimilartoitemsonthe50-questionversionoftheSILLunderstudyhere.

Tosummarize,theSILLappearstoenjoyahighdegreeofreliabilityin its variousversions and the languages inwhich ithasbeenem-ployed.However,thereliabilityhasbeenfortheSILLasawhole,withtheexceptionofBrownetal.(�996),whereseveralofthescaleswereratherlow.Thisstillleavesthequestionofvalidity,whichbasedonthesourcesdiscussedseemsfarfromestablished,andhasledustoaskthefollowingresearchquestions.

Page 6: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

3737robson & MidorikAwA

Research Questions

�.HowreliableistheJapaneselanguageversionoftheSILLforJapaneseuniversitystudents?

2.TowhatdegreeistheJapaneselanguageversionoftheSILLvalidforJapaneseuniversitystudents?

Method

Thepresent study isbasedon twoadministrationsof theofficiallytranslatedSILL(Oxford�990/�994)tothesamegroupof�53Japaneseuniversitystudents.Thegroupwascomprisedof��0first-andsecond-yearfemalesand43first-andsecond-yearmalesstudyingataprivatewomen’suniversityandaprivatecoeducationaluniversityinTokyo.TheirEnglishproficiency levelwasapproximately lowintermediate.Thefirstadministrationwasconductedatthebeginningofthespringsemester.Asecondadministrationwasconductedduringthebeginningofthefallsemesterusingaversioninwhichtheorderoftheitemshadbeenrandomized.Therewerenochangesinthemakeupofthegroupofsubjectsforthetwoadministrationsofthequestionnaire.Inaddi-tion,post-administrationinterviewswereconductedwithtenrandomlyselectedstudents,fourmalesandsixfemalestogetfeedbackonwhatthestudentsthoughtaboutthequestionnaire.Theinterviewswerecon-ductedindividuallyinJapanesebytheJapanesenativespeakerauthorofthisstudywitheachoftheinterviewees.Theywerequestionedabouttheirthoughtsoneachofthe50itemsandtheirresponsesweretakendownintheformofnotes.

Analysis

ThedatacollectedfromthetwoadministrationsoftheSILLwerefirstanalyzedforitemstatisticsfollowedbydescriptivestatisticsforthesixpartsaswellastheentireSILL.Thealphalevelforallstatisticaldecisionswassetat .05.BothadministrationswerethenexaminedforinternalconsistencyusingCronbach’salphaforeachofthesixpartsaswellasoverallreliabilityforbothadministrations.Next,eachTimeOneitemwascomparedtoitsidenticalTimeTwoitemusingthePearsoncorrelation.Theresultingcorrelationswerethensquaredtodeterminethedegreeofsharedvariance.Thesquaredvalueofthecorrelationcoefficientcanbeinterpretedastheproportionofsimilaritybetweenthetwoitems(Hatch&Lazaraton,�99�).ThisprocedurewasrepeatedforthesixpartsoftheSILLandfortheentireSILLaswell.Finally,thetwoadministrationswere

Page 7: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

3838

JALT JournAL

examinedusingprincipalcomponentanalysis(PCA),whichisatypeofexploratoryfactoranalysis,withvarimaxrotationandeigenvaluessetatone.Thesearethetypicalproceduresforcarryingoutfactoranaly-sis.Asiscommon,loadingsof.30andabovewereconsideredstrongenoughforinclusioninagivenfactor(Hatch&Lazaraton,�99�).IntheinitialuseofPCA,theanalysiswasallowedtoselectasmanyfactorsascouldbefoundwithaneigenvalueover�.00;however,asecondPCAwasrunonbothadministrationsinwhichtheanalysiswasforcedtochoosesixfactorsbasedonOxford’stheorizedgrouping.ScreeplotsforallPCAswerealsocalculated.TheseadditionalprocedureswereconductedtoprovidetheSILLwithasmanyopportunitiesaspossibletosupplysupportforitstheoreticalbasis.Finally,thenotestakendur-ingtheinterviewswereexaminedtodeterminethetypesofdifficultiesthestudentshadunderstandingthequestionnaireitemsandhowtheirdifficultiescomparedtooneanother.

Results

Table�showstheitemsthemselveswiththeirgroupings,themeanoneachitemandthestandarddeviation,withTable2showingthemeansandstandarddeviations for the itemson thesecondadministration.Table3provides thedescriptive statistics for the six subsectionsoftheSILLandtheentireSILLforbothadministrations.Thedistributionsarealleitherpositivelyornegativelyskewedandthosewithskewnessstatisticsat�.0orgreaterareproblematic(Brown,�997).Theseskeweddistributionscanreducethetestreliabilityandareviolationsoftheas-sumptionsofnormalityforthecorrelationstatisticsandfactoranalysis,whichcouldadverselyaffecttheseresults.

Table�:MeanScoresandStandardDeviationsfortheItemsandTheirStrategyTypes,TimeOne(n=�53)

Item Statement Type M SD

� IthinkofrelationshipsbetweenwhatIalready knowand newthingsIlearninEnglish. Memo 2.79 0.942 IusenewEnglishwordsinasentencesoI canrememberthem. Memo 2.56 0.953 IconnectthesoundofanewEnglishword andanimageorpictureofthewordtohelp meremember. Memo 3.02 �.094 IrememberanewEnglishwordandan imageorpictureofasituationinwhichthe wordmightbeused. Memo 2.63 �.�25 IuserhymestoremembernewEnglishwords. Memo 2.4� �.��6 IuseflashcardstoremembernewEnglishwords. Memo 2.�9 �.42

Page 8: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

3939robson & MidorikAwA

7 IphysicallyactoutnewEnglishwords. Memo �.80 0.888 IreviewEnglishlessonsoften. Memo 2.66 0.669 IremembernewEnglishwordsorphrasesby rememberingtheirlocationonthepage,onthe board,oronastreetsign. Memo 2.56 �.24�0 IsayorwritenewEnglishwordsseveraltimes. Cog 3.98 0.99�� ItrytotalklikenativeEnglishspeakers. Cog 3.09 �.23�2 IpracticethesoundsofEnglish. Cog 3.40 �.05�3 IusetheEnglishwordsIknowindifferentways. Cog 2.89 0.96�4 IstartconversationsinEnglish. Cog 2.�7 0.86�5 IwatchEnglishlanguageTVshowsspokenin EnglishorgotomoviesspokeninEnglish. Cog 3.25 �.09�6 IreadforpleasureinEnglish. Cog 2.77 0.97�7 Iwritenotes,messages,lettersorreportsinEnglish.Cog 2.�9 �.06�8 IfirstskimanEnglishpassage(readoverthe passagequickly)thengobackandreadcarefully. Cog 3.39 �.09�9 Ilookforwordsinmyownlanguagethatare similartonewwordsinEnglish. Cog 2.39 �.�620 ItrytofindpatternsinEnglish. Cog 2.8��.072� IfindthemeaningofanEnglishwordbydividing itintopartsthatIunderstand. Cog 2.70 �.�822 Itrynottotranslateword-for-word. Cog 2.96 0.9923 ImakesummariesofinformationthatIhear orreadinEnglish. Cog �.97 0.9224 TounderstandunfamiliarEnglishwords, Imakeguesses. Comp 3.44 0.9225 WhenIcan’tthinkofawordduringa conversationinEnglish,Iusegestures. Comp 3.65 �.�626 ImakeupnewwordsifIdonotknowthe rightonesinEnglish. Comp 2.23 �.��27 IreadEnglishwithoutlookingupeverynewword. Comp 3.07 �.0528 Itrytoguesswhattheotherpersonwillsay nextinEnglish. Comp 2.35 0.9929 IfIcan’tthinkofanEnglishword,Iuseaword orphrasethatmeansthesamething. Comp 3.8� 0.9430 ItrytofindasmanywaysasIcantousemyEnglish.Meta 2.60 �.0�3� InoticemyEnglishmistakesandusethat informationtohelpmedobetter. Meta 3.37 �.0�32 IpayattentionwhensomeoneisspeakingEnglish. Meta 3.60 0.9833 ItrytofindouthowtobeabetterlearnerofEnglish. Meta 2.73�.0734 IplanmyschedulesoIwillhaveenoughtime tostudyEnglish. Meta 2.3� 0.8935 IlookforpeopleIcantalktoinEnglish. Meta 2.�9 �.0336 Ilookforopportunitiestoreadasmuchas possibleinEnglish. Meta 2.50 0.9737 IhavecleargoalsforimprovingmyEnglishskills. Meta 2.94 �.2938 IthinkaboutmyprogressinlearningEnglish. Meta 3.09 �.0439 ItrytorelaxwheneverIfeelafraidofusingEnglish.Aff 2.80 �.0740 IencouragemyselftospeakEnglishevenwhen Iamafraidofmakingamistake. Aff 3.07 �.�64� IgivemyselfarewardortreatwhenIdowell

Page 9: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

4040

JALT JournAL

inEnglish. Aff 3.43 �.0942 InoticeifIamtenseornervouswhenIam studyingorusingEnglish. Aff 3.08 �.�643 Iwritedownmyfeelingsinalanguagelearningdiary. Aff �.480.8644 ItalktosomeoneelseabouthowIfeelwhen IamlearningEnglish. Aff �.99 0.9945 IfIdonotunderstandsomethinginEnglish,Iask theotherpersontoslowdownorsayitagain. Soc 4.�4 0.8846 IaskEnglishspeakerstocorrectmewhenItalk. Soc 2.65 �.�947 IpracticeEnglishwithotherstudents. Soc 2.24 �.0�48 IaskforhelpfromEnglishspeakers. Soc 2.69 �.2449 IaskquestionsinEnglish. Soc 2.44 �.0950 ItrytolearnaboutthecultureofEnglishspeakers. Soc 3.03 �.2�

Note:ThestatementforeachitemisintheEnglishoriginalfromwhichtheJapanesetranslationwasmade.KeyforStrategyType:Memo=Memory,Cog=Cognitive,Comp=Compensation,Meta=Metacognitive,Aff=Affective,Soc=Social

Table2:MeanScoresandStandardDeviationsfortheItems,TimeTwo(n=�53)

Item M SD Item M SD Item M SD Item M SD

� 2.99 �.�4 �6 3.95 0.89 3� 3.39 �.06 46 2.08 �.042 3.25 �.�3 �7 2.5� �.24 32 3.�5 �.04 47 2.97 �.223 3.52 0.90 �8 3.24 �.�7 33 2.88 �.07 48 3.47 �.044 2.36 0.82 �9 2.22 �.06 34

Table3:DescriptiveStatisticsfortheSILLandSubsections,TimesOneandTwo(n=�53)

Measure M SD Min Max Range Skew

SILL,TimeOne �39.00 24.60 66 207 �4� -.24Memo,TimeOne 22.64 4.66 �� 36 25 -.04Cog,TimeOne 39.95 7.65 �8 6� 43 -.20Comp,TimeOne �8.33 3.60 8 28 20 -.04Meta,TimeOne 25.03 6.34 9 40 3� -.�6Aff,TimeOne �5.84 3.64 7 27 20 .07

Page 10: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

4�4�robson & MidorikAwA

Table4givesthereliabilityforthesixpartsandtheoverallreliabilityforbothadministrations.WhiletheSILLasawholeforbothtimesoneandtwohasveryhighreliabilityat.93,severalofthesubsectionsarevery low. Inparticular, theTimeOne reliabilities forMemo,Comp,andAffareunacceptablylow.ThesameistrueforMemo,Comp,Aff,andSocinTimeTwo.Theresultsforthesecondmeasureofreliability,test-retest,areshowninTables5and6.Thedegreeofsharedvariancefortheitemsdoesnotexceed46percentwithsomeaslowas3,4,5,and7percent.Theaverageforalltheitemsisjust�9.5percent.Forthesubsections,thesharedvarianceissimilarlylow,withtheonlyexcep-tionbeingfortheSILLasawholeat58percent.

Soc,TimeOne �7.�8 4.89 6 30 24 .��SILL,TimeTwo �44.58 25.22 63 229 �66 -.26Memo,TimeTwo 26.67 4.94 �� 4� 30 -.29Cog,TimeTwo 37.84 7.9� �6 66 50 .�3Comp,TimeTwo �8.85 3.4� 8 27 �9 -.42Meta,TimeTwo 26.2� 5.7� 9 44 35 -.�6Aff,TimeTwo �7.45 3.73 6 27 2� -.38Soc,TimeTwo �7.54 3.65 6 26 20 -.25

KeyforStrategyType:Memo=Memory,Cog=Cognitive,Comp=Compensation,Meta=Metacognitive,Aff=Affective,Soc=Social

Table4:InternalConsistencyfortheSILLandSubsections,TimesOneandTwo(n=�53)

Measure Alpha Measure Alpha

SILL,TimeOne .93 SILL,TimeTwo .93

Memo,TimeOne .63 Memo,TimeTwo .66

Cog,TimeOne .80 Cog,TimeTwo .83

Comp,TimeOne .67 Comp,TimeTwo .58

Meta,TimeOne .85 Meta,TimeTwo .79

Page 11: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

4242

JALT JournAL

Table5:PercentageofSharedVarianceBetweenSILLItems,TimesOne&Two(n=�53)

Items RSquared Items RSquared

� .03 26 .282 .�6 27 .�03 .�6 28 .�04 .2� 29 .�45 .�8 30 .246 .07 3� .�87 .�8 32 .�88 .07 33 .249 .�3 34 .�8�0 .�8 35 .4��� .34 36 .25�2 .26 37 .42�3 .�2 38 .�6�4 .29 39 .24�5 .29 40 .22�6 .29 4� .34�7 .27 42 .08�8 .�4 43 .2��9 .�4 44 .0520 .�4 45 .�42� .27 46 .4622 .07 47 .2423 .�� 48 .2824 .07 49 .0725 .36 50 .04

Table6:PercentageofSharedVarianceBetweentheSILL&Subsections,TimesOne&Two(n=�53)

Measure RSquared

Memo .25Cog .36Comp .�4Meta .35Aff .�7Soc .26SILL .58

KeyforStrategyType:Memo=Memory,Cog=Cognitive,Comp=Compensation,Meta=Metacognitive,Aff=Affective,Soc=Social

Page 12: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

4343robson & MidorikAwA

Tables7and8showtheresultsofthefirstPCAswitha�5-factorso-lutionforTimeOneanda�3-factorsolutionforTimeTwo.Wewouldexpectthefactoranalysistogroupitems�through9inonefactor,items�0through23inasecondfactor,items24through29inathirdfactor,items30through38inafourthfactor,items39through44inafifthfac-tor,anditems45through50inasixth.However,theresultsfortheTimeOnePCAshowveryfewitemsloadingtogether.Thegreatestgroupofitemsloadingtogetherisinfactor�4withitems46through49together;however,beyondthis,therearenogreatergroupsofloadingsthanjusttwoorthreeitemstogether.Factoronetakesup23percentofthetotalvariancewiththeotherfactorsaccountingforconsiderablyless,whichisconfirmedbytheeigenvalues.Inaddition,thecommunalities,whichshowthedegreetowhichthefactorsareaccountingforeachitem,arenotparticularlyhighexceptforitems24,25,35and36.AsimilarstateofaffairsisfoundfortheTimeTwoPCA;however,therearenogroupsofloadingsgreaterthantwo,makingtheresultsappearevenlesssystem-aticthanwiththoseoftheTimeOneanalysis.Again,almostallthetotalvarianceisbeingaccountedforbyfactorone.Also,withtheexceptionsofitems�3and�4,thecommunalitiesarenotparticularlyhigh.

Tables9and�0showtheattempttoforcetheSILLintoasix-factor

Table7:PrincipalComponentAnalysis,TimeOne(n=�53)

Item FactorLoadings Factor� Factor2Factor3 Factor4 Factor5Communalities

35 0.85 0.9�36 0.85 0.9�23 0.64 0.4730 0.60 0.5850 0.44 0.6�32 0.72 0.5729 0.66 0.5�4� 0.59 0.533� 0.56 0.6240 0.53 0.5839 0.45 0.5020 0.70 0.579 0.63 0.35�9 0.58 0.442� 0.39 0.458 0.79 0.4934 0.49 0.54�6 0.45 0.57Table7cont...

Page 13: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

4444

JALT JournAL

Item FactorLoadings Factor� Factor2Factor3 Factor4 Factor5Communalities

44 0.79 0.38� 0.32 0.38

Eigenvalues ��.54 3.�9 2.40 2.02 �.98PercentofTotalVariance 23.08 6.37 4.8� 4.04 3.95

Item FactorLoadings Factor6 Factor7Factor8 Factor9 Factor�0Communalities

24 0.9� 0.9�25 0.9� 0.9�27 0.43 0.5926 0.79 0.3933 0.63 0.5338 0.56 0.5537 0.52 0.5��2 0.38 0.53�0 0.78 0.3845 0.4� 0.496 0.79 0.4342 0.53 0.35

Eigenvalues �.69 �.60 �.45 �.33 �.22PercentofTotalVariance 3.39 3.20 2.9� 2.66 2.45

Item FactorLoadings Factor�� Factor�2 Factor�3 Factor�4 Factor�5Communalities

�8 0.68 0.4722 0.65 0.50�3 0.37 0.453 0.74 0.464 0.58 0.5�7 0.46 0.4�5 0.46 0.44�7 0.73 0.49�4 0.59 0.65�5 0.58 0.42�� 0.48 0.5948 0.70 0.5746 0.64 0.6547 0.52 0.6343 0.52 0.4649 0.5� 0.6�Table7cont...

Page 14: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

4545robson & MidorikAwA

28 0.74 0.422 0.47 0.36

Eigenvalues�.�9 �.�8 �.�2 �.08 �.04PercentofTotalVariance2.38 2.36 2.24 2.�6 2.09

Note:Onlyitemswithloadingsequaltoorover0.30areindicatedinthetable

Table8:PrincipalComponentAnalysis,TimeTwo(n=�53)

Item FactorLoadings Factor� Factor2 Factor3 Factor4 Factor5Communalities

�7 0.72 0.6635 0.7� 0.7626 0.65 0.594� 0.6� 0.5630 0.58 0.52�9 0.53 0.5522 0.46 0.48�2 0.43 0.4�8 0.39 0.5�2� 0.38 0.549 0.69 0.5838 0.64 0.4844 0.62 0.4940 0.56 0.5437 0.55 0.56�0 0.54 0.5�32 0.49 0.55�� 0.48 0.5542 0.66 0.6�28 0.65 0.53�6 0.63 0.6245 0.63 0.4729 0.54 0.5�5 0.38 0.42�5 0.67 0.4247 0.59 0.533� 0.5� 0.55� 0.5� 0.533 0.46 0.494 0.38 0.37�3 0.85 0.9��4 0.85 0.9�

Eigenvalues ��.94 3.02 2.70 2.09 �.89PercentofTotalVariance 23.88 6.04 5.4� 4.�8 3.79

Page 15: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

4646

JALT JournAL

Table8cont...Item FactorLoadings Factor6 Factor7 Factor8 Factor9 Factor�0Communalities

2 0.76 0.4236 0.45 0.5346 0.85 0.48�8 0.73 0.5�20 0.52 0.6624 0.49 0.4725 0.42 0.4643 0.77 0.4650 0.49 0.4�49 0.39 0.5�6 0.62 0.387 0.43 0.42

Eigenvalues �.58 �.46 �.39 �.36 �.20PercentofTotalVariance 3.�6 2.9� 2.79 2.73 2.40

Item FactorLoadings Factor�� Factor�2 Factor�3 Communalities

48 0.72 0.3739 0.43 0.6434 0.37 0.4923 0.75 0.5333 0.40 0.7027 0.72 0.3�

Eigenvalues �.�6 �.�0 �.05PercentofExplainedVariance 2.3� 2.20 2.09

Note:Onlyitemswithloadingsequaltoorover0.30areindicatedinthetable.

solution.Herewehaveaclearerpictureofwhythefirstfactoristak-ingupsomuchofthetotalvariance,althoughwithTimeTwo,thereislessofaconcentrationofitemsinthefirstfactor.Nonetheless,theloadingsforbothPCAsshowacombinationofrelatedandunrelateditems fromthesixsubgroups loading together.Figures�and2givevisualrepresentationsoftheeigenvaluesthroughscreeplots,which,ifwecountthenumberoffactorstotheleftofthepointwherethelineturnsstronglytotheright,seemtoindicatethataonefactoranalysisoftheSILLwouldbemostappropriate.

Theinterviewsrevealedsomeveryinterestingproblemsthequestion-

Page 16: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

4747

Table9:PrincipalComponentAnalysiswithSixForcedFactors,TimeOne(n=�53)

Item FactorLoadings Factor� Factor2 Factor3 Factor4 Communalities

35 0.79 0.9�36 0.79 0.9��4 0.69 0.6547 0.68 0.6449 0.68 0.6�30 0.67 0.5946 0.67 0.6548 0.60 0.5723 0.55 0.47�7 0.5� 0.4950 0.49 0.6240 0.47 0.5828 0.44 0.42�6 0.43 0.574 0.42 0.5��5 0.39 0.42�3 0.35 0.4526 0.3� 0.3920 0.72 0.572� 0.66 0.45�9 0.59 0.4422 0.49 0.50�8 0.44 0.4739 0.43 0.509 0.43 0.357 0.40 0.423 0.39 0.465 0.35 0.4432 0.69 0.57�� 0.62 0.5945 0.57 0.49�2 0.56 0.5329 0.55 0.5�38 0.5� 0.5533 0.5� 0.533� 0.49 0.6237 0.44 0.5��0 0.39 0.388 0.69 0.4934 0.49 0.532 0.48 0.366 0.44 0.43

Eigenvalues ��.54 3.�9 2.40 2.02PercentofTotalVariance 23.09 6.37 4.8� 4.04

Page 17: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

4848

JALT JournAL

Table9cont...Item FactorLoadings Factor5 Factor6 Communalities

43 0.53 0.4644 0.53 0.38� 0.43 0.384� 0.37 0.5324 0.86 0.9�25 0.86 0.9�27 0.44 0.50

Eigenvalues �.98 �.69PercentofTotalVariance 3.96 3.39

Note:Onlyitemswithloadingsequaltoorover0.30areindicatedinthetable.

Table�0:PrincipalComponentAnalysiswithSixForcedFactors,TimeTwo(n=�53)

Item FactorLoadings Factor� Factor2 Factor3 Factor4 Communalities

38 0.63 0.4844 0.58 0.4932 0.57 0.5539 0.57 0.6437 0.57 0.563� 0.56 0.5524 0.53 0.4740 0.52 0.544 0.47 0.3725 0.47 0.46�5 0.40 0.4235 0.68 0.7622 0.66 0.4826 0.65 0.59�7 0.59 0.66�3 0.57 0.9��4 0.57 0.9�6 0.54 0.387 0.5� 0.424� 0.5� 0.5630 0.5� 0.5233 0.44 0.7023 0.39 0.532� 0.39 0.54�0 0.38 0.5��6 0.72 0.6242 0.65 0.6�

Page 18: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

4949robson & MidorikAwA

Table�0cont...20 0.65 0.6628 0.64 0.5329 0.6� 0.5�45 0.49 0.47�8 0.47 0.5�5 0.47 0.423 0.37 0.4934 0.37 0.4947 0.59 0.5349 0.57 0.5�36 0.57 0.53� 0.52 0.5348 0.49 0.3750 0.48 0.4�8 0.44 0.5��2 0.32 0.4�

Eigenvalues ��.94 3.02 2.70 2.09PercentofTotalVariance 23.88 6.04 5.4� 4.�8Item FactorLoadings Factor5 Factor6 Communalities

9 0.62 0.58�� 0.58 0.55�9 0.56 0.5543 0.39 0.462 0.69 0.4227 -0.52 0.3�46 -0.42 0.48

Eigenvalues �.89 �.58PercentofTotalVariance 3.79 3.�6

Note:Onlyitemswithloadingsequaltoorover0.30areindicatedinthetable.

naireposedfortherespondents.Themajorityofstudentsinterviewedhaddifficultyunderstandingitems�,5,6,7,�4,�9,20,22,26,43,44,and47.ThemostcommonlycitedreasonfortheirlackofunderstandingwasunfamiliarJapaneseorEnglishexpressions.Thiswasparticularlytrueforitems5,6,22,and43.Anotherreasonrespondentsgavefortheirdifficultyinunderstandingwasthattheycouldnotimaginethesituation.

Discussion

TheresultsreportedaboveprovideahighlevelofreliabilityfortheSILL

Page 19: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

5050

JALT JournAL

Figure�:ScreePlot,PrincipalComponentAnalysis,TimeOne(n=�53)

Figure2:ScreePlot,PrincipalComponentAnalysis,TimeTwo(n=�53)

Page 20: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

5�5�robson & MidorikAwA

asawhole,whichisproblematic,astheSILLshouldbemeasuringsixdifferent typesof strategies,notonegrand strategy type.Moreover,thealphasforthesubsectionsshowasimilarlylowlevelofreliabilityaswasfoundbyBrownetal.(�996).Onereasonforthiscouldbethenumberofitemsinthesubsections,wherethelongersubsectionssuchasCognitiveStrategieshavehigherreliability.Lengthisanimportantfactorinreliability,aslongermeasurestendtobemorereliable(Bach-man,�990).Moreover,aswasnotedpreviously,allthedistributionsareskewed,whichmustalsobeaffectingthelevelofreliability.Forexample,SocialStrategiesTimeOnehasfairlyhighreliability,butatTimeTwoitdropsto.59.However,thereisalsoanincreaseintheskewbetweentimesone(.��)andtwo(-.25).Nevertheless,theseskeweddistributionscannotfullyexplaintherelativelylowreliabilityastheCognitiveStrate-giessubsectionhasaconsistentlevelofreliabilityfromTimeOnetoTimeTwo,butskeweddistributionsof-.20and.�3.Thetest-retestreli-abilityasindicatedbythepercentageofsharedvariancefortheitems,subsectionsandentiremeasureshowthattheSILLishighlyunreliable.Itisimportanttorememberthatreliabilitycanbemeasuredseveraldif-ferentways,andthatdependenceonasingleapproachcanberisky.Onereasonforthelowfigureshasbeenfoundinotherstudieslookingateitherbeliefsorstrategies(forexampleGaies&Sakui,�999),wherethestudentswerefoundtochangeovertime.Althoughitisdifficulttodeterminetheexactreasonsforchangewithoutconductingextensivepost-administrationinterviews,studentsmayinterpretthequestionsonagivenmeasureinlightoftheircurrentlearningsituationandnotlearn-ingsituationsingeneral.Moreover,theeffectsoftrainingandlearningmustalsobetakenintoaccount.Inaddition,itisimportanttoremem-berthatstrategiesarenotpersonalitytraits,whichhavebeenshowntoremainstableovertimeandacrosssituations(seeAngleitner,�99�).Thus,itishardlysurprisingthatthepercentageofsharedvarianceshouldbesolowbetweenthetwoadministrations.However,thereareotherpossibleexplanationsforthelowlevelsofreliability.Again,theskeweddistributionscouldbeadverselyaffectingtheresults,oritispossiblethatthepopulationsurveyedwastoohomogeneous.Thesubjectsareallfromasinglelanguagebackgroundandculturewithclosesimilaritiesinageandpossiblyeducationalexperience.Theskeweddistributionsarelikelypartoftheexplanation.Nonetheless,theJapaneseversionoftheSILLwasdesignedtoexaminejustthistypeofpopulation.Moreover,theeducationalbackgroundofthisgroupofsubjectsisprobablynotallthathomogeneous.Thestudentsatthewomen’suniversitycomefromawideareanorthandwestofTokyoandattendedbothprivateandpublichighschoolswherethereareeducationaldifferencesfromone

Page 21: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

5252

JALT JournAL

schooltothenext.Theco-edschoolsubjectsaresimilarinthisregard.ItalsoseemsreasonabletoexpectthatmostadministratorsorteacherswillusetheSILLundersimilarconditionsinJapan.

ThefactoranalysisresultsdonotconfirmOxford’ssixstrategycat-egoriesevenwhenattempting to force theanalysis intoa six-factorsolution.Infact,theSILLiseithermeasuring�5or�3differenttypesofstrategies,orevenjustoneasindicatedbytheeigenvaluesandscreeplots.Thereareanumberofpotentialreasonsforthis.Thelowreli-abilityisanimportantfactorasisthesizeofthepopulation.HatchandLazaraton(�99�)recommendatleast35subjectspervariableforPCA,whichinthisstudywouldnecessitateannsizeofabout�,750subjects.Withasamplesizeofonly�53,thereisconsiderablelossofstatisticalpower.Nonetheless,other studieswith larger sampleshave shownsimilarresults(Hsiao&Oxford,2000)andbasedonthosefoundhereaswellasinBrownetal.(�996),itwouldseemsafertolimittheSILLtoonegrandlanguagelearningstrategiesfactorinsteadoftryingtobreakitintotheorizedgroups.

Attemptingtolabeleachofthesestrategytypesisverydifficult.Thereseemstobealmostnosystemtothefactorloadings,although,someofthefactorscanbetentativelylabeled.Forexample,TimeOnefactor�4seemstoberelatedtoOxford’sSocialStrategies,whilefactor2con-tainsitemsfromtheAnalyzingandReasoningsubgroupwithinCogni-tiveStrategies.Factor�2seemstobetheMemoryStrategysubgroupApplyingImagesandSounds.ThefactorsolutionforthesecondtimeshowsanevengreatermixingofitemsfromdifferentstrategygroupsalmostnecessitatingacompleteabandonmentofOxford’scategories.However,bylookingatthewording,wecanapplytentativelabels.Forexample,factortwocanbeinterpretedasvariousspeakingstrategies.Inaddition,thereseemtobegroupingsofitemsinbothTimeOneandTimeTwobasedonthetypeofactionexpressedbytheverbinJapa-nese.AnexamplewouldbeTimeOnefactorfour,wherethesubjectsseemtoplaceemphasisonsuchactionsas“review,”“read,”and“plan.”TheattempttoforcetheSILLintosixfactorsforTimeOneresultedinwhatlookslikeaone-factorsolutionincludingsomeitemsfromeachsubsection.IftheseresultshadbeenrepeatedinTimeTwo,therewouldhavebeenanopportunitytosupportaone-factorsolutionbasedonthisdata.Unfortunately,theitemsinthefirstfactordiffer.

Theproblemsstudentshadunderstandingthequestionnairewerepartiallyrevealedbythepost-administrationinterviewsconductedwithaverysmallsample.Thesestudentswereunfamiliarwithsuchexpres-sionsorsituationsaskokoroniegaku(makingamentalpicture)initem4,inotsukau(userhymes)initem5,“flashcards”initem6,andkarada

Page 22: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

5353robson & MidorikAwA

dehyエnshite(physicallyactout) in item7.Forexample,duringtheadministrationofthequestionnaire,themajorityofstudentscouldnotreadthecharacterin,whichmeansrhymeinJapanese,anddidnotknowitsmeaningwhenitwasreadtothem.Moreover,otheritemsthatwereincomprehensiblewereonesthatreflectamoreWesternapproachtolearningstrategiesthanonewithwhichJapanesestudentsarefamiliar,suchaswithitems4and7.Inaddition,studentshaddifficultyrelatingmanyof the situationspresented in thequestionnaire to theirownlearning.First,thequestionnairepresumesanESLlearningsituation,wherethesituationinJapanisclearlyEFL.Thus,thesestudentshavefewopportunitiesfortargetlanguageuseoutsideoftheclassroom.Ofthe�0interviewees,3hadexperiencedstudyinginanEnglish-speak-ingcountryandhadfewcomprehensionproblemswiththelearningsituationspresented.However, for the remaining7 students, targetlanguagestudyandusewaslimitedtotheclassroom,library,home,train,ortheirEnglishSpeakingSociety(ESS)meetingsandtheyfoundmanyofthelearningsituationsinthequestionnaireunimaginableorstrange.Moreover,aswasnotedabove,theintervieweesrespondedtoitemsbasedontheircurrentlearningsituationandnotlearningsitua-tionsingeneral.

Conclusion

Thesimplestconclusiononecandrawfromthisinitialattemptatde-terminingthereliabilityandvalidityoftheJapaneselanguageversionoftheSILListhatitisneitherreliablenorvalidbasedonthisstudentsample.AlthoughtheSILLhasshownahighdegreeofinternalreliabilityfortheentirequestionnaire,itclaimstomeasuresixdifferentstrategiesandthusmustbeanalyzedassixdifferentmeasures.Infact,thehighdegreeofreliabilityfortheentireSILL,asnotedabove,isnotnecessarilyagoodthing.Thesubsectionshaveagenerallylowandunacceptablealphalevel.Moreover,thereareseriousquestionsabouthowreliabletheresultsarewhengiventothesamegroupmorethanonceandhowvalidthecategoriesusedtogrouptheitemsonthequestionnaireare.Inotherwords,whiletheSILLmayindeedbemeasuringlanguage-learn-ingstrategies,itdoesnotseemtobemeasuringgroupsofstrategiesinthemannerOxfordhasclaimed,atleastfortheselearners.Itwouldseemreasonable,basedonthehighreliabilityfortheentireSILL,theeigenvalues,andscreeplots, todescribe theSILLasageneralmea-sureoflanguage-learningstrategiesandnotameasureofsixdifferentstrategytypes.Theresearchersbelievethattheseconclusionscanbedrawnbasedonthesedatainspiteofpotentialproblemswithnsize,

Page 23: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

5454

JALT JournAL

thepossiblyhomogeneouspopulation,skeweddistribution,andlowreliability.

Asdiscussedpreviously, themethodsOxfordused tovalidateanearlierversionoftheSILLaresomewhatsuspect.Takentogetherwiththelackofestablishedreliabilityorvalidityforthelaterversions(de-spiteclaimsbyYamato,2000,p.�42,tothecontrary),thoseusingtheEnglishversionoftheSILLwillnotbeabletorelyontheresults.More-over,HsiaoandOxford’s(2000)confirmatoryfactoranalysisdoesnotprovidemuchconfidenceeither.CautionaryusebecomesevenmorenecessarywiththeJapanesetranslation,asitisnowanewquestion-naire thathasnotgone througha rigorous reliabilityandvalidationprocess.Theseissuesandproblemsarenotjustaboutstrategies,butrelatetoanyuseofaself-reportquestionnaire.Itshouldbeclearfromthisanalysisthatsimplytakingaquestionnaire,translatingitintoanotherlanguage,administeringittoagroupofstudentsandthenusingthere-sultsformakingeducationalpolicydecisionsareveryunwisepractices.Moreover,anyquestionnairemustreflecttheactuallearningsituationsofthetargetpopulation,theirstrategyuse,thetypeoflanguagewithwhichtheyarefamiliar,andanyculturaldifferencesthatmightaffecttheoutcome.TheresearchershopethatthroughthisinitialattemptatvalidatingOxford’squestionnaireotherresearchersandlanguage-teach-ingprofessionalswilltakeamorecautiousapproachtoquestionnaireuseandinterpretation.

AcknowledgementsThisisamuchrevisedandexpandedversionofapaperdeliveredatAILA99inTokyo.Theauthorswouldliketothankthetwoanonymousreviewersfortheirsuggestionsandvaluablecomments.

GordonRobsonisProfessoratShowaWomen’sUniversity.Hisresearchinterestsincludereadingstrategiesandindividuallearnerdifferences.

HidekoMidorikawa isProfessor at ShowaWomen’sUniversity.Herresearchinterestsincludereadingstrategiesandteachereducation.

ReferencesAllport,G.W.(�937).Personality:Apsychological interpretation.NewYork:

Holt,Rinehard&Winston.

Angleitner, A. (�99�). Personality psychology: Trends anddevelopments.EuropeanJournalofPersonality,5,�56-�7�.

Page 24: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

5555robson & MidorikAwA

Bachman,L.F.(�990).Fundamentalconsiderationsinlanguagetesting.Oxford:OxfordUniversityPress.

Brown, J.D. (�988).Understanding research in second language teaching:A teacher’s guide to statistics and researchdesign. London:CambridgeUniversity.

Brown,J.D.(�997).Statisticscorner,questionsandanswersaboutstatistics:Skewnessandkurtosis.Shiken:JALTtesting&evaluationSIGnewsletter,�(�),�6-�8.

Brown,J.D.,Robson,G.,&Rosenkjar,P.(�996).Themotivation,anxiety,learningstylesandproficiencyof Japanesecollege students.UniversityofHawai’iWorkingPapersinESL,�5(�),33-72.

Chapelle,C.(�994).AreC-testsvalidmeasuresforL2strategyresearch?SecondLanguageResearch,�0(2),�57-�87.

Gaies,S.,&Sakui,K.(�999).Symposiumonlearners’beliefsaboutlanguagelearning.Paperpresentedatthe�2thWorldCongressofAppliedLinguistics,Tokyo,Japan.

Gardner,R.C.,&Lambert,W.E.(�972).Attitudesandmotivationinsecond-languagelearning.Rowley,MA:NewburyHouse.

Griffee,D. (�999). Translatingquestionnaires fromEnglish into Japanese:Is itvalid?InA.Barfield,R.Betts, J.Cunningham,N.Dunn,H.Katsura,K.Kobayashi,N.Padden,N.Parry&M.Watanabe(Eds.),OnJALT98:Focusontheclassroom(pp.�76-�80).Tokyo:JapanAssociationforLanguageTeaching.

Guilford,J.P.(�940).AninventoryoffactorsSTDCR,manualofdirectionsandnorms.BeverlyHills:SheridanSupplyCo.

Guilford,J.P.&Fruchter,B.(�973).Fundamentalstatisticsinpsychologyandeducation.NewYork:McGrawHill.

Hatch,E.&Lazaraton,A.(�99�).Theresearchmanual:Designandstatisticsforappliedlinguistics.NewYork:NewburyHousePublishers.

Horwitz,E.K.,Horwitz,M.B.,&Cope,J.(�986).Foreignlanguageclassroomanxiety.TheModernLanguageJournal,70,�25-�32.

Hsiao,T.-Y.,&Oxford,R.L.(2000).Learningstrategyuseandtargetlanguages:confirmatory factor analysis across languages. Paper presented at theannualmeetingoftheTeachersofEnglishtoSpeakersofOtherLanguages,Vancouver,Canada.

McCrae,R.R.(�989).WhyIadvocatethefive-factormodel:JointanalysesoftheNEO-PIandotherinstruments.InD.M.Buss&N.Cantor(Eds.),Personalitypsychology:Recenttrendsandemergingdirections(pp.237-245).NewYork:Springer-Verlag.

Messick,S.(�989).Validity.InR.E.Linn(Ed.),Educationalmeasurement(3rded.,pp.�3-�03).NewYork:Macmillan.

Oxford,R.L.(�990).LanguageLearningStrategies:WhatEveryTeacherShouldKnow.NewYork:NewburyHouse.

Page 25: How Reliable and Valid is the Japanese Language …jalt-publications.org/files/pdf-article/jj-23.2-art2.pdfconcern in test development and use is demonstrating not only that test scores

5656

JALT JournAL

Oxford,R.L.(�994).Gengogakushusutoratejiiエgaikokugokyoushigashitteokankerebanaranaikoto[LanguageLearningStrategies:WhatEveryTeacherShouldKnow](M.Shishido&N.Ban,Trans.).Tokyo:Bonjinsha. (originalworkpublished�990)

Oxford,R.L.(�996).Employingaquestionnairetoassesstheuseoflanguagelearningstrategies.AppliedLanguageLearning,7,(�&2),25-45.

Oxford, R. L., Park-Oh,Y., Ito, S.,& Sumrall,M. (�993). Factors affectingachievementinasatellite-deliveredJapaneselanguageprogram.AmericanJournalofDistanceEducation,7,�0-25.

Takeuchi,O.(�993).AstudyoflanguagelearningstrategiesandtheirrelationtoachievementinEFLlisteningcomprehension.BulletinoftheInstituteforInterdisciplinaryStudiesofCulture,�0,�3�-�4�.

Watanabe,Y.(�990).Externalvariablesaffectinglanguagelearningstrategiesof JapaneseEFL learners: Effects of entrance examination, years spentat college/university, and stayingoverseas.Unpublishedmaster’s thesis,LancasterUniversity,Lancaster,U.K.

Yamato,R.(2000).Awarenessandrealuseofreadingstrategies.JALTJournal,22(�),�40-�64.

(ReceivedAugust5,2000;revisedMay3,200�)