Validity-and-washback-in-language-testing

://ltj.sagepub.com/content/13/3/241Theonlineversionofthisarticlecanbefoundat: DOI:10.1177/026553229601300302199613:241LanguageTestingSamuelMessickValidityandwashbackinlanguagetesting  Publishedby:    : : : : : WhatisThis? -Nov1,1996VersionofRecordatZHEJIANGUNIVERSITYonOctober14,2014ltj.sagepub.comDownloadedfromatZHEJIANGUNIVERSITYonOctober14,2014ltj.sagepub.comDownloadedfromValidityandwashbackinlanguagetestingSamuelMessickEducationalTestingServiceWashback,aconceptprominentinappliedlinguistics,referstotheextenttowhichtheintroductionanduseofatestinfluenceslanguageteachersandlearnerstodothingstheywouldnototherwisedothatpromoteorinhibitlanguagelearning.Someproponentshaveevenmaintainedthatatest’svalidityshouldbeappraisedbythedegreetowhichitmanifestspositiveornegativewashback,anotionakintotheproposalof’systemicvalidity’intheeducationalmeasurementliterature.Thisarticleexaminestheconceptofwashbackasaninstanceoftheconsequentialaspectofconstructvalidity,linkingpositivewashbacktoso-calledauthenticanddirectassessmentsand,morebasically,totheneedtominimizeconstructunder-representationandconstruct-irrelevantdifficultyinthetest.ThecurrenteducationalreformmovementintheUSAputsconsider-ablestockinthenotionthatperformanceassessments,asopposedtomultiple-choicetests,willfacilitateimprovedteachingandlearning(ResnickandResnick,1991;Wiggins,1989;1993).Someproponentsevenclaimthatperformanceassessments,especiallythosethatareauthenticanddirect,arelikelytobe’systemicallyvalid’inthattheyinduce’intheeducationsystemcurricularandinstructionalchangesthatfosterthedevelopmentofthecognitiveskillsthatthetestisdesignedtomeasure’(FrederiksenandCollins,1989:27).Akindrednotionprominentinappliedlinguistics,especiallyinBri-tain,iscalled’washback’,whichistheextenttowhichthetestinflu-enceslanguageteachersandlearnerstodothings’theywouldnotnecessarilyotherwisedo’(AldersonandWall,1993:117).Aswithso-calledsystemicvalidity,somewritersinvokethenotionof’wash-backvalidity’,holdingthatatest’svalidityshouldbegaugedbythedegreetowhichithasapositiveinfluenceonteaching(Morrow,1986).Intheassessmentofskills,testshavingbeneficialwashbackarelikelytobecriterionsamples.Thatis,inthecaseoflanguagetesting,theassessmentshouldincludeauthenticanddirectsamplesofthecommunicativebehavioursoflistening,speaking,readingandwritingofthelanguagebeingleamt.Ideally,themovefromlearningexer-cisestotestexercisesshouldbeseamless.Asaconsequence,foropti-malpositivewashbackthereshouldbelittleifanydifferencebetweenatZHEJIANGUNIVERSITYonOctober14,2014ltj.sagepub.comDownloadedfrom242activitiesinvolvedinlearningthelanguageandactivitiesinvolvedinpreparingforthetest.Althoughonlysparselyinvestigatedtodate,evidenceofwashbackistypicallysoughtintermsofbehaviouralandattitudinalchangesinteachersandlearnersthatareassociatedwiththeintroductionoftestshavingimportanteducationalconsequences(AldersonandWall,1993).WithrespecttoUSAeducationreform,amorestringentclaimhasbeenmadeinvolvingnotonlychangesinteacherandlearnerbehavioursbutalsoinlearneroutcomes.Towit,’evidenceforsys-temicvaliditywouldbeanimprovementin[thetested]skillsafterthetesthasbeeninplacewithintheeducationalsystemforaperiodoftime’(FrederiksenandCollins,1989:27).However,suchformsofevidenceareonlycircumstantialwithrespecttotestvalidityinthatapoortestmaybeassociatedwithpositiveeffectsandagoodtestwithnegativeeffectsbecauseofotherthingsthataredoneornotdoneintheeducationalsystem.Techni-callyspeaking,sucheffectsshouldnotbeviewedastestwashbackbutratherasowingtogoodorbadeducationalpracticesapartfromthequalityofthetest.Furthermore,atestmightinfluencewhatistaughtbutnothowitistaught,mightinfluenceteacherbehavioursbutnotlearnerbehaviours,ormightinfluencebothwithlittleornoimprovementinskills.Hence,washbackisaconsequenceoftestingthatbearsonvalidityonlyifitcanbeevidentiallyshowntobeaneffectofthetestandnotofotherforcesoperativeontheeducationalscene.Indeed,ifitexists,washback’islikelytobeacomplexphenomenonwhichcannotberelateddirectlytoatest’svalidity’(AldersonandWall,1993:116).Inanyevent,washbackisonlyoneformoftestingconsequencethatneedstobeweighedinevaluatingvalidity,andtestingconsequencesareonlyoneaspectofconstructvalidityneedingtobeaddressed.Neithertestingconsequencesingeneralnorwashbackinparticularcanstandaloneasastandardofvalidity.Hence,oneshouldnotrelyonwashback,withallitscomplexityanduncontrolledvariables,toestablishtestvalidity,Morrow(1986)andFrederiksenandCollins(1989)notwithstanding.Rather,onecaninsteadturntothetestpropertieslikelytoproducewashback-namely,authenticityanddirectness-andaskwhattheymightmeaninvalidityterms.Next,weexaminetheimplicationsofauthenticityanddirectn