311 Measuring M gVariatio oninPerceptiono ofAccepttability:A AMagnitu ude EstimattionInve estigation nofNethe erlandsa andBelgianDutch h LaurieeA.Stowe ,,RobertHartsuikerb,M MagdaDevossb,JackHoeeksemaa a UniversityofG Groningen,bGhentUnivversity a Abstrract Wereeportastudyyofthepercceivedaccepptabilityofthhreeconstruuctionswith variable formss:theformo oftheadjectiivesindefiniiteneuterno ounphrases,theproducttionofinterrruptions oftheeverbalclustterandthep positionofveerbalparticleesinverbclu usters.Theseconstruction nsdiffer infrequencybetw weennortherrnstandard DutchusediintheNetherlandsandtthestandard dvariant usedinBelgium.TTheperceptio onofacceptabilityfollow wedthefrequ uencyofprod duction. 1. IIntroduction nalongtheN NorthͲSouth haxisinDuttchwas Inthiispaper,theeperceptionofsyntactticvariation invesstigated. M Much has been b writteen on NortthͲSouth vaariation, ussing corporra (e.g. Nieuw wborg 196 68, De Cubber 1973 3, Grondelaaers, Speellman and Carbonez 2001), collecctionsofprroductiond datafromtaaperecordin ngsorelicittation(e.g. Pauwels,19 953;De Vrien ndͲDeMan, 1969;Stroo op,1970;D DeSchutter,1974;Gerrritsen,1991 1;Barbiers, Bennis, DeVogelaer,DeevosandVaanderHam m,2005).Thesestudiessthusfocussonproducctionof nusage.Reelativelylittllework variousstructureesandprefferencediffferencesas reflectedin been done on the isssue of whether percep ption of these structu ures show parallel has b prefeerencesbyeelicitinginfo ormantjudggments,with htheexceptionofHaesseryn(1990 0). IInthepreseentpaper,w weusethe technique ofmagnitudeestimationtogain amore refineed understtanding of NorthͲSoutth differencces in percception of acceptabilitty in a numb berofareassthathavebeenidentifiedintheliteratureassareasofreelativelysign nificant syntaacticvariatio on,namely(1)adjectivveͲnounagreeementinn nounphrasees;(2)interruption byno onverbalmaaterialinveerbclusterss;and(3)sp plittingofp particleverb bsinverbclusters. Sinceewewerep primarilyintterestedin thebasicd distinctionb betweenSo outhern(=B Belgian) and Northern ((=Netherlan ndic) Dutch, we avoid ded manipu ulating variiables that would nguish vario ous regionss within Beelgium from m one ano other, or reegions with hin the distin Netherlands. Ad dditionally since the Dutch spoken in thee southern provinces of the berofcharacteristicswithBelgian Dutch,wed decidednotttouse Netherlandsshaaresanumb m those pro ovinces in our o experim ments, in orrder to keeep the comparison particcipants from simpleandbinarry.Thus,wh heneverwespeakof“N NorthͲSouth hvariation,””weactuallyymean differences that are fo ound betweeen Belgian n Dutch and d the stand dard Dutch of the the d westeern,eastern nandnorthernprovinccesoftheNeetherlands. 312 2. SyntacticVariationbetweenNorthernandSouthernDutch Before turning to the experiments, it will be useful to take a brief look at the three syntacticvariablesthatwechosetostudy. 2.1 Adjectivenounagreement AdjectiveͲnounagreementinDutchis simpleon themorphologicallevel.Thereareonly twoformsoftheadjective,theshortoruninflectedformconsistingofjustthestemofthe adjective,andthelongorinflectedforminwhichaschwahasbeenaddedtothestem(in wordsendinginavowel,thisschwaisautomaticallydeleted).Thesyntacticconditioning ofagreementisabitmorecomplex,sinceitdependsontwofactors:nominalgenderand typeofdeterminer.StandardNorthernDutchonlyhastwogenders,neuterandcommon gender (because the original Germanic feminine and masculine genders merged in the early modern period). Standard Southern Dutch maintains a distinction between masculineandfeminineinitssystemofpronominalreference,butnotintheinflectional system.Moreover,thegenderdistinctionhascompletelydisappearedintheplural,where all nouns are treated as common gender, as will be evident from the examples in (1) below. In Northern Dutch, singular neuter nouns with an indefinite determiner (such as een ‘a, one’ or geen ‘no’) are preceded by shortforms,while allother combinations of nounanddeterminerrequirethelongformoftheadjective.Compare: (1) a. hetoudehuis [singular;neuter;definite] theoldhouse b. eenoudhuis [singular;neuter;indefinite] anoldhouse c. deoudeman [singular;commongender;definite] theoldman d. eenoudeman [singular;commongender;indefinite] anoldman e. deoudehuizen [plural;definite] theoldhouses f. geenoudehuizen [plural;indefinite] nooldhouses AsnotedinHaeserynetal.(1997,section6.4.1.3),southernDutchsometimesemploys theshortformindefiniteneuternounphrasessuchas 313 h hetbruinpaaard (2) a. tthebrownh horse b. h hetgroothu uis tthelargeho ouse beungramm maticalinNo orthernDuttch,whileth heyareapparently Theseeformsarejudgedtob gram mmaticalinSSouthernDu utchproducction.Howacceptabletheyareinperceptionis anoth hermatter.Inproductiionthestem mformoccu ursfairlyinfrequently. 2.2 IInterruption nswithintheverbalclu uster The second and d third con nstructions investigateed here havve to do w with the orrder of nterruptioneffectsinttheverbclu uster,in elementsinthe verbclusteer.Thefirstconcernsin h nonͲverbal elements appear w within the cluster. c Intterruption effects e have been which studiedbyagreeatmanylin nguists.InD Dutch,asin German,th hemainverrbanditsauxiliary orm an unin nterrupted sequence aat the end of a subord dinate clause, e.g. verbss tend to fo Everss (1975), Deen Besten ((1989), Haeeseryn (1990 0). An example can b be seen in ((3); the verbaalclusterisindicatedbyitalics. (3) o omdatdedamesdebrieveninhuntasha addengesto opt b becausetheeladiestheelettersintheirbagsh hadput ‘‘becausetheladieshad dputtheletttersintheirbags’ d orders, in n which non nverbal elem ments appeear in the m middle of th he verb Alternative word clusteeroccasionallyappear,,especially intheusaggeofSoutheernspeakerrs(Koelmans1965, Den Besten and d Edmondso on 1983, Haaegeman an nd Van Riem msdijk 1986 6), a pheno omenon ure as ‘Verb Projection Raising’ and elsewh here as referred to in tthe generattive literatu 3) writes in n this conn nection of ‘cluster ‘interrruption of the verb cluster.’ EEvers (2003 creep pers’.Comp pare: omdatdedamesdebrievenhaddeninhun ntasgesto opt (4) o b becausetheeladiestheelettersha adintheirbagsput ‘‘id’ mans(1965)notedthatnonverbalitemsthattformanidiomaticunit(e.g.keepaneye Koelm oninEnglishoriitsDutchveersioninhettooghoudeen)withtheemainverbarefarmorrelikely 314 to occur in the verb cluster than nonidiomatic nonverbal items. Koelmans’ observation was based on corpus research; it is one goal of our study to see if first, the difference between Northern and Southern variants of Dutch is reflected in the acceptability judgmentsofnativespeakersaswellandsecondiftheeffectofidiomaticitycanbeseenin perceptionandifsoitaffectstheSoutherngrouponly. 2.3. Splittingofparticleverbs Closely related to the previous construction is the splitting of particle verbs. Standard Dutch(bothnorthernandsouthern)typicallyallowstwooptionsforparticleverbsinthe verb cluster: the particle may either appear inside the cluster, typically adjoined to the mainverb,orwithanotherverbalelementintervening,leadingtoasplittingoftheverbͲ particlecombination. (5) …hoehetteamdewerkzaamhedenzouindelen. …howtheteamthetaskswouldapportion ‘…howtheteamwouldapportionthetasks.’ (6) …hoehetteamdewerkzaamhedenin zoudelen ‘id’ (7) …hoehetteamdewerkzaamhedenin zouwillendelen …howtheteamthetaskswouldwantͲtoapportion ‘…howtheteamwouldwanttoapportionthetasks’ VerbͲparticle combinations are actually special cases of the verbal idioms mentioned aboveinconnectionwithinterruptionsoftheverbalcluster.Whatisdifferentaboutthem is that particles are always acceptable as cluster creepers, both in the North and in the South.Whatwewouldliketoknow,however,iswhethertherearesignificantdifferences between sentences like (6) and sentences like (7) in terms of preference or relative acceptability.Thereareissuesastotherelativeacceptabilityofthepossiblepositionsof the particle (cf. Bennis 1992, Koopman & Szabolcsi 2000). Our primary concern in this study is whether there is a difference between the Northern and Southern participants with regard to the acceptability of the split version. De Cubber (1973) noted that nonͲ adjacent position is less common in Southern Dutch. However, we also manipulate the distancebetweentheparticleanditsverbbyvaryingthesizeoftheverbalclusterfrom twotothreeelements,whichmayservetoaccentuatedifferencesinpreferencebetween thesplitandnonͲsplitversionsoftheconstruction. 315 3.InvvestigatingVariationin nPerception onswhichw wearestudyyingneverd differto Letusemphasizeeatthispoiintthattheconstructio e that only one vversion of the construcction is useed in the So outh and only the the extent otherrintheNorrth.Bothinflectedanduninflected dversionso oftheadjecttiveareposssiblein theSSouthernvariant,buttheuninflectedformismorecommonthaninNorthernDutch. Likew wise,interru uptionoftheverbalclu uster,particcularlyinidiiomaticverbphrases,iismore comm mon in sou uthern Dutcch, although h still not particularly p frequent. Separation of the particclefromitssverboccurrsfairlyfreq quentlyinb bothvariantts,butisclaimedtobemore comm mon in Norrthern Dutcch (De Cubber, 1973). For this reeason, binaary grammaaticality judgm ments are p probably no ot sensitive enough to reveal the effects of interest: bo oth will frequ uentlybejud dgedgramm matical. To in nvestigate w whether subtle differeences in freequency in production n are mirro ored by differrences in aacceptabilityy in percep ption, it is important to find a way to allo ow the informantstoexxpresstheirrjudgmentssevenincaaseswherettheperceivveddifferencesare o 20) such aas a LikertͲscale, may not be smalll. NͲpoint scales (e.g. acceptablee from 1 to suitable for this, because they speccify the nu umber of d distinctions in advancce. Our experiments theerefore useed Magnitu ude Estimation (ME), a technique importeed into 996),inwhichparticipaantsgivearrelative linguisticsfrom psychophyssics(seeBardetal.,19 judgm mentofeacchstimulus incomparissontoarefferencestim mulus.Particcipantsare usually free to choose an arbitraryy value for the reference themseelves. This technique t h has the advantagethatttheparticip pantscanad ddasmanyyintermediaatepointsaasarenecesssaryin contrrast to a LLikert scale.. Importanttly, the tecchnique has been dem monstrated d to be sensitivetograd dationsinju udgmentsofftheaccepttabilityofseentences,yieldingdataathatis hatisrobusstenoughto oshowstatisticallysign nificantdiffeerences(e.gg.,Bard repliccableandth etal.,1996;Keller&Alexop polou,2001;Feathersto on,2005a,b b;seeSoracee&Keller2005for ouseMEinordertoco ompare anovverview).Assfarasweaareaware, ourstudyissthefirstto accep ptabilityjud dgmentsforrspeakersoftwodifferrentvariantssofonelan nguage.Wediscuss group pcompariso onswiththistechniqueelater. 4. EExperimenttalMethodss Participantss 4.1 P s North hern and So outhern Du utch speakeers participaated in all tthree experriments The same descrribedbelow w,whichwerreinterleavvedwitheacchotherand dactedasd distractersfo oreach otherr.Thegrou upofNorth hernDutchsspeakersco onsistedof 56firstyearstudentssatthe 316 FacultyofArtsoftheUniversityofGroningeninNortheastNetherlands(50F;6M;mean age=21.4;s.d.=4.71).FortheSouthernDutchgroup,56studentsofpsychologyatthe UniversityofGentinthenorthwestofBelgium(40F;16M;meanage=18.5;s.d.=2.04) participated. These participants were also selected forinclusionfrom a largergroup on the basis of being native speakers of Northern or Southern Dutch without extensive exposure to the other variant, based on a language experience questionnaire that they filled in. Some additional subjects were excluded to make the number of participants fromeachgrouponeachlistequal(seematerials). 4.2 Materials Theexperimentswereinterleavedinasingledatacollectionandservedasdistractersfor each other. For this reason, we will describe all three sets of materials together. Magnitude estimation encourages participants to judge relative acceptability, and it is important to include a number of very clearly ungrammatical items and very clearly acceptable items to calibrate the extremes of the scale. A description of fillers which servedtoachievethiswillbeprovidedattheendofthissection,aswilladescriptionof theallocationofthematerialsandfillerstoexperimentallists. 4.2.1 Experiment1:AdjectiveͲNounAgreement Weconstructed12sentencepairswhichcontainedaprepositionalphrasewithadefinite neutergendernounphrase,asin(8).6Thiscouldappearwithanadjectiveintheinflected form –e, which is standard in Northern Dutch, or without the inflection which is acceptable for a large number of adjective noun combinations in Southern Dutch. This variability between language variants is indicated by the question mark before the sentence.Thesentencesvariedsomewhatinstructure,buthadincommonthatthenoun 6 InflectionproductionpreͲtest.Anumberoffactorsseemtoaffectthefrequencyoftheuninflectedformwith definiteneutergendernounsinSouthernDutch,includingtheexactcombinationsofadjectiveandnoun.The major goal of the current experiment was to investigate the degree to which the perception of acceptability mirrorsthedegreetowhichtheformsareproduced.Forthisreason,wechosetomaximizethedegreetowhich Southernspeakersarelikelytoproducetheactualsequenceswhichweusedintheexperiment.Therefore,we conductedapretestinwhichwepresentedaquestionnairewithindefiniteneutergendernounphrases(e.g., eenbelangrijkgesprek),inwhichtheuninflectedformiscorrect,andaskedparticipantstoorallygeneratethe definiteform.Theythenfilledinwhethertheyhadusedthe–eformornot.Weusedcombinationsforwhichat least 70% of our sample of 10 had generated the uninflected form. Note that this pretest, if anything, may underestimate the frequency of the uninflected form, since in a conscious generation task like this one the uninflectedformmightbeunderreported. 317 phrassecontainin ngthetargeetadjective appearedinaprepositionalphrasepresenteedafter thein nflectedverrb. EErwerdentiijdenshetb belangrijke ggesprekeen npaarafsprrakengemaaakt (8) a. b. ??Erwerdenttijdenshetb belangrijk ggesprekeen npaarafsprakengemaaakt Lit. TTherewereduringtheiimportantcconversation nafewagreeementsmaade ‘A Afewagreeementsweremadedurringtheimp portantconvversation.’ EErwerdentiijdensdelan ngevergadeeringeenpaarpuntenbesproken (9) a. b. **Erwerdentijdensdela angvergadeeringeenpaarpuntenbesproken lit TTherewereduringthellongmeetin ngacoupleͲͲofpointsdiiscussed ‘A Acoupleofpointswereediscussedduringthelongmeetin ng.’ Particcipants mayy judge thee uninflecteed forms to o be gramm matical beccause they do not noticcethemissin ng–e.Toco ontrolforth his,twelvessentencepaairsasin(9))wereconsttructed h contained d a commo on gender rather than n a neuter gender ph hrase. The control which condition is also importan nt in our b betweenͲgro oups compaarison. Th he –e inflecction is consiidered to b be equally obligatory in these phrases in both b Northeern and So outhern Dutch h, which allowed us to assess th he relative aacceptabilitty of the un ninflected d definite comm mon gendeer form botth within aand across groups. Eaach de pairr was matcched in syntaacticstructu urewithaheetpairinorrdertoredu ucedifferencesinaccep ptabilityratingdue todiffferencesbeetweenconstructions. We predicted p th hat for Nortthern speakkers, there should be no difference in accep ptability betw ween the un ninflected n neuterͲ and common gender g con nditions (i.e. a main efffect of inflecctionandno ointeractio onwithgend der).ForSo outhernspeeakers,theu uninflected neuter gender conditio on should be more aacceptable than the uninflected u common gender nificantinteractionbetw weengendeerandinflecction). condition(asign 4.2.2 Experim ment2:Non nverbalInteerruptionsin ntheVerbC Cluster nwithnonvverbalmateerialinsidettheverbalcclusterisarcchaicinDuttch,but Thecconstruction particcularlyinidiomsthisw wordordertypeisstillo occasionallyyproducediinSouthernDutch. InNo orthernDutch,accordin ngtostandarddescrip ptionsinterrruptionbyn nonverbalm material (with htheexcepttionofpartticlesfromp particleverbs)isconsid deredungraammaticaleevenin idiom matic phrasees (Koelmans 1965, H Haeseryn 19 990). To tesst whether interruptio ons are relatiivelyaccepttableinidio omaticsenteences,wecconstructed12sentenccepairscon ntaining 318 commonidiomaticverbphrases(e.g.,(10),suchasuithetoogverliezen(loosetrackofor literallyoutoftheeyelose). (10) a. Datwarenkwestiesdiedemannenuithetoog hadden verloren b. Datwarenkwestiesdiedemannenhadden uithetoog verloren Lit. Thosewereissuesthatthemenhadouttheeyelost(losttrackof) ‘Thosewereissuesthatthegentlemanhadlosttrackof.’ (11) a. Datwarenbrievendiededamesinhuntas hadden gestopt b. *Datwarenbrievendiededameshadden inhuntasgestopt lit Thosewerelettersthattheladieshadintheirbagsplaced. ‘Thosewerelettersthattheladieshadputintheirhandbags.’ For each idiomatic item, a nonͲidiomatic sentence pair was constructed matched in sentence structure, like the example in (11), in order to test whether the asymmetry in acceptabilitybetweenidiomnaticandnonͲidiomaticsentencesisalsofoundinperceived acceptability. Based on Koelmans’ production data the version of these sentences with nonverbal interruptions are expected to be equally unacceptable regardless of idiomaticityfortheNortherngroups(maineffectofinterruptionandnointeractionwith idiomaticity),buttheyareexpectedtobemoreacceptableforSoutherndialectspeakers, particularly in he case of the idiomatic sentences (interaction between idiomaticity and interruption). 4.2.3 Experiment3:SplittingParticleVerbs Weconstructed24sentencesetslikethosein(12)containingparticleverbcombinations like indelen, which means apportion or, syntactically more apposite, carve up. These sentences occurred in four versions: one version with the particle and verb together in clauseͲfinal position in an embedded clause (12a), another version with the particle fronted to a position before a single auxiliary or modal in the verb cluster (12b), and similarversionsthatcontaintwoauxiliaryand/ormodalelementsinthecluster(e.g.12c and d). The occurrence of the particle both adjacent to the verb and separated from it occursinbothNorthernDutchandSouthernDutch,butthenonͲadjacentpositionappears to occur less frequently in Southern Dutch (De Cubber 1973). For the experimental sentencestherewassomevariabilityinthestructureofthemainclauseandtransitivityof the embedded clause, but all sentences contained an embedded clause with a particle verb. 319 (12) a. D Dechefvroeeghoehettteamdewerkzaamhedenzouindeelen. b. ??Dechefvro oeghoehettteamdewerkzaamhed denin zoud delen Lit TThemanageeraskedhow wtheteamthe tasksw wouldapportion ‘TThemanageeraskedhowtheteam mwouldapp portionthettasks.’ c. D Dechefvroeeghoehettteamdewerkzaamhedenzouwilleenindelen d. ??Dechefvro oeghoehettteamdewerkzaamhed denin zouw willendelen n Lit. TThemanageeraskedhow wthe teamthetasksw wouldwantttoapportion n. ‘TThemanageeraskedhowtheteam mwouldwan nttoapporttionthetaskks.’ his case it w was not po ossible to provide p con ntrol condittions which h both grou ups are In th hypothesizedtofindequalllyunaccepttableasinttheprevioussexperimen nts,soour control dtofindequ uallyaccepttable(thead djacent conditionisthefformwhichbotharehyypothesized ons). If De Cubber’s (1973) production dataa is correctt and is parrallel to perceived versio accep ptability, w we expect to t find thatt the Northern group p finds all versions off these senteencesequalllyacceptab ble.TheSou utherngrou upontheottherhandisexpected tofind both versionsoffthesenten ncewiththeparticlen notadjacentttotheverrblessacceeptable. Thevvariationin thecompleexityofthe verbalclusterwasincludedtoprrobethedegreeto which hsplittingttheparticle isaccepted d,sincewith halongerd distance,un nacceptabiliitymay bestronger. onstructionofFillersan ndLists 4.2.4 Co orts of fillerr sentences provided cclear extrem me values ffor the maggnitude A number of so mation scalee, with the simple can nonical intraansitive and d transitivee sentences at the estim highendoftheaacceptabilittyscale(14ͲͲ15)andtheemissingͲarrgument,exxtraͲargumeent,and without fronttedͲparticlesentencesaattheloweendofthesccale(16Ͳ18)).Sentencesswithandw te(19 9Ͳ20)wereo originallyintendedasp partofanaadditionalexperiment, butduetodesign probllemswillno otbediscusssedfurtherrhere.Additionally,weeusedasettofsentenccesthat proviidedtwofurtherpointssalongtheaacceptabilittyscale.Tofillinthemiddleoftheerange, senteenceswere includedco ontaininga shortnoun nphraseafteralongad dverbialphrase,in otherr words, ssentences with an u unacceptable ‘heavy NP shift’ (e.g., (21))). This theorretically prrovides a midway p point betw ween clearly grammaatical and clearly ungraammatical.Twelveothersentencees(e.g.,(22))containedappropriateeheavyNPshift(a longNPfollowin ngashortad dverbial).Th henumberofitemsofeachtypeo offillerisprrovided wingtheexaamplesgiveen. follow 320 (14) Depuberheefttegenzijnbroergejammerd 12 Lit. Theteenagerhadtohisbrothercomplained (15) Detennisserheeftdefinalebereikt 12 Lit. Thetennisplayerhadthefinalsreached (16) *Dejongenheeftzijnmoedergejammerd 12 Lit. Theboyhashismothercomplained 12 (17) *Derennerheeftindewedstrijdbereikt Lit. Therunnerhasinthecontestreached (18) *Deoppasrealiseerdedatweghetjongetjeliep 12 Lit: Thebabysitterrealizedthatawaythekidran (19) Demanbegontewerken 6 Lit: Themanbegantowork (20) *Demanbegonwerken 9 Lit: Themanbeganwork (21) ?Dejongenvondbijdeschoolwaarhijelkedagspeeldehetpoesj 12 Lit. Theboyfoundnexttotheschoolwhereheeverydayplayedthekitten (22) Demanzagindetuineenlange,slankedamegekleedinwittezijde 12 Themansawinthegardenatallslenderwomandressedinwhitesilk Topreventeffectsofhavingalreadyseenandratedanearlyidenticalsentence,versions oftheexperimentalsentenceswereallocatedtothefourlistsusingaLatinsquaredesign, so that subjects saw an equal number of each condition in an experiment (6) with no repeated items. When allocating experimental sentences to conditions on the lists, care wastakenthatthevariablesentencestructureswereevenlyspreadacrosstheconditions. Allfillersentencesappearedoneachofthefourlistsinordertoprovidethesamescale for magnitude estimation. Six blocks were made containing sentences of each type of filler sentence and one sentence of each condition of each experiment. Each block of sentenceswasthenrandomlyorderedandtheblockswereconcatenatedwitheachother toprovideanorderinwhichthesentencestypeswerethoroughlymixedandwereevenly spreadacrossthelist.Asecondversionofeachlistwasalsocreatedwiththesentencesin reversed order, giving eight versions of the experiment, each of which was rated by an equalnumberoftheparticipantsinthefinalanalysis.7 7 The versions presented in Ghent and in Groningen were almost identical, except that one filler sentencecontainedthewordhesp(‘ham’)intheSouthernversion,butham(‘ham’)intheNorthernversion,and in another two filler sentences, the neuter determiner was used in the Southern version and the common genderdeterminerintheNorthernversion. 321 4.3 P Procedure participantssfirstreadaawritteninstructionco oncerningthetasksandthencarrriedout Thep two versions o of the maggnitude esttimation taask. In the first version they became b acquaainted with h the conceept of doin ng magnitude estimattion on thee basis of aa short trainiing using lines and esttimating th he differencce in length h relative to o a referencce line. They were then n asked to ccarry out th he same co omparison b but now on n the basis of how h the senteences differr in accepttability from m a referen nce sentencce. The refference much senteence contained an inappropriate heavy NP shift like th he fillers w which served d as an intermediatepo ointonthe acceptabilittyscale(i.e.,example (21)).Thep participants gavea numeerical weigh ht to the reference seentence and then pro oceeded to estimate h how far awayyfromthisrreferenceo othersenten nceswere,o onascaleincludingbo othlessacceeptable ormo oreacceptaable.Thereferencesen ntencewas presentedtthroughouttthelistafteereach group p of ten sentences to help the participant to maintain the same relative scaale, and subjeects were in nstructed to o give the reference the t same rrating each time they saw it. Particcipantswho ofailedtodosowereeexcludedfro omtheanalyysis. 4.4 DataAnaalysisComm montoAllTh hreeExperim ments were free to o generate their own scale of accceptabilityy (some Because the participants w ome .001 to 1), we first normaalized the data usingg ZͲscores. In this used 1Ͳ4000, so mean and standard devviation weree calculated d over the data from all four calculation, a m experiments forr each subjeect, and each score was recoded as (Score –– Mean)/Sttandard his means the extrem me tails of each distriibution beccome comp parable, deviaation. By th 8 despiitetheactualnumbersusedforsccoring. A AfterrecodingthedatatoZͲscorees,themeaanoftherattingsforeaachconditio onwere calculatedforeaachsubjectandforeacchitem.Forreachexperiment,ANO OVAswerecarried out o over the paarticipant m means (F1) and item m means (F2).. Effects wiill be regarrded as signifficant if they are sign nificant at alphaͲlevel .05 in both the parrticipant an nd item analyysis,asboth hitemsandparticipanttsarerando omfactors (Clark,1973 3).Forthe sakeof readaabilitywed donotreportthefullsstatisticswh hentheFͲraatioisbelow w1andtheerefore 8 Weused thisprocedureeratherthantthemorecom mmontransform mationtolog scores(seeBaardetal., 1996),, because wee found that the groups used u sufficien ntly different scales that tthere were significant interacctions betweeen group and o other factors for virtually any set of cond ditions when the t log transfo ormation proced durewasused d,evenwhentthepatternoffdatawasiden ntical.Othereexperimentersshavereporteedsimilar difficulties with beetweenͲgroupss comparison ns (e.g., Feattherstone, 20 005b; Hopp, 2009). The ZͲscore transfo ormation app pears to be somewhat s mo ore reliable in n this respectt, although reemnants of tthe scale differeencescanbeseeeninthefigu uresbelow. 322 clearly not significant. In general, our interest is in the effects of the language variants usedbythetwogroups,andwewillthereforeconcentrateoninteractionsofotherfactors withgroup(althoughseefootnote3). 5. ResultsandDiscussion 5.1 Experiment1:AdjectiveͲNounAgreement:ResultsandDiscussion The results of the experiment in which we investigated the acceptability of uninflected adjectiveformsinneutergendernounphrasesareshowninFigure1below. BothgroupsshowedsignificantmaineffectsofGender[Southern:F1(1,52)=78.84, p<.001;F2(1,20)=14.58,p=.001;Northern:F1(1,52)=40.39,p<.001;F2(1,20)= 7.12,p<.05]andInflection[Southern:F1(1,52)=47.44,p<.001;F2(1,20)=46.90,p< .001;Northern:F1(1,52)=160.40,p<.001;F2(1,20)=129.23,p<.001].However,only the Southern group showed a significant interaction between Gender and Inflection [Southern:F1(1,52)=31.41,p<.001;F2(1,20)=28.76,p<.001;Northern:F1andF2< 1]. An overall interaction between Group (Ghent/Groningen), Gender, and Inflection [F1 (1, 104) = 13.76, p < .001; F2 (1, 20) = 8.48, p < .001] suggests that the two groups do indeeddifferinhowtheyreacttotheinflectionlessdefiniteneuter. Previous work (Haeseryn 1990) has suggested that the use of the uninflected form with neuter gender (het) words is relatively more frequent and acceptable in Southern thaninNorthernDutch.Theresultsofthecurrentexperimentclearlyconfirmthisclaim. TheSouthernparticipantsfindtheuninflectedformwithhetmuchmoreacceptablethan withde(p<.001forbothsubjectsanditems).Furthermore,theresultssuggestthatthe current generation of Southern speakers find the uninflected form approximately as acceptableastheinflectedformshowingvirtuallynodifferenceinacceptabilitybetween thetwovariants(neithersubjectsoritemssignificant.Theuninflectedformwithcommon gender de is regarded as quite unacceptable relative to the inflected form by both NorthernandSoutherngroups(p<.001forbothsubjectsanditems),sotheeffectcannot beexplainedasreadersmissingtheomittedͲe. 323 Figurre1:MeanZͲscoreratingsforsenttencesconttaininginflectedandun ninflected adjecctivesinneu utergender(het)andcommongen nder(de)deefinitenoun nphrases 0,6 0,4 0,2 uninfl 0 infl Ͳ0,2 Ͳ0,4 Ͳ0,6 de het South de het orth No Theseeresultsprrobablyoveerestimatettheacceptabilityoftheeomitted––e,sinceweechose lexicaal items thaat are produced frequently in this form. An interestingg follow up to this studyywouldbe tosystematicallyvary theproducctionfrequeencyforvarriouscombinations and see s to whaat extent th his correlatees with perrceived acceeptability fo or each seq quence. Howeever,notetthattheinflectedformsswerejudggedtobeeq quallyaccep ptable.Sincethese wereenottheforrmwhichw wasfrequenttlyproduceedinthepreeͲtest,they shouldhavvebeen judgeed to be lesss acceptab ble if the results are pu urely due to o the frequ uency of ind dividual comb binations. 5.2 EExperiment2: Verb bClusterIntterruptions::Resultsan ndDiscussio on bclusterintterruptionsinembeddeedclausesis Therresultsoftheexperimentwithverb show wninFigure2below. 324 MeanAcceptabilityRating(ZͲscore) Figure2:MeanZͲscoreratingsforembeddedclausescontainingidiomaticandnonͲ idiomaticphrasesprecedingorinterruptingverbalclusters 0,6 0,3 0 Ͳ0,3 idiomatic Ͳ0,6 nonͲidiomatic Ͳ0,9 Ͳ1,2 Ͳ1,5 interruption nointerruption interruption nointerruption South North TheSoutherngroupshowedasignificantInterruptionxIdiomaticityinteraction[F1(1,52) =53.54,p<.001;F2(1,20)=17.76,p<.001],aswellasmaineffectsofInterruption[F1 (1,52)=189.70,p<.001;F2(1,20)=133.14;p<.001]andIdiomaticity[F1(1,52)=34.47,p < .001; F2 (1, 20) = 175.34; p < .05]. The interaction reflects a significantly increased acceptabilityfor theinterruptioninidiomatic verbphrases(F1(1,52)=74.055,p<.001; t2(22)=5.460,p<.001),aswellasanonͲreliabletendencytowarddecreasedacceptability foridiomaticsentenceswithoutintrusion(F1(1,52)=3.773,p<.058;t2(22)=Ͳ711,p<.3). ThegroupfromGroningenalsoshowedasignificantInterruptionxIdiomaticityinteraction [F1(1,52)=44.82,p<.001;F2(1,20)=16.61,p=.001]andamaineffectofInterruption [F1(1,52)=822.52,p<.001;F2(1,20)=696.79,p<.001],buttheyshowednoeffectof idiomaticity [F1(1, 52) = 1.34, p > .25; F2 < 1 ]. As in the Southern group, idiomaticity modifiedtheacceptabilityofsentenceswithintrusions(F1(1,52)=24.883,p<.001;t2(22) =6.023,p<.001),andshowedatendencytodosoaswellforthosewithout(F1(1,52)= 15.43, p < .001; t2(22) = Ͳ1.788, p Ͳ .088). The two groups show essentially the same pattern,althoughitissomewhatmoreextremefortheNortherngroup. In earlier forms of Dutch nonͲverbal material within the verbal cluster was grammatical regardless of idiomaticity, but it is currently normally regarded as 325 ungraammatical in Standard d Netherlan nds Dutch aand is at leeast heavilyy nonͲprefeerred in South hernDutch..Fromcorp pusstudies((e.g.,Koelm mans,1965),,itappears thatinterru uptions oftheverbalclu usterinemb beddedclau usesoccurm morecomm monlyinSou uthernDutcchthan orthern Duttch, particularly in idio oms, although it remaiins infrequeent. Althou ugh the in No North herners’rattingsaremo orenegativeeforbothin nterruptionconditions,,itisdifficulttobe sure that this iss meaningfu ul. Overall,, both grou ups show th he same paattern, desp pite the hGroup.Additionally, theNortheerngroupteendstoshow wamoreeextreme interaactionswith scaleintheirjud dgmentsinggeneral,eveenintheno ormalizedZͲscoreratin ngs.Thisshowsup for cconditions which w are almost cerrtainly equally unacceeptable in both variants, cf. uninfflectedadjectivesincommongend dernounph hrasesinExp periment1above. W What we ccan conclud de is that the t Northerrn group also finds th he intrusion n more accep ptable for idiomatic phrases than n for nonͲid diomatic ph hrases, just like the So outhern group p.Sinceintterruptions havebeen reportedto ooccurmorrefrequenttlywithidioms,we alsop predictedth hatacceptabilitymighttfurtherbe qualifiedbyidiomaticity,particularlyfor the Southern ggroup. As suggested by Koelmans (1965)), idiomaticcity modifieed the ofverbalinttrusions.Ho owever,thissincreasedacceptabilityisnotlim mitedto unaccceptabilityo the SSouthern group, g as shown by the interacttions reporrted for bo oth groups above. Intereestingly,theereisatrendinthedataforidio omaticityto havethecconverseefffectfor bothgroupsasw well.Thesen ntencescon ntainingidio omaticverbphrasesweerelessacceeptable matic senten nces when the highlyy frequent continuou us order w was not than nonͲidiom employed,eventhoughitisthefrequentclearly grammaticaalorderforrthisconstrruction. hat the driive toward continuity in idiomattic phrasess is a stron ng one. This suggests th dtoreachsignificance intheitem manalysisw whichsuggesststhat Howeever,theefffectsfailed theeeffectmayb belimitedto ocertainitems. SplittingParticlleVerbs:ReesultsandD Discussion 5.3 EExperiment3: htheaccep ptabilityofffronting Figurre3displaysstheresultssofthethirrdexperimeent,inwhich a verrbal particlee to the po osition before auxiliaryy and modaal verbs in eembedded clauses waseexamined. Wevaried thenumberofverbclu usterelemeentsbetweeentheparticleand verb as well, to further pro obe the deggree to whiich splittingg the verb aand particlee is less ptable. accep A Analyses off the two groups sep parately do o not illuminate the crucial c inteeraction betw ween Group andParticlle Splitting [F1(1, 104) =14.26, p < .001;F2((1,20) =8.36; p < .01], sinceboth groupssho owsimilarpatterns.However,takkingthesho ortversionssalone, n between G Group and Splitting [FF1(1,104) = 19.860, p < .001; theree is a clear interaction F2(1,20)=4.606 6,p<.044],,whichrefleectsthefacctthatonlytheSoutherngroupsh howsan 326 effect of Particle Splitting when only one element intervenes between the verb and the particle [Southern: F1(1,52) = 34.093, p < .001; F2(1, 20) = 9.826, p = .005; Northern: F1(1,52)=2.162,p>1;F2(1,20)<1]. Figure3MeanZͲscoreratingsforsentencescontainingsplitandnonͲsplitparticles,with shortorlongverbalclusters 0,7 0,6 0,5 0,4 0,3 0,2 0,1 0 Ͳ0,1 long short South long split nonͲsplit short North Although nonͲsplit particle verbs are considered to be interruptions within the verb cluster,theyarenotconsideredlessgrammaticalthanthesplitvariantingeneral.Infact, they seem to be the preferred form. This is confirmed by the main effect of splitting shownbybothgroups.Ithasbeennoted(DeCubber1973)thatsplittingaparticlefromits verbwithintheverbalclusterislesscommoninSouthernDutchthaninNorthernDutch. We investigated whether this pattern can also be seen in the perceived acceptability of theseconstructions.Aspredicted,theeffectismoreobviousfortheSoutherngroupthan for the Northern group. This is confirmed by the significant interaction between Group andParticleSplittingreportedabove.9Theresultsdemonstratethatinperception,asin production, the language variants differ. However, even though less common in the 9 ThisresultcannotbeattributedtothegeneraltendencyoftheNortherngrouptoassignamoreextreme scaleintheirratings,becauseinthiscasetheyareconsistentlymorepositivethantheSoutherngroupfor thelesspreferredversion.Thisinteraction,then,canbeacceptedwithoutfurtherreservation. 327 south h, they rem main relativeely acceptab ble. Note tthat most of o the conditions in the table aboveareratedascomparaabletotheb bestalternaativesinExp periments1and2. nadditionalvariable,w wevariedthelengthoftheseparattionbetweeentheparticleand Asan itsveerbwithan additionalaauxiliaryor modalcom mponent.A Additionaldiistancediminished accep ptabilityinb bothgroupssascanbe seenintheesignificanttinteraction nbetweenP Particle andLLength[F1(1 1,104)=8.5 53,p<.001;;F2(1,20)=12.11,p<.01].Thisin nteractionb bolsters the conclusion c that in geeneral the nonͲsplit vversion is preferred, even though the prefeerenceislessspronounccedfortheN Northernsp peakers. GeneralD DiscussionaandConclussion 6. Inthiisstudy,weeusedexpeerimentalm methodstoiinvestigate threedifferrenceswhicchhave been claimed to o exist betw ween the No orthern and d southern Dutch language varian nts, the produ uction of uninflected aadjectives in definite neuter n gend der noun ph hrases, splittting of particcleverbs,andnonverb balinterrupttionsintheverbalclustterinembeeddedclausees.We have demonstraated that this sort of experimen nt can give insight into o the relationship ween frequency of vvarious constructions in produ uction and their perceived betw accep ptabilityinccomprehenssion. TTheresultssupporttheehypothesissthatwhen ntwoconstrructionsdiffferinfrequencyof produ uction, theey will also o differ in n perceived d acceptab bility. Fo or examplee, both Experiments2aand3 includedaless acceptable variant,bu utoneofth hosewas reelatively frequ uentforbotthvariants andtheoth herwasverryinfrequen nt.Thisiscclearlyrefleectedin theacceptabilityyratingsacrrossthetwo ogroups.M Moreimporttantly,wecaanindeedsseethat wogroupsd differinhow wacceptabletheyfind dsomeform ms.Thisisp particularly clearin thetw Experiments1and3.Theffactthatweecanseeth hesedifferencesbetweeengroups evenin method relatiivelysubtle differencesslikethediffferencein particlesplittingshowsthatthism hassomescopeforapplicattionintheffuture. H However fu uture experiments neeed to be concerned with spuriouss interaction ns with group p which maay be prod duced due to group d differences in the man nner in which the magn nitudeestim mationscaleeisapplied.Itisnotcleearwhythississoproblematic,buttgroup differrences havee emerged in several other studies as well (Feathersto one, 2005b;; Hopp, 2009). For single group studies, the logarithmicc scale is th he most common man nner of mparable,bu utthisleadsstoextremeinteractio ons.EvenusseofZͲ makingthejudggmentscom minatethe issue.Infutturestudiess,itmight bebetter b to ousea scoreesdoesnottotallyelim very extensiveLLikertͲscale((e.g.1to10 00;e.g.Hop pp,2009)w withanumberofexamp plesfor e values and in ntervening vvalues usingg fillers as in the currrent study, before the extreme initiatingratingssforthetarrgetitems. Thismayelliminatetheegroupdiffferencesand dmake 328 the ratings more reliable. Replications of the current studies might confirm that the apparentgroupdifferencesinExperiment1and2reallyexist.Despitethesedifficulties, elicitation of acceptability judgments seems suitable for further investigation of differences between language variants, particularly since it can be webͲbased, allowing large sets of data to be collected, adding to our arsenal of methods for investigating languagevariation. 7. References Barbiers,Sjef,HansBennis,GuntherDeVogelaer,MagdaDevos,MargreetvanderHam. 2005.SyntacticAtlasoftheDutchDialects.Vol.1:Pronouns,Agreementand Dependencies.Amsterdam:AmsterdamUniversityPress. Bard,EllenGurman,Robertson,Dan,&Sorace,Antonella(1996).Magnitudeestimation oflinguisticacceptability.Language72(1),32Ͳ68. Bennis,Hans(1992).‘LongHeadMovement:ThePositionofParticlesintheVerbalCluster inDutch.’In:LinguisticsintheNetherlands1992,ed.byR.BokͲBennemaandR.van Hout,,Amsterdam:JohnBenjamins,37Ͳ47. Besten,Hansden(1989).StudiesinWestͲGermanicSyntax.Doctoraldissertation,Catholic UniversityofBrabant. Besten,HansdenandJerroldA.Edmondson(1983).TheVerbalComplexinContinental WestGermanic.OntheFormalSyntaxoftheWestgermania,ed.byWerner Abraham,Amsterdam/Philadelphia:JohnBenjamins,155Ͳ216. Clark,HerbertH.(1973).ThelanguageͲasͲfixedͲeffectfallacy:Acritiqueoflanguage statisticsinpsychologicalresearch.JournalofVerbalLearningandVerbalBehavior, 12(X),335Ͳ359. DeCubber,W.1973,Desplitsingvanscheidbaarsamengesteldewerkwoordenin hedendaagsproza,StudiaGermanicaGandensiaXIV,39Ͳ63. Evers,Arnold(1975).ThetransformationalcycleinDutchandGerman.Bloomington: IndianaUniversityLinguisticsClub. Evers,Arnold(2003).VerbalClustersandClusterCreepers.In:VerbConstructionsin GermanandDutch,ed.byPieterA.M.SeurenandGerardKempen,Amsterdam:John Benjamins,43Ͳ89. Featherston,Sam.(2005a).Magnitudeestimationandwhatitcandoforyoursyntax: somewhͲconstraintsinGerman.Lingua115,1525Ͳ1550. 329 Feath herston,S.((2005b).Universalsand dgrammaticcality:WhͲcconstraintsiinGermanaand EEnglish.Ling guistics,43((4),667Ͳ711 1. Gerrrritsen,Marinel(1991).AtlasvandeeNederland dsedialectssyntaxis.Am msterdam:K KNAW. Grondelaers,Steefan,DirkSp peelmanandAnCarbonez.2001.R Regionalevaariatieindee p postverbaleedistributievanpresen ntatiefer. Haeggeman,Lilian neandHenkvanRiemssdijk(1986)VerbProjectionRaisin ng,Scope,an ndthe TTypologyoffRulesAffecctingVerbs..LinguisticInquiry17,4 417Ͳ466. Haeseryn,Walteer(1990).Syyntactischenormeninh hetNederlands.Eenem mpirischond derzoek n naarvolgorrdevariatieiindewerkw woordelijkeeeindgroep.D Dissertation n,Catholic U Universityo ofNijmegen n. Haeseryn,W.,K..Romijn,G.Geerts,J.d deRooijand dM.C.vand denToorn(1 1997).Algem mene N NederlandseeSpraakkun nst,2nd,revvisededition,Groningeen/Deurne:MartinusN Nijhoff aandWolterss/Plantyn. Hopp p,H.(2009)..ThesyntaxxͲdiscourseinterfaceinnearͲnativeeL2acquisition:OffͲlin neand o onͲlineperfformance.B Bilingualism:LanguageandCognition,12(4),4 463Ͳ483. Keller,Frank,&A Alexopoulou,Theodoraa(2001).Ph honologycompeteswithsyntax: eexperimentalevidenceefortheinteeractionofw wordorderandaccenttplacementtinthe rrealizationo ofInformatiionStructurre.Cognition,79,301Ͳ3 372. Koelm mans,L.,19 965.Ietsoveerdewoord dordebijsam mengesteldepredikateninhet N Nederlands.DeNieuweeTaalgids5 58,156Ͳ165.. Koop pman,HildaandAnnaSSzabolcsi(20 000).Verbalcomplexess.Cambridge,MA:MITͲͲPress. Nieuw wborg,E.(1 1968).Dedis istributievanhetonderrwerpenheetlijdendvoo orwerp. A Antwerpen::Plantyn. Pauw wels,A.(195 53)Deplaatsvanhulpw werkwoord,verledendeeelwoordeninfinitiefinde N Nederlandsebijzin.Leu uven,Symon ns. Soracce,Antonellla&Keller,Frank(2005 5).Gradiencceinlinguistticdata.Lingua,115,14 497Ͳ 1 1524. DeScchutter,G.((1974).“Weezenvissen.Dialektgeografievaneeenkonstruktie.”Taaleen TTongval26,70Ͳ85. Stroo op,Jan(1970).‘Systeem mingesprokkenwerkwo oordsgroepeen.’TaalenTongval22 2,128Ͳ 1 147.Reprintedin:J.Strroop,ed.,N Nederlandsd dialectonderrzoek.Amstterdam:Huiisaan d dedriegracchten,1983,,247Ͳ264. DeVrriendtͲdeM Man,M.J.,1969,Frequeentievanwo oordenensstructurenin nspontaan g gesprokenN Nederlands..Brussels:D Didier.
© Copyright 2024 ExpyDoc
