ebook img

Aspect-Based Agreement Reversal in Neo-Aramaic PDF

12 Pages·2012·0.3 MB·English
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Aspect-Based Agreement Reversal in Neo-Aramaic

Aspect-Based Agreement Reversal in Neo-Aramaic Laura Kalin and Coppe van Urk UCLAandMIT 1. Introduction This paper discusses a pattern of aspect-based agreement reversal in Neo-Aramaic languages, in which the functionof agreementmarkersswitches between perfectiveand imperfectiveaspect. In the examples in (1a–b), for instance, the agreement morpheme that references the object in one aspect referencesthesubjectintheother. (1) Functionofagreementmarkersswitchesbetweenaspects: a. qtil-´ı-le. ˙ kill.PERF-S.3PL-L.3MS ‘Hekilledthem.’ b. qatl-´ı-le. ˙ kill.IMPF-S.3PL-L.3MS ‘Theykillhim.’ (ChristianBarwar;Khan2008:167,282) This reversal is not an instance of aspect-based split ergativity, since agreementin both aspects has a nominative-accusativealignment. As such, agreementreversal persists even in unaccusatives, as seen in (2a–b): the morpheme that references the unaccusative subject in each aspect is the same one that referencesthetransitivesubjectin(1a–b). (2) Reversalextendstounaccusatives: a. Ma´mom´ıt-le q@`damta-w... Mamodie.PERF-L.3MSmorning-CONJ ‘Mamodiedinthemorningand...’ b. lE´-qEm-i ra`ba. NEG-grow.IMPF-S.3PLmuch ‘Theydonotgrowmuch.’ (ChristianBarwar;Khan2008:132,195) Thisisanunusualpatternsinceitisbasedonthesamedistinctionasmanyergativesplits(namely, aspect),butdoesnotappeartoinvolveanyergativity.Inthispaper,wearguethatagreementreversaland splitergativityarenonethelessdrivenbythesameprocess. Wewillfirstshowthat,despiteappearances, there is an asymmetry between the perfective and the imperfective. We then propose that agreement reversal arises because imperfective Asp introduces an additional ϕ-probe, drawing on work on split ergativitywhichsuggeststhataspectsplitsarisebecausenonperfectiveaspectsareassociatedwithmore structure(Laka2006;Coon2010;Coon&Preminger2011).Inthisway,ourproposalprovidessupport for this view of aspect-based splits in case and agreement, since it allows for the apparentlydisparate patternsofsplitergativityandagreementreversaltobegivenaunifiedtreatment. ∗ OurthankstoByronAhn,SabineIatridou,AnoopMahajan,DavidPesetsky,MashaPolinsky,OmerPreminger, Norvin Richards, and Carson Schu¨tze for helpful discussions and advice. We also owe a great debt to Laura McPhersonandKevinRyanfortheirworkonSenaya.Weusethefollowingabbreviations: ACC=accusative,CL= clitic,CONJ=conjunction,DEM=demonstrative,F=feminine,FUT=future,GEN=genitive,IMPF=imperfective, L=L-suffix,M=masculine,NEG=negation,P=preposition,PERF=perfective,PL=plural,S=S-suffix,S/SG= singular,1=1stperson,2=2ndperson,3=3rdperson.Theauthorsarelistedinalphabeticalorder. © 2012 Laura Kalin and Coppe van Urk. Proceedings of the 30th West Coast Conference on Formal Linguistics, ed. Nathan Arnett and Ryan Bennett, 184-194. Somerville, MA: Cascadilla Proceedings Project. 185 Thepaperisorganizedasfollows. Section2introducesthepatternofagreementreversal. Section 3 discusses the implications of the PCC effect found in the perfective and outlines the basics of our proposal. Section 4 outlines how the presence of an additional ϕ-probe on imperfective Asp can derive the pattern of agreement reversal, and presents evidence for this account from the behavior of ditransitivesandfromanotherunusualaspectsplitinSenaya. Finally,section5showsthatouraccount tiesinnaturallywithrecentworkonsplitergativity(Laka2006;Coon2010;Coon&Preminger2011) andtheideathatnonperfectiveaspectsincludeadditionalprepositionalstructure(Demirdache&Uribe- Etxebarria2000;Coon2010). 2. The patternofagreement reversal Many of the northeasternNeo-Aramaiclanguageshave adoptedaspect-based splits in agreement, apparentlyasaresultofcontactwithvarioussplitergativeKurdishlanguages(Doron&Khan2012).In a number of Neo-Aramaic languages, this aspect split takes on an unusual form, in that no ergativity appears to be involved. Instead, both sides of the split have a nominative-accusative alignment in agreement,butthesubjectmarkingofoneaspectistheobjectmarkingoftheother. WecallthispatternagreementreversalanditsurfacesinChristianBarwar,ChristianQaraqosh,and relatedvarieties(Khan2002,2008;Coghill2003;Doron&Khan2012).1 WefocusonChristianBarwar andChristianQaraqoshhere,becausethesearethemostwell-documented(Khan2002,2008). Agreementreversalmanifestsitselfinthemorphologythatattachestotheverb. TheNeo-Aramaic languageswediscusshereallmakeuseoftheverbaltemplatein(3),wherethetermsS-suffixandL-suffix refertodifferentsetsofagreementmarkers.2 (3) VerbaltemplateinNeo-Aramaic: Verbstem-S-suffix-L-suffix TheparadigmsforthesemorphemesinChristianBarwar,fromKhan(2008),areinTables1and2. Table1: S-suffixesinBarwar Table2: L-suffixesinBarwar Singular Plural Singular Plural 1stperson -@n -@x 1stperson -li -l@n 2ndperson -@t -itu 2ndperson -lux(m.)/-l@x(f.) -lE´xi 3rdperson -∅(m.)/-a(f.) -i 3rdperson -le(m.)/-la(f.) -lE/-la An importantfact about these paradigms, which holds in all agreementreversallanguages, is that the default S-suffix (3rd person masculine singular) has a null spell-out. This will be important in our analysis, aswe willproposethattheprobebehindS-suffixesispresentinsomederivationsbutfailsto registeragreementandsospellsoutasdefault,whichisnull. TheagreementmarkersinTables1and2combinewithanumberofdifferentverbalbases,which areformedbymeansofroot-and-patternmorphology. Wewillbeconcernedwiththeimperfectiveand perfectivebases. SomeexamplesoftheseinBarwararegiveninTable3. Table3: Barwarverbalbases Root Imperfectivebase Perfectivebase pθx(‘toopen’) paθ@x pθix m-sˇlx(‘tostrip’) msˇal@x msˇol@x m-plx(‘touse’) mapl@x mupl@x 1 DoronandKhan(2012)treatagreementreversalasancaseofsplitergativityinvolvingextendedergativity(i.e.a systeminwhichergativegeneralizestounaccusatives)intheperfective. Wedonothavethespacetodiscusstheir proposalhere,thoughwenotethattheexistenceofsuchergativesystemsiscontroversialandthattheiraccountdoes notprovideaprincipledexplanationofthedirectionalityoftheagreementasymmetriesbehindagreementreversal. 2 AlthoughweadoptthisterminologyfromKhan(2002,2008),thetermL-suffixisamisnomer,aswewillargue thattheL-suffixesareclitics,followingDoronandKhan(2012). Thisismotivated, forinstance, bythefactthat L-suffixes,butnotS-suffixes,canappearoutsideofotherencliticmaterial,specificallytheencliticauxiliary. 186 BothofthesebasestakeS-suffixesandL-suffixesinaccordancewiththeverbaltemplatein(3). Thefunctionsoftheseagreementmarkersreversebetweenaspects,however. Whenattachedtothe imperfectivebase,theS-suffixmarkssubjectagreementandtheL-suffixmarksagreementwithadefinite orpronominalobject. Theexamplesin(4a–c),againfromChristianBarwar,illustrate. (4) IMPERFECTIVE S-suffix=subject,L-suffix=object: a. mEy-@´n-na ’ay-ba`xta. bring.IMPF-S.1SG-L.3FSDEM-woman ‘Ishallbringthatwoman.’ b. xosˇe´ba la´-palx-i na`sˇe. SundayNEG-work.IMPF-S.3PLpeople ‘OnSunday,peopledonotwork.’ c. ’a´namE´θ-en ’as@`rta. ˙ I die.IMPF-S.1SGevening ‘Ishalldieintheevening.’ (ChristianBarwar;Khan2008:115,132,135) Thesubjectofatransitive,unergative,andunaccusativeallpatternalike(markedwithS-suffixes),while objectsaretreateduniquely(markedwithL-suffixes). Withtheperfectivebase,however,L-suffixesnolongermarkobjects,butcross-referencethesubject, eveninanunaccusative.S-suffixesnowfunctionasobjectmarkers,andsoonlyappearintransitives.3 (5) PERFECTIVE S-suffix=object,L-suffix=subject: a. xawr-a˘waθ-i brat-i grisˇ-a-la. friend-PL-1SGdaughter-1SGpull.PERF-S.3FS-L.3PL ‘Myfriendspulledmydaughter.’ b. kalbanwix-le. dog bark.PERF-L.3MS ‘Thedogbarked.’ c. brat-i qim-la. daughter-1SGrise.PERF-L.3FS ‘Mydaughterrose.’ (ChristianBarwar;Doron&Khan2012:231) Wewillrefertothispatternasagreementreversal. Itcanbeschematizedasin(6). (6) Verbalbase S-suffix L-suffix Imperfective - SubjAgr - ObjAgr Perfective - ObjAgr - SubjAgr Thisaspectsplitisquiteunusual.Itisreminiscentofsplitergativity,sinceitshowsupinaplacetypical ofergativesplits,namelybetweentheperfectiveandimperfective(e.g.HindiandChol). However,both sidesoftheaspectsplithaveafullynominative-accusativealignment,sothatthereisnorealergativity anywhereinthesystem. 3. The perfective andthe PCC We will start by developing our account of the perfective, which we take to have the most basic syntax. We argue that, despite appearances, there is only one ϕ-probe in the perfective, which has 3 Justlikeobjectagreementintheimperfective,thisagreementissubjecttoadefinitenessrestriction. Wewillnot tobetooconcernedwiththishere,thoughthereareatleasttwoaccountsofthisthatarefullycompatiblewithour theoryofagreementreversal.Oneoptionistosaythatindefiniteobjectsundergopseudo-incorporation,inthesense of Massam(2001) andDayal (2011), sothat theyareinaccessible foragreement. Another optionisthat definite objectsraisetotheedgeofalowerphase(e.g.VP),followingthetreatmentofSakhadifferentialobjectmarkingin BakerandVinokurova(2010). Thismovementstepwouldmakedefiniteobjects,butnotindefiniteobjects,visible foragreement.TheformerapproachisadoptedfortheSenayafactsin(22a–b)inKalinandvanUrk(toappear). 187 to license both the subjectand the object. Thisprobeclitic-doublesthe subject, while registeringtrue agreementwiththeobject,creatingtheappearanceoftwoseparateagreementmorphemes.Theempirical motivationforthisproposalcomesfromthe existenceofa StrongPCC effectbetweenthesubjectand theobjectintheperfective,andoursyntaxfortheperfectivemirrorstheBe´jarandRezac(2003)theory ofthisphenomenon. 3.1.ThePersonCaseConstraint ThePERSON CASECONSTRAINT(PCC)referstotheungrammaticalityofcertaincombinationsof person when two argumentsoccupy the same domain (Bonet 1991). We will be concernedhere with theStrongPCC,whichrestrictsthelowerargumentto3rdperson.InGreekditransitives,forinstance,a directobjectcliticinthecontextofanindirectobjecthastobe3rdperson: (7) DirectobjectcliticofGreekditransitiveshastobe3rdperson: a. Tha tu to stilune. FUTCL.GEN.3SGCL.ACC.3SGsend-3PL ‘Theywillsendittohim.’ b. *Thatu se stilune. FUT CL.GEN.3SGCL.ACC.2SGsend-3PL ‘Theywillsendyoutohim.’ (Greek;basedonAnagnostopoulou2005) WecanstatetheStrongPCCasfollows: (8) StrongPCC(Bonet1991): FortwoargumentsinadomainX,thelowerargumenthastobethirdperson. Intheagreementreversallanguages,aStrongPCCeffectobtainsbetweenthesubjectandobjectofthe perfective,suchthattheobjecthastobe3rdperson:4 (9) Barwarobjecthastobe3rdpersonintheperfective: a. *grisˇ-an-le. pull.PERF-S.1FS-L.3MS ‘Hepulledme.’ b. *grisˇ-at-le. pull.PERF-S.2FS-L.3MS ‘Hepulledyou.’ c. grisˇ-´ı-le. pull.PERF-S.3PL-L.3MS ‘Hepulledthem.’ (ChristianBarwar;Doron&Khan2012:232) Most accounts of the PCC assume that PCC effects arise when two arguments compete for the attentionofoneϕ-probe(e.g.Anagnostopoulou2003;Be´jar&Rezac2003;Nevins2007;Rezac2011).5 Assuch,wetaketheexistenceofaPCCeffecttosuggestthatthereisonlyoneϕ-probeintheperfective ofagreementreversallanguages,whichhastolicenseboththesubjectandtheobject. To be precise, we adoptthe Be´jar and Rezac (2003)accountof the StrongPCC. Be´jar andRezac propose that a ϕ-probe can sometimes agree with multiple arguments because ϕ-probes consist of separate person (π) and number (#) probes, which probe independently. These probes can end up targetingdifferentDPsif theonethatprobesfirst, whichBe´jarandRezacassume isthepersonprobe, triggers a change in the DP it agrees with. For Be´jar and Rezac, this change is clitic-doublingof the 4 Inordertoexpressa1stor2ndpersonobjectwiththeperfective,theselanguagesmakeuseoftwostrategies.The objectcanbeembeddedunderapreposition,inwhichcaseallpersonsareacceptable,ortheperfectiveisexpressed periphrastically,byputtingaperfectiveprefixontheimperfectivebase(agreementisjustasintheimperfective). 5 Note that the syntactic signature of the PCC means that a morphological analysis of agreement reversal, as suggested by Baerman (2007) for Hertevin, is not appropriate for Barwar. For detailed argumentation that the PCCissyntactic,seeRezac(2011). 188 goal DP, which they assume makes argumentsinvisible for further probing.6 If the person probe is a clitic-doubler, the highest argument becomes inaccessible to further probing. The number probe will thenenduptargetingadifferentDP.Inthisway,oneϕ-probecanagreewithmultiplearguments. Thetreein (10)representsthissituation. PersonprobesfirstandagreeswithDP . Aspartofthis 1 Agreerelation,thepersonprobeclitic-doublesDP ,sothatDP becomesinvisibleforfurtherprobing. 1 1 Numberthenprobes,ignoringDP ,andagreeswithDP . 1 2 (10) π,# CL DP 1 X DP 2 Be´jarandRezacproposethattheStrongPCCeffectemergesinthisenvironment.Specifically,they arguethattheStrongPCCeffectarisesbecause1stand2ndpersonDPsarespecialinthattheyrequire personagreementtobelicensed. TheycallthislicensingrequirementthePLC,statedasfollows: (11) PersonLicensingCondition(PLC;Be´jar&Rezac2003): Interpretable1st/2nd-personfeaturesmustbelicensedbyenteringintoanAgreerelationwithan appropriatefunctionalcategory. ThePLCmeansthat,inthesituationin(10),onlythehigherargumentcanbe1stor2ndperson,since only the higher argument,DP , entersinto person agreement. In contrast, the lower argumentin (10) 1 DP ,onlyagreeswithanumberprobeandsowillviolatethePLCif1stor2ndperson. 2 The PLC is a licensing requirement that is independent of case. Be´jar and Rezac assume that agreementwithanyϕ-probe,whetheritisapersonornumberprobe,sufficesforcaseassignment. As such,a3rdpersonDPinthelowerpositionin(10)iscase-licensedbyvirtueofagreeingwithanumber probe. And, since a 3rd person DP is not subjectto the PLC, it does not need to agree with a person probe to be licensed. In this way, the lower argumentis restricted to 3rd person and the Strong PCC effectisderived. 3.2.Agreementintheperfective ToderivetheStrongPCCeffectinagreementreversallanguages,weimplementtheBe´jarandRezac (2003)account.Wewilltreattheperfectiveinagreementreversallanguagesasanenvironmentinwhich oneϕ-probeagreeswithmultiplearguments.Inparticular,weproposethatChristianBarwarandrelated languagesdifferfromothernominative-accusativelanguagesinthatTcarries aϕ-probe, butnotv, so thatvisinactive.7 Asaresult,thereisnoheadthatisdedicatedtolicensingobjects;thus,intransitives, boththesubjectandtheobjecthavetobelicensedbyT. To be precise, we propose that the person probeon T in agreementreversallanguagesis a clitic- doubler, which clitic-doubles the DP it agrees with. The reflex of this relation is an L-suffix, which we take tobe a clitic, followingDoronandKhan(2012;see also fn. 2). In theperfective,intransitive and transitive subjects are then referenced by L-suffixes, because they are the highest argument and thereforetheargumentthatthepersonprobeonTagreeswithandclitic-doubles. Thenumberprobeon Tsubsequentlyagreeswithanobject,ifoneispresent.Weproposethatthisagreementisspelledoutby 6 We can see that cliticization makes DPs invisible for probing in languages like Greek and Basque, in which clitic-doubledargumentsdonotcountforintervention. Onewayoftheoreticallyimplementingthisobservationis proposedbyAnagnostopoulou(2003),whosaysthatthetailofaclitic-doublingchainisanA’-traceandthehead ofthechainisnolongeraphrase,butahead. 7 IthastobeTthatisactiveandnotv,becausethiswouldnotmapstraightforwardlyontoaPCCconfiguration. Specifically,iftheϕ-probewereonv,thenwewouldhavetomakeanadditionalstipulationaboutthedirectionality ofprobing(upwardsthendownwards)inordertoaccountforthefactthatthePCCaffectsobjectsandnotsubjects. Inaddition,whiletheperfectivecouldbeaccountedforthiswiththisstipulation,itdoesnotallowimperfectiveAsp tointerfereinthedesiredwayintheimperfective,asisneededtoderiveagreementreversal;see§4.1. 189 an S-suffix, which markstrue agreement. These assumptionsimplementthe Be´jar and Rezac account andsoderivetheprofileofagreementintheperfective. To see exactly how this works, we will run through the derivations of perfective transitives and intransitives. Consider first an example of a perfectivetransitive (12), in which an L-suffix marks the subjectandanS-suffixtheobject.Thishasthesyntaxin(13). (12) PERFECTIVE TRANSITIVE S-suffix=object,L-suffix=subject: xawr-a˘waθ-i brat-i grisˇ-a-la. friend-PL-1SGdaughter-1SGpull.PERF-S.3FS-L.3PL ‘Myfriendspulledmydaughter.’ (ChristianBarwar;Doron&Khan2012:231) (13) TP AspP T CL π,# Asp vP PERF Subj v VP V Obj Inthisderivation,persononTprobesfirst,licensingthesubject,andtriggeringcliticdoubling.This cliticdoublingspellsoutasanL-suffix,sinceL-suffixesarethecliticseries.NumberonTnowprobes.It ignoresthesubjectbecausethesubjectisclitic-doubled(seefn.6),andagreeswiththeobject,licensing itaswell. NumberagreementwiththeobjectisspelledoutasanS-suffix. ThisderivesthePCCeffect, sincetheperfectivetransitivemapsexactlyontotheBe´jarandRezac(2003)configurationin(10). The objectoftheperfectiveonlyagreesfornumber,andsoviolatesthePLCif1stor2ndperson. Thisaccountextendstoperfectiveintransitives,like(14),whichhavethestructurein(15). (14) PERFECTIVE INTRANSITIVE L-suffix=subject: brat-i qim-la. daughter-1SGrise.PERF-L.3FS ‘Mydaughterrose.’ (ChristianBarwar;Doron&Khan2012:231) (15) TP AspP T CL π,# Asp vP PERF v VP V Subj Inthisstructure,persononTprobesthesubjectandtriggerscliticdoubling(becauseπindiscriminately clitic-doubles the argument it agrees with). This clitic spells out as an L-suffix. Number on T now probes. Sincethesubjecthasbeenclitic-doubled,itisnolongervisibleforprobing. Assuch,number findsnothingandspellsoutasthedefaultS-suffix(3sgmasculine),whichisnull. Thisconcludesouraccountof agreementin the perfective. Intransitiveandtransitive subjectsare markedthe same waybecausebothare clitic-doubledbythe personprobeonT. Transitiveobjectsare different,becausetheyarethetargetofrealagreement,namelywiththenumberprobeonT. 190 4. The syntaxofagreement reversal In this section, we develop our account for the imperfective. We propose that imperfective Asp introducesanadditionalϕ-probe,whichcausesanapparentreversalinagreement. Thisprobe,andnot T,willtargetthesubject,becauseAspismergedfirst. Tcanthenignorethesubjectandlicensetheobject instead. ThisexplainswhythereisnoPCCeffectintheimperfectiveandwhyimperfectiveobjectsare markedinthesamewayasperfectivesubjects—bothareprobedbyT. 4.1.Anadditionalϕ-probeintheimperfective Asnoted,theimperfectivediffersfromtheperfectiveinthatthereisnoPCCeffect:theobjectisnot limitedto3rdperson. Instead,we getfullagreementforpersonandnumberwithboththesubjectand theobject,as(16)illustrates. (16) NoPCCeffectintheimperfective: ’u-bt-amr-@`n-nux. ˙ CONJ-FUT-say.IMPF-S.1SG-L.2MS ‘AndIshallsaytoyou.’ (ChristianBarwar;Khan2008:175) Assuch, we assumethatthereare two ϕ-probesin the imperfective,so thatthe subjectandthe object can be licensed separately for person. In particular, we propose that imperfective Asp introduces an additionalϕ-probe. This ϕ-probe is not a clitic-doubler, so that it only registers agreementwith (but doesnotclitic-double)thesubject. Inaddition,weassumethatS-suffixesspelloutrealagreementregardlessofwhatheadinitiatesit. As a result, both agreementthat is triggeredby imperfectiveAsp and agreementtriggered by number onTisspelledoutasanS-suffix. Imperfectivesubjectsandperfectiveobjectsarethenmarkedwiththe samesuffixnotbecausetheyagreewiththesamehead,butbecausebotharethetargetofrealagreement. To see how this proposal derives agreement reversal, we will examine how transitives and intransitivesarederived.Considerfirstimperfectiveintransitives,liketheunaccusativein(17),inwhich thesubjectisreferencedbyanS-suffix. Thesehavethestructurein(18).8 (17) IMPERFECTIVE INTRANSITIVE S-suffix=subject: ’a´namE´θ-en ’as@`rta. ˙ I die.IMPF-S.1SGevening ‘Ishalldieintheevening.’ (ChristianBarwar;Khan2008:132) (18) TP T AspP Asp vP IMPF ϕ-probe v VP V Subj In this structure, the ϕ-probe on imperfective Asp is the ϕ-probe that is merged first and so it agrees with the subject. Because S-suffixes spell out agreement (regardless of where agreement originated), thisspellsoutasanS-suffix. PersonandnumberonTthenprobe,butfindnoactivegoal,andsospell outasthedefaultS-suffix,whichisnull. 8 AlthoughwerepresentAspasanundifferentiatedprobehere,itisalsoassumedtoconsistofaseparateperson andnumberprobe.Sinceneitherisaclitic-doubler,theseprobeswilljustalwaysagreewiththesameargument. 191 Recall now that, in imperfectivetransitives, fullobject agreementis marked with an L-suffix, the samemorphemethatmarkssubjectagreementintheperfective,asintheexamplein(19). We assume theunderlyingstructurein(20). (19) IMPERFECTIVE TRANSITIVE S-suffix=subject,L-suffix=object: ’u-bt-amr-@`n-nux. ˙ CONJ-FUT-say.IMPF-S.1SG-L.2MS ‘AndIshallsaytoyou.’ (ChristianBarwar;Khan2008:175) (20) TP AspP T CL π,# Asp vP IMPF ϕ-probe Subj v VP V Obj Inthistree,theϕ-probeonimperfectiveAspagreeswiththesubject,asabove,andthisagreement spellsoutasanS-suffix. WeproposenowthatBarwarmakesavailableamovementsteptospec-TPfor thesubject,sothatitgetsoutofthewayoftheagreementprobeonT.Thisallowstheobjecttobeagreed withwithoutthesubjectintervening.9 ThepersonprobeonTnowclitic-doublestheobject,creatingan L-suffix. Thenumberprobedoesnotfind anargumentto agreewith, becausethe availablearguments have either been clitic-doubledor movedout of the way of the probe on T, and so it spells out as the defaultS-suffix,whichisnull. ThisaccountderivestheappearanceofagreementreversalintheNeo-Aramaiclanguageswehave examined. Agreement reversal boils down to the interaction of two factors. First, the ϕ-probe on imperfectiveAspcausestheimperfectiveobjecttobetreatedliketheperfectivesubject(bothareprobed byπonT).Second,becausetrueagreementspellsoutuniformlywithS-suffixes,perfectiveobjects(true agreementwithT)aremarkedjustlikeimperfectivesubjects(trueagreementwithAsp). 4.2.Evidenceforthisaccount This section briefly discusses some evidence for our account. First of all, the current proposal derives some puzzling facts about a common strategy for expressing multiple objects in ditransitives in these languages. Specifically, it makes an interesting prediction aboutditransitivesthat we show is correct.Asthetreein(20)illustrates,thenumberprobeonTremainsfreeinouranalysisofimperfective transitives,unlikeinperfectivetransitives(15),inwhichallprobesagree. Ifanadditionalargumentis available,wethenpredictthatitcanbeagreedwithintheimperfective,butnotintheperfective.Further, weshouldthenseethePCCresurfaceandaffectthelowestargument. 9 Thisisnotacountercyclicderivation.Instead,theclaimhereisbasicallythatthereisanEPPfeatureonTwhich isunorderedwithregardtoT’sotherfeatures, sothattheorderingofprobingandmovement isfree. Weassume thenthatthereisreallymovementofthesubjecttospec-TPinallofthederivationsdiscussed. Crucially,however, orderingthemovementsteptospec-TPbeforeprobingresultsinungrammaticalityintheperfective,sinceitprevents thesubjectfrombeinglicensed.Assuch,movementtospec-TPhastobeorderedafterprobinginthesederivations and so does not affect the conditions on agreement. For more on the idea that arguments can move out of the wayofaϕ-probeinthisfashion,seeHolmbergandHro´arsdo´ttir(2003),AnandandNevins(2006),Legate(2008), SigurDssonandHolmberg(2008),Preminger(2011),andHalpert(2012). 192 Inanumberoftheagreementreversallanguages,includingChristianQaraqoshandAlqosh(Khan 2002;Coghill2003),weseeexactlythis,thoughwithatwistthatwewillshowprovidesstrongsupport for our analysis.10 We illustrate with Qaraqosh (Khan 2002). In Qaraqosh, ditransitivesindeed allow both objects to be marked on the verb with the imperfective base (and not with the perfective base!). Intheseconstructions,a strongPCC effectobtains,so thatonly3rdpersondirectobjectsare tolerated (21a).Ifthedirectobjectis1stor2ndperson,theindirectobjecthastobeexpressedinaPP(21b): (21) StrongPCCeffectinQaraqoshditransitives: a. k-ew-i-lux-ila. INDIC-give.IMPF-S.3PL-L.2MS-COP.3FS ‘Theygivehertoyou.’ b. k-ew-´ı-lax gda´l-eh. INDIC-give.IMPF-S.3PL-L.2FSto-P.3MS ‘Theygiveyoutohim.’ (ChristianQaraqosh;Khan2002:143–144) This is exactly what our analysis predicts: the indirect object is the higher of the two objects, and is probedbypersononTandclitic-doubled(surfacingasanL-suffix,thecliticseries);thedirectobjectis thelowerobject,and,assuch,onlythenumberprobeonTagreeswithit,restrictingitto3rdperson. However,thereisaninterestingwrinkleinthispattern.Observethat,in(21a),thedirectobjectisnot expressedwithanS-suffix,asperfectiveobjectsare. Instead,agreementwiththedirectobjectishosted ontheencliticauxiliary-i,whichalsoservesasthecopula. Weproposethatthisauxiliaryisinsertedto hostthe strandednumberagreementonT, adoptinga viewofauxiliariesin whichtheyare insertedto hoststrandedinflectionalfeatures(seeBjorkman2011forextensivediscussion).11 Assuming this, our analysis straightforwardlyderivesan otherwise puzzlingconstellation of facts aboutthis way of realizingagreementin ditransitives. We derivethe presenceof a strongPCC effect. More strikingly, our analysis explainswhy it is the highest verbalelement(the enclitic auxiliary)that expressesagreementwiththelowestargument(thedirectobject). Finally,ourproposalderivesthefact thatthisstrategyisavailableonlywiththeimperfectiveandnotwiththeperfective. Anotherpieceofevidencefortheanalysisofagreementreversalpresentedherecomesfromanother unusualaspectsplitfoundintheNeo-AramaiclanguageSenaya(Kalin&vanUrk,toappear).12 Inthis language,agreementreversalisonlypartial. JustasinBarwarandrelatedlanguages,theimperfective uses an S-suffix to mark the subject and an L-suffix to crossreference the object. In the perfective, however,Senayadiffersfromthe languagesdiscussed. An L-suffixreferencesthe subject, butthereis noobjectagreementandnoS-suffix(22a–b),sothatdefiniteorpronominalobjectscannotbelicensed atallintheperfective(cf. fn. 3). (22) PERFECTIVE NoS-suffix,L-suffix=subject: a. axniidmex-lan. we sleep.PERF-L.1PL ‘Weslept.’ b. axniixaksuutaksuu-lan. we a book write.PERF-L.1PL ‘Wewroteabook.’ (Senaya) In Senaya then, we overtly see what we argued is true for the agreement reversal languages underlyingly: thereisoneagreementprobeintheperfective,buttherearetwo intheimperfective. As such, the present account can straightforwardlybe extended to Senaya, if we assume that, in Senaya, the L-suffix has been re-analysed as true agreement on T, while S-suffixes spell out agreement on imperfective Asp. There is then no clitic doubling in Senaya, and the number and person probes on 10 Otheragreementreversallanguages,likeBarwar,simplydonotallowobjectagreementwithmultipleobjectsin aditransitiveandinsteadalwaysexpresstheindirectobjectinaPP(Khan2008). 11 Crucially,insertionoftheauxiliaryisnottriggeredbytheL-suffix,sinceitisacliticandnotaninflectionalfeature onT.Inthisway,thispatternalsoprovidessomeindirectsupportfortheideathattheL-suffixisaclitic. 12 ThedatainthissectioncomesfromfieldworkandmorphologicalanalysisbyLauraMcPhersonandKevinRyan. 193 Tagreewiththesameargument. Inthisway,theapproachdevelopedhereallowsforaunifiedaccount oftwounusualaspectsplits,agreementreversalandSenaya’ssystemofpartialagreementreversal.This approachtoSenayaagreementisfurtherexploredinKalinandvanUrk(toappear). 5. Accounting foraspect splits Our syntax for agreement reversal languages derives agreement reversal from the idea that the perfective has one ϕ-probe and the imperfective has two. The question that arises now is why there shouldbeanadditionalϕ-probeintheimperfective. Inthissection,wesuggestthattheanswerliesin workonaspect-basedsplit ergativitywhicharguesthatsuch splitsarise becausenonperfectiveaspects introduceadditionalprepositionalstructure(Laka2006;Coon2010;Coon&Preminger2011). Recent work on split ergativity explores the hypothesis that nonperfective aspects are associated withadditionalstructure(Laka2006;Coon2010;Coon&Preminger2011). Thisworkismotivatedby Dixon’s(1994)observationthatthereisaconsistentdirectionalitytoaspectsplits. Dixonpointsoutthat aspect-basedergativesplitsarealwaysergativeontheperfectiveside(23),asschematizedin(24). (23) Dixon’sobservation(1994:99): “...If a split is conditioned by...aspect, the ergative marking is always found...in perfective aspect.” (24) Universaldirectionalityofaspectsplits: (ERG/ABS)perfective>>imperfective>>progressive(NOM/ACC) Coon (2010) proposes that Dixon’s observation derives from the idea, developed in work on tense and aspect by Demirdache and Uribe-Etxebarria (2000), that nonperfective aspects involve a prepositionalpredicateabsentintheperfective. Coonpointsoutthatsuchanadditionalpredicatemay also introduce a new case/agreement domain, shifting a language out of its canonical case/agreement pattern. Split ergative languages are then really ergative throughout, but the presence of additional case/agreementdomainsinsomeenvironmentsdisruptsthispattern. This approach to split ergativity has been explored in two ways. For Basque and Chol, Laka (2006) and Coon (2010), respectively, propose that this extra prepositional predicate is expressed as an independentverb, so that nonperfectiveaspects are biclausal. Coon and Preminger (2011)suggest thattheprepositionalpredicateintroducedbynonperfectiveaspectsmayintroduceanadditionalphase boundary(Coon&Preminger2011),disruptingaMarantzianassignmentofergativecase. Theseallinvolvenaturalpropertiesofprepositions. Sinceprepositionsarepredicates,theycanbe expressedasamainverb.Inaddition,prepositionsarephases,sothatwemightexpectthispropertytobe preservedaswell. Anotherpropertyofprepositionsisthattheycanintroduceϕ-probesandassigncase. Ifwe taketheaboveviewofsplitergativityliterally, wemightimaginethatthesepropertiestoocould effectanaspectsplit. Thisiswhatweproposeliesbehindtheadditionalϕ-probeintheimperfectivein Neo-Aramaic.Whatdistinguishesagreementreversallanguagesisthat,inordertoexpressnonperfective aspect, they make use of a prepositional predicate that retains a ϕ-probe. As a result, there is an additionalϕ-probeintheimperfective. Thestructureofperfectiveandimperfectiveaspectinagreementreversallanguagesisthen(25a–b), whereAsp introducesaPwhichhasretaineditsϕ-probe. IMPF (25) a. Perfectiveaspect: b. Imperfectiveaspect: TP TP T AspP T AspP ϕ-probe ϕ-probe Asp vP PERF Asp PP IMPF ... vP P ϕ-probe ...

Description:
examples in (1a–b), for instance, the agreement morpheme that references the object in one aspect references the Agreement reversal manifests itself in the morphology that attaches to the verb. Lingua 113, 997–1019. Kalin
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.