BetweentheHistoricalLanguagesandtheReconstructedLanguage:AnAlternativeApproachtotheGerundive+“DativeofAgent”ConstructioninIndo-European*

SerenaDanesi,CynthiaA.Johnson,&JóhannaBarðdalUniversityofBergen,GhentUniversity&GhentUniversity

Abstract

ItisarguedbyHettrich(1990)thatthe“dativeofagent”constructionintheIndo-European languagesmost likely continues a construction inherited fromProto-Indo-European.Intworecentproposals(Danesi2013,Luraghi2016),itisarguedthat the “dative of agent” contains no agent at all, although the two proposalsdifferwithregardto thereconstructabilityof the“dativeofagent”construction.Luraghiarguesthatitisanindependentsecondarydevelopmentfromanoriginalbeneficiary function (cf. Hettrich 1990), while Danesi maintains that theconstructionisreconstructableforanearlierproto-stage.ElaboratingonDanesi’sapproach,weanalyzegerundiveswiththe“dativeofagent”insixdifferentIndo-Europeanlanguagesthatbridgetheeast–westdivide,namely,Sanskrit,Avestan,Ancient Greek, Latin, Tocharian, and Lithuanian. Scrutiny of the data revealssimilaritiesatamorphosyntactic level,asemantic level(i.e.modalmeaningandlowdegreeoftransitivity),andalso,tosomeextent,atanetymological level.Ananalysis involvingamodal readingof thepredicate,withadative subject andanominative object, is better equipped to account for the particulars of the“gerundive + nominative + dative” construction than the traditionalagentive/passive analysis. The proposal is couched within the theoreticalframeworkofConstructionGrammar, inwhich thebasicunitof language is theConstruction, i.e. a form–functioncorrespondence,andnoprincipleddistinctionbetween lexical items and complex syntactic structures is assumed. As thesestructures are by definition units of comparanda, required by the ComparativeMethod, they can be successfully utilized in the reconstruction of a proto-constructionforProto-Indo-European.

Keywords: dative of agent, gerundives, non-canonical case-marking, syntactic

reconstruction,constructiongrammar * Forvaluablecommentsanddiscussions,wethankLeonidKulikov,RomanoLazzaroni,SilviaLuraghi,ElisabethRieken,ananonymousreviewerofIF,theNonCanCaseresearchteaminBergen,theEVALISAresearchteaminGhent,andtheaudiencesatAppraisingCaseinBergen(2012)andatthe21stInternationalConferenceofHistoricalLinguisticsinOslo(2013)whereanearlierversionofthispaper(Danesi2013)waspresented.ThisresearchwassupportedbygenerousresearchgrantsawardedtoJóhannaBarðdal(PI)fromtheNorwegianResearchCouncil(NonCanCase,grantnr.205007)andtheEuropeanResearchCouncil(EVALISA,grantnr.313461).

2

1. IntroductionThisarticletacklestwoseparateissueswithinhistoricalsyntaxofIndo-European:first,the so-called “dative of agent” construction that co-occurs with gerundive forms inmany of the Indo-European subbranches, and second, the issue of reconstruction ofsyntax. The “dative of agent” is a function of the dative case that is most notablydocumentedfornon-finiteformssuchasgerundives,i.e.verbaladjectivesthathavealsobeen described as “future passive participles” (on dative of agent, cf. Green 1913,Hettrich 1990, 2014, George 2005, Luraghi 2016; for “future passive participle”, cf.Smyth 1963: 81 for Greek, Macdonnell 1916: 186 for Vedic Sanskrit, for example).Verbal adjectives are, as thename implies, adjectivesderived fromverbs.Gerundivesagreewiththenounstheyarepredicatesof,andthisnountypicallycorrespondstoanargument of the root verb; hence, gerundives show behavior characteristic of bothverbsandadjectives.Toconcretize, in theLatinexample in(1)below, thenominativenounisres‘affair’andthegerundiveagreeingwithitisagitanda‘whichistobemoved’.Acorrespondingfiniteverbwithagitari‘move’intheactivediathesiswouldhaveresasitsobjectintheaccusativecase,whilethesubjectwouldbeinthecanonicalnominative.(1) non agitanda res erit not move.GER.NOM affair.NOM be.3SG.FUT

‘Mustnotthematterbeagitated?’(Cic.Verr.2.5.179)

Moreover, inconstructions involvinggerundives,aspecificmodalmeaningexpressingthespeaker’sstanceisfound,indicatingnecessityfortheeventtohappeninthefuture,i.e.aneventthatshouldoroughttobecarriedout(cf.alsoDanesi, Johnson&Barðdal2016).

Gerundivescanoccurwithasingleargumentinthenominativeasin(1)above,witha single argument in thedativeas in (2a)below,orwithmultiple arguments, ofwhichonemaybethenominativeandtheotherdativeasin(2b)below.Theselasttwowill be referred to with the combined label of GER+(NOM+)DAT construction in theremainder of this article (for a discussion of different subconstructions of the moreschematicGER+(NOM)+(DAT)construction,seeSection4.1below).(2) a. Latin

alieno more vivendum estmihiof.another.ABL manner.ABL live.GER.NEUTbe.PRES.3SGme.DAT‘Imustliveaccordingtothemoodofanother’(Ter.And.1,1,125)

b. Latin Haec precipua colenda est nobis

this.NOM.F especially cultivate.GER.NOM.Fbe.3SG we.DAT ‘Thisvirtuewemustespeciallycultivate’(Cic.DeOr.2.148) WhiletherelationshipbetweenthedifferentsubconstructionsoftheGERconstructionarecertainlyoflinguisticinterest,wefocusontheconstructionswithadativeargument(2a–b)inthisarticle.

TheGERconstructionsallexpressdeonticmodality,involvingnotionslikeduty,obligation, prohibition, and permission. Expressions of deontic modality typically

3

evaluate thecontentofanutterance,basedon the judgmentofan individualor somemoral/social convention (cf., for instance, Palmer 2001). Such utterances differ fromepistemicexpressions,whichinsteadevaluateastatement’struthvalidity.Bothtypesofmodal expressions, deontic and epistemic, share the same structure throughout theIndo-Europeanlanguagefamily(Hettrich1990:64ff).

While the constructions described above have traditionally been analyzed aspassive-like, and following this, thedativehasbeenanalyzedas ademotedagent,wearguethatboththemodalityandthedativeargument’ssemanticrole isapropertyofthe construction as a whole, a construction which we analyze as a sub-type of theordinary Oblique Subject Sonstruction which is found more widely across the Indo-Europeanlanguagefamily(cf.Conti2009,Luraghi2010,Barðdaletal.2012,2013,Dahl&Fedriani2012,Matasović2013,Danesi2014,Fedriani2014,Viti2016,interalia). Given theoccurrenceof theGER+(NOM+)DATconstructionacross sixdifferentIndo-European languages (and five different branches) and its categorization as anobliquesubjectconstruction,suchaconstructioncanandshouldbereconstructed forProto-Indo-European. However, in the field of historical linguistics, syntacticreconstructionisonlyjustbeginningtoreceivedueattention,asupuntilthemid-1990s,suchataskwasmetwithsevereskepticism.Unlikethelexicon,syntaxdoesnotprovidewell-defined entities for comparison. The Comparative Method operates on theprinciplethatunitsthataretobecompared, i.e.comparanda,arearbitrarypairingsofform and meaning which, as such, must be directly inherited from an earlier proto-stage. Since syntactic structures are traditionally considered to lack a meaningcomponentoftheirown(forclaimstothateffectandacriticismthereof,seeHarrison2003 and Klein 2010, respectively), identifying cognate entities seems like anunattainable enterprise (see however Barðdal & Eythórsson 2017). Consequently,syntactic structures were largely excluded from historical-comparative frameworks,except,perhaps,asanextensionoflexicalreconstructionininstancesof“archaic”frozensyntax(Watkins1976,Jeffers1976,Winter1984,amongothers). During the last two decades or so, syntactic reconstruction has been a fieldcomingintoitsown,incontrasttohistorical–comparativeresearchforthelast150–200yearswhichhasmoreorlessfocusedonlexicalandphonologicalreconstruction.Threedifferent schools of research on syntactic reconstruction can be identified withinhistoricallinguisticstoday(cf.Eythórsson&Barðdal2016).Theseare:● the traditional Indo-Europeanist school (for instance, Lühr 2008, Hock 2013,

Keydana 2013, Kulikov & Lavidas 2013, Cotticelli Kurras & Rizza 2013, Viti2014)

● thegenerativistschool(Hale1987a–b,Garrett1990,Willis2011,Walkden2014)● theconstructiongrammarschool(Barðdal&Eythórsson2012a–b,2017,Barðdal

2013,2015,Barðdal&Smitherman2013,Barðdaletal.2013,Danesietal.2017).Theproblemforthefirsttwostrandsofresearchaboveisthatthesearenotnecessarilywell equipped for the enterprise of carrying out syntactic reconstruction; the Indo-European paradigm lacks a formal representational system to model itsreconstructions, while the representational system of the generative paradigm is notbuilt to adequately explicate themeaningpart of the form–meaning correspondencesunderlyingtheuseoftheComparativeMethod.

We argue, in line with the nascent but ever-growing research interest insyntactic reconstruction, that comparison among syntactic structures and subsequent

4

reconstruction is possible and worth more attention, following Harris & Campbell(1995), Gildea (1998), Kikusawa (2002), Bowern (2008), Barðdal & Eythórsson(2012a–b), Barðdal (2015), Barðdal et al. (2013) and Eythórsson & Barðdal (2016).More specifically, we argue that on a Construction Grammar analysis, theGER+(NOM+)DAT structuremust be regarded as a construction, i.e. a form–meaningpairing of its own, resulting in the dematerializing of the comparanda problem forsyntactic reconstruction,as there isnomeaningfuldifferencebetween thesyntaxandlexicon modules in that framework. That is, since constructions are form–meaningcorrespondences in synchronic Construction Grammar, the leap to historical form–meaningcorrespondencesisminimal.ConstructionGrammar,inaddition,disposesofacomprehensive representational formalism in which all aspects of grammar can beexplicated, in turn allowing for the precise modeling of the form–meaningcorrespondencesneededtolayoutthedetailsofanunabridgedreconstruction.

Moreover, the strength of our reconstruction is bolstered by the fact that thestructuresin(2)shareimportantmorphosyntacticandsemanticproperties;attheveryleast,suchformsarenotordinarypassivesineachlanguageandthuscannotbederivedbyextendingthepassiveanalysis.Inotherwords,suchstructuresareexceptionalwithregardtothemorphosyntax(intheexpressionofmodality)andthesyntax(intheuseofa non-canonical argument structure) of the language, as well as being semanticallyspecific in theirmeaning. Themodalmeaning entailedby the verbal adjective furthermotivatestheargumentstructureofthisconstruction,i.e.theuseofthedativecaseforthe so-called “agent”. In contrast, the alleged demotion of an agent of a passive doesneither provide a satisfactory account of the argument structure of the GERconstructionnorofthisparticularuseofthedativecase.

Inthisarticlewethusproposeaconstructionof thetypeGER+(NOM+)DATforProto-Indo-European that expresses deonticmodality of obligation/necessity throughthe combination of a verbal adjective and a dative argument. While Hettrich (1990)proposes that the“dativeofagent”continuesaProto-Indo-Europeanconstruction,werefine his analysis to more accurately accommodate the facts of the data, which areneither consistent with a passive reading of the gerundive, nor, following from thepassivereading,anindependentagentreadingofthedativeargument.However,unlikeLuraghi (2016), we take the function of the dative in the individual languages to beinherited from Proto-Indo-European, rather than involving parallel independentdevelopmentsofthebeneficiaryfunctionofthedativeevolvingintotheobligedreferentinmodalcontextsinasmanyasfivebranchesofIndo-European.

In Section 2we present the facts of the GER+(NOM+)DAT construction acrossLatin,Greek,Sanskrit,Avestan,Tocharian,andLithuanian.WecomparethesefactswiththetraditionalanalysisinSection3,showingthattheyareindeedincompatiblewithatraditional passive analysis, althoughwe argue, likeHettrich but contra Luraghi, thatsucha construction canbe reconstructed, albeit in adifferentmanner.Wealsoargueagainsttwoadditionalanalysessuggestedinthe literature,namelythepossessiveandthebenefactiveanalyses.InSection4weproposeinsteadanovelreconstructionoftheGER+(NOM+)DAT that aligns the construction with ordinary oblique subjectconstructions observed across the early Indo-European daughter languages. Ouranalysis is presented in Sign-Based Construction Grammar, employing the SBCGformalism.Finally,Section5containsasummaryofthecontentandconclusionsofthisarticle.

5

2. Gerundivesinsixearly/archaicIndo-EuropeanlanguagesIn this section we investigate gerundives and their characteristics in Latin, AncientGreek,Sanskrit,Avestan,Tocharian,andLithuanian,respectively.These languagesarechosenfirstandforemostbecausetheyallhaveaverbaladjectivethatcombineswithadative argument, but fortuitously they also span the traditional east–west divisionwithin Indo-Europeandialectology,which in turn lends support to the reconstructionforProto-Indo-EuropeanthatweproposeinSection4below.Furthermore,thefirstfivelanguagesareconsideredeither“classical”or“ancient”Indo-Europeanlanguageswhilethelast,Lithuanian,isregardedasparticularlyconservative;thecharacterizationofthelanguagesasmoreancientor conservativealso servesasapointofdeparture for theproposedreconstructioninSection4.

Traditionally, gerundives are referred to as “future passive participles”, cf. theearlygrammarhandbooks,e.g.Smyth(1963:81)onGreek,wherethisformislabeledasthe “future passive participle” and Macdonnell (1916: 186) specifically on VedicSanskrit, where the terms “gerundive” and “future passive participle” are usedinterchangeably.The term“futurepassiveparticiple” is stillused in recent studiesonthegerundiveinclassicallanguages,e.g.Jasanoff(2006:195).Also,theterm“dativeofagent” is used to describe a dative argument with passive forms in, for instance,Goodwin(1900:252).Hence, thetraditionalanalysisofgerundives is indeedthatofapassive, thoughGoodwin (1900: 252) notes that this dative, in Greek at least, is alsousedtoindicatethe“agent’sinterestintheresultofthecompletedaction...”.

We take issue with the passive analysis of the GER+(NOM+)DAT constructionandshowbelowthatagentsofpassivesaregenerallynotmarkedwithadativecaseinthe six languages discussed, but with other morphosyntactic means, such as withinstrumentalsorprepositionalphrases; this in turnrulesoutapassiveanalysis.Thus,we compare the use of the dative with the case marking of agents in passiveconstructionsinthesixlanguagesdiscussedbelow.

In contrast, in each language discussed below, the GER+(NOM+)DATconstructionexpressesdeonticmodality,i.e.eitherobligationornecessity.Thelocusofmodality is not attributable to any specific lexical item or category found in theconstruction,butrathermustbeattributedtotheconstructionasawhole.Thatis,theGER+(NOM+)DAT construction is semantically non-compositional. We explore thenatureofthenon-compositionalityoftheGER+(NOM+)DATconstructionineachofthelanguagesunderinvestigationinthefollowingsubsections(foradiscussionofwhetherthemodalmeaningmaybederivedfromthegerundivesuffix,seeSections3.1and4.1below).

Anoteonterminologyisinorderhere.Asmentionedabove,thedativeargumentis traditionally labeled “dativeof agent” (cf.Green1913,Hettrich1990, 2014,George2005, Luraghi 2016). In what follows, we argue that this label is not accurate, as a)agents are not typically dative in the languages described and b) the argument itselfdoesnotevenexpressanagent.InordertoavoidanypotentialterminologicalconfusionaccompanyingtheuseoftermsforestablishedthematicrolesindescribingthedativeintheGER+(NOM+)DATconstruction,weprefertousetheterm“protagonist”toportraythemainparticipantof theverbal event.Our terminological choicehighlights the factthatthisargumentisthefocusoftheintendedactionaroundwhichtheeventisframed,and,whatismore,suchatermisnotinherentlylinkedtoanysemanticorsyntacticroleormorphologicalcase.

6

2.1 LatinAmajormorphological characteristicof theGER+(NOM+)DATconstruction is that themainparticipant,ortheprotagonist,oftheverbaleventisexpressedinthedativecase(Kühner & Stegmann 1955: 730, Menge 2000: 734). The examples in (3) show theGER+(NOM+)DAT construction with the so-called “dative of agent”, where thegerundive agrees with the (nominative) object of the verbal event, and not with thedative.(3) a. Consolandus hic mihist

console.GER.NOM.M this.NOM.M me.DAT+be.PRES.3SG ‘Imustconsolethisman’(Pl.Bacch.625)

b. Caesari universa uno tempore Caesar.DAT all.together.NEUT.PLone.ABL time.ABL

agenda eratdo.GER.NEUT.PL be.IMPF.3SG

‘Caesarhadtodoallthethingsatonetime’(Caes.Gall.2,20)

c. gerendus est tibi mosbear.GER.NOM.M be.PRES.3SG you.DAT custom.NOM.M.SGadulescentibus, Crasseyoung.DAT.PL Crassus.VOC

‘You,OCrassus,mustcomplywiththewishesofyoungmen’(Cic.DeOr.1.105)

The “dative of agent” also occurs in the constructionwithout a nominative, as in (4),wherethegerundiveisinthedefaultneuterform,theobjectisinthecaserequiredbytherootverbwhensuchaverbistransitive,andtheprotagonistoftheeventis inthedative:(4) suo cuique iudicio est utendum

own.ABL each.one.DATjudgment.ABL be.PRES.3SG use.GER.NEUT‘Eachmanmustusehisownjudgment’(Cic.Nat.Deor.3.1.)

Thedativeargumentistraditionallylabeled“dativeofagent”inLatin(seeforexampletheclassicschoolgrammarsbyAllen&Greenough1903:233–234andBennett1914:166ff,thetraditionalgrammarLeumann-Hoffman-Szantyr1972:96–97,andtherecenttypological/descriptive discussions such as Van Hoecke 2011: 15). However, asmentioned previously, this label is amisnomer—dative case is generally not used toexpresstheagentinLatin.Rather,theagentofpassivesisregularlyexpressedbymeansofprepositionalphrases,most commonlyaorab, togetherwithanounphrase in theablativecase,asshownin(5),orbythebareablativeifitdenotesaninanimatereferent.(5) Pompeius… a Catone aspere… est accusatus

Pompey.NOMby Cato.ABL harshly be.PRES.3SG accused.PPP‘Pompeywasharshlyaccused…byCato’(Cic.Fam.1.5b)

7

Theagentivephrasewitha/abandanablativecanco-occurwithagerundive,butonlyin order to avoid ambiguity in contexts in which the dative could also express abeneficiary argument, as in (6). The phrase ab eo is used in place of the dative ei,becauseinthiscontexttheutterancewiththedativecouldmean‘forhimwho…’.(6) Supplicatio ab eo qui ante dixit

supplication.NOM.F by him.ABL who.NOM beforespeak.PF.3SGdecernenda non fuitdecree.GER.NOM.F not be.PF.3SG‘Thesupplicationshouldnothavebeendecreedbyhimwhospokebefore’

(Cic.Phil.14.11)Thus, the “dative of agent” and the agentive prepositional phrase with a/ab are notinterchangeable, although they show some functional similarities. Basic passivesentences select for the ablative case followingapreposition,while gerundives selectfordatives.Thedifferencesindistributionofcasemarkersforsuch“agents”acrossthetwoconstructionsisgiveninTable1.Table1:TheDistributionofCaseMarkersfor“Agents”intheGER+(NOM+)DAT

constructionandthepassiveconstructioninLatin

GerundivesPassives

Latin

DAT a+ABLab+ABLABL

Furthermore, the so-called “dative of agent” is restricted almost entirely to theGER+(NOM+)DAT construction. It is occasionally attested with perfect passiveparticiples,butagain,notformarkingagentivitybuttosignalthenon-agentiveroleofthe protagonist (Gildersleeve& Lodge 1984). This is especially truewhenused in anadjectivalsense,asin(7)below:(7) puella… amata nobis quantumamabitur

girl.NOM loved.PPP.NOM.F us.DAT as.much.aslove.FUT.PASS.3SGnullanobody.NOM.F‘Thegirl[whowas]lovedbyusasnoonewillbeloved’(Cat.,VIII,5)

Inexample(7),thedativewiththepastperfectparticipleamata‘loved’issemanticallycloser toanexperiencer,rather thananagent, in that itdenotes thepersonwhofeelsloveforthegirlandtowhomthegirlisdear.Inthatsense,thisisnotanexampleoftheGER+(NOM+)DATconstructionexpressingobligationornecessity.

Hence, we argue that the dative that occurs in the GER+(NOM+)DATconstructionhas been improperly labeled as “dative of agent”. This dative designatesthemainparticipantoftheverbalact,theprotagonist,whoisaffectedbytheobligation,necessity,ordesiretoperformtheeventdenoted.Itisneitherapossessor,experiencernorbeneficiary.

Tosummarizethecontentofthissectionsofar:

8

● Latingerundiveshaveamodalmeaningofobligationand/ornecessity;● Latinhasaspecificdevice,i.e.thedativecase,formarkingtheprotagonistof

eventsdenotedbythesegerundives;● ThedativeisnotageneraldeviceformarkingtheagentinLatin.

Toconclude, theGER+(NOM+)DATconstruction isnot interchangeablewithapassiveconstruction.

Instead, the dative in the Latin GER+(NOM+)DAT construction is semanticallymoresimilartothesubject-likedativeofmodalpredicates,e.g.oportet‘itisnecessary’,licet‘itispermitted’,necesseest‘itisnecessary’,opusest‘itisnecessary’,etc.,whereana/abphrasecontainingthedemotedagentisalsoexcluded.Thatis,thedativecanonlyoccur as a direct argument of the predicate in such examples in Latin,without beingintermediated by a preposition. Compare the dative usage in the GER+(NOM+)DATconstructionsabovewiththatofthedativeargumentin(8)below:(8) huius nobis exempla permulta opus

this.GEN us.DAT examples.NOM very.many.NOM.PL need.NOMsuntbe.PRES.3PL‘Weneedverymanyexamplesofthis’(Cic.Inv.2.19.57)

In example (8), the dative argument nobis ‘us’ is more subject-like from a semanticstandpoint than thenominativeargumentexempla ‘examples’, in spiteof the fact thatexempla has the canonical subject case marking, nominative. That is, the participantwhofeelsneed,theprotagonist,isinthedativecase,andtheobjectoftheneedingeventisinthenominativecase.Theverbalpredicatein(8)iscompositional,namelyopussunt‘be in need’,whichmaybe an intransitive/one-placepredicate, occurring onlywith anominativesubject,oratransitive/two-placepredicate,occurringwithtwoarguments,i.e.adativesubjectandanominativeobject(forfurtherdiscussiononsuchalternations,seeSection4.1below).ThisisexactlyparalleltothecasemarkingoftheargumentsoftheGER+NOM+DAT constructionwhere the dativemarks the obliged participant, theprotagonist, and the nominative marks the object of the obligation. The similaritybetween these two constructions is striking and forms the basis of our analysis inSection4below.2.2 GreekTheanalogousAncientGreek formcorresponding to theLatingerundive is theverbaladjectivein-τέος.AsinLatin,thisformexpressesnecessityandtheparticipantwhoisobligedtoperformtheevent,theprotagonist,isexpressedwiththedativecase,exactlyas in Latin. The examples in (9) below exemplify this use of the dative in theGER+(NOM+)DATconstructionwherethegerundiveagreeswiththenominativeobjectoftheverbalpredicateandtheprotagonistismarkedinthedativecase.

9

(9) a. ὠφελητέα σοι ἡ πόλις serve.GER.NOM.F you.DAT the.NOM.F city.NOM.Fἐστίbe.PRES.3SG‘youmustbenefitthecity’(Xen.Mem.3.6.3)

b. ἡμῖν… πάντα ποιητέα

us.DAT all.NEUT.PL do.GER.NEUT.PL‘Wemustdoeverything’(Xen.An.3,1,35)

c. ποταμὸς δ᾽ εἰ μέν τις καὶ

river.NOM PTC whether PTC PRON.NOM and ἄλλος ἄρα ἡμῖν ἐστι another.NOMPTC us.DAT be.PRES.3SG διαβατέος οὐκ οἶδαcross.GER.NOM not know.PERF.1SG‘WhetherwemustcrossanyotherriverIdonotknow’(Xen.An.2,4,6)

In contrast, the examples in (10) illustrate the so-called “dative of agent” in theconstructionwithoutanominative:(10) a. φημὶ δὴ διχῇ βοηθητέον εἶναι

say.PRES.1SGPTC in.two.ways assist.GER.NEUT be.INFτοῖς πράγμασιν ὑμῖνthe.DAT.PL act.DAT.PL you.DAT.PL‘Isaythatyoumustgiveassistancetothecircumstancesintwoways’

(Dem.1.17)

b. οὔτε μεθεκτέον τῶν πραγμάτωνand.not have.a.share.GER.NEUT the.GEN.PLgovernment.GEN.PLπλέοσιν ἢ πεντακισχιλίοις (εἴη)more.DAT.PL than five.thousand.DAT.PL be.OPT.3SG‘Andnomorethanfivethousandshouldhaveashareinthegovernment’

(Thuc.8.65)

c. τὸν θάνατον ἡμῖν μετ᾽ εὐδοξίαςthe.ACC death.ACC us.DAT PREP honor.GEN αἱρετέον ἐστίchoose.GER.NEUT be.PRES.3SG‘Wemustchoosethedeathwithhonor’(Isoc.6.91)

UnlikeLatin, theGERconstruction inGreeksometimeshasanaccusativeargument inplaceofthedative,shownin(11)below:(11) ἰτέον ἂν εἴη θεασομένους, ἔφη

go.GER.NEUT PTC be.OPT.3SG look.on.ACC.M.PL say.AOR.3SG ὁ Σωκράτης

the.NOM Socrates.NOM ‘Socratessaid:“Weshouldgoandsee’”(Xen.Mem.3.11.1)

10

However, neither the accusative nor the dative are the ordinary cases of a demotedagent of passives in Ancient Greek. Rather, demoted agents are usually expressed bymeansofprepositionalphrases,mainlyὑπό+genitive,asshownin(12).(12) ὑπ᾽ Ἀχαιῶν/ ροτροπάδην φοβέοντο μελαινάων ἀπὸ

by Achaens.GENwith.headlong.speedfear.3PL black.GEN.PLawayνηῶνships.GEN‘By the Achaeans theywere put to flightwith headlong speed, away from theblackships’(Hom.Il.16.304)

Infact,theaccusativeisneverusedtoexpressanagent,andthe“dativeofagent”hasaverylimiteddistributioninAncientGreek(seealsoBarðdal&Danesi2014).TherearesporadicattestationsfoundinHomer,asexample(13)shows.(13) ἀνάσσονται δ᾽ ἐμοὶ αὐτῷ

rule.PRES.MP.3PL PTC me.DAT self.DAT‘(Thecities)areruledbymyself’(Hom.Od.4.777)

AndafterHomer,thisusageofthedativeoccursalmostexclusivelywiththeperfectandpluperfectpassive,cf.example(14).(14) πολλαὶ θεραπεῖαι καὶ παντοδαπαὶ τοῖς

many.NOM cures.NOM and of.every.kind.NOM the.DATἰατροῖς εὕρηνταιphysicians.DAT find.PERF.MP.3PL

‘Manyandvariouscureshavebeendiscoveredbythephysicians’(Hip.8.39)Ithasbeenarguedthatthisanomaloususeofthedativeinthepassivemayhavearisenwiththeperfectbecausetheperfectisstativeandnoteventive.Theperfectwouldthenformanoppositionwiththepresentandtheaoristwhichareeventiveanddynamicandmark the agent in the nominative case (George 2005: 79ff; on the expression of theagentinAncientGreek,seeLuraghi1986,1995,2000,Conti1999,amongothers).SeealsoDaues(2006)onthemediopassiveinHomericGreekwherepassivesareanalyzedas being functionally “extreme middles”. This entails that the reading of themediopassiveaseitherpassiveormiddle isderivedfromthetransitivityofthe lexicalverb.Onsuchananalysis,thedativein(13–14)couldbeanextensionofthedativeinDat-Nomconstructionswithmiddleverbs.However, irrespectiveofhowthisaberrantuseofthedativefordemotedagentsarose, this isnotaproductivepatterninAncientGreek.

The differences in distribution of case markers for “agents” across theGER+(NOM+)DATconstructionandthepassiveinAncientGreekisgiveninTable2.

11

Table2:TheDistributionofCaseMarkersfor“Agents”intheGER+(NOM+)DATconstructionandthepassiveconstructioninAncientGreek GerundivesPassives

AncientGreek

DATACC

ὑπό+GENDAT(sporadicinHomer)

To summarize, the behavior of the gerundive in Ancient Greek is very similar to thebehaviorof thegerundive inLatin.The combinationwithdativeand the fact that thegerundive itself is indifferent tovoice suggest that theGER+(NOM+)DATconstructionshouldnotbeanalyzedasapassiveconstructionatall.Again,thedativeargumentintheGER+NOM+DAT construction is not an agent, as traditionally argued, but is insteadsimilartothesubject-likedativefoundinconstructionswithmodalpredicates,e.g.μένει‘itremains’asin(15)below.

(15) τοῖς πᾶσιν ἀνθρώποισι κατθανεῖν μένει

the.DAT.PL all.DAT.PL men.DAT die.AOR.INF remain.PRES.3SG‘Allmenareexpectedtodie’(Eur.Fragments733;Stob.Flor.124.29)

Theverbμένειmeans‘tobedoomedtodosomething’whenoccurringwithadativeandinfinitive;ἀνάγκηἐστίmeans ‘it isnecessary’or ‘tobeobliged todosomething’whenoccurringwithadativeandaninfinitive,andlikewiseχρήἐστίmeans‘thereisneed’or‘tohavetheneedtodosomething’whenitoccurswithanaccusativeandandinfinitive,and finally ἐπαμμένει means ‘it awaits’ or ‘to be destined to do something’ whenoccurringwithadativeandaninfinitive.Inordinaryfiniteuseswithoutthedative,theverbμένειsimplymeansto‘remain’:(16) στήλη μένει ἔμπεδον

stone.NOM.SG remain.PRES.3SG in.the.ground‘stoneremainsintheground’(Il.17.434)

Themodalmeaningisthereforenotaninherentpartofthemeaningofthislexicalitem,but is only found when this verb is used with a dative combined with an infinitive,which in turn shows that both the modal meaning and the dative come from theconstructionitself.ItiswellknownthatmodalconstructionsoftenhavedativesubjectsandthisiscertainlythecaseinAncientGreek(cf.Danesi,Johnson,&Barðdal2016).

In addition,Greekgrammarhandbooks (cf.Goodwin1900, Smyth1963) agreethat the gerundive is interchangeablewith the impersonalδεῖ ‘there isneed’ togetherwithaninfinitive,asshowninexample(17)below:(17) θεοῖσι προσβαλεῖν χθονὶ/ ἄλλην δεήσει

gods.DAT put.to.INF hearth:ACC other.ACC.SG be.needful.FUT.3SGγαῖανworld.ACC‘Thegodswillhavetoaddanotherearthtoourworld’(Eur.Hipp.941)

12

ExactlyasintheGER+DATconstruction,theobligedparticipantisinthedativecaseandthepredicateπροσβαλεῖνδεήσει‘itwillhavetoadd’evenhasthesamemeaningasthegerundiveinacorrespondingGER+DATconstruction. The similarity between the Dative Modal Construction and the GER+DATconstructionisconfirmedbythefactthatδεῖ‘thereisneed’togetherwithaninfinitivecanalsocombinewithaso-called“accusativeofagent”,shownin(18)below,analogoustotheGER+DATconstructionin(11)above:(18) δεῖ ἐμὲ ἐν κοίτῃ σῇ

be.needful.PRES.3SG me.ACC in bed.DAT you.DATκατυπνῶσαιfall.asleep.INF

‘Imustfallasleepinyourbed’(Hdt.7.16C)

Tosummarizethecontentofthissection:● Greekgerundiveshaveamodalmeaningofobligationand/ornecessity;● Greekhasaspecificdevice,i.e.thedativecase(andtheaccusativecase,toamore

limited degree), for marking the protagonist of the events denoted by thesegerundives;

● ThedativeisotherwisenotageneraldeviceforexpressingtheagentinGreek.Toconclude,exactlyasinLatin,theGER+DATconstructionisnotinterchangeablewithapassiveconstructioninAncientGreek.2.3 SanskritIn Sanskrit, verbal adjectives analogous to the gerundives in Latin andAncientGreekare attested since the Ṛgveda, which is the most ancient Vedic Sanskrit document,datingapproximatelytothefirstmillenniumBCE.

Sanskritgerundivesarederivedwithvarious suffixes.Theearliestattestedarethefollowing(cf.Macdonnell1916:186–187).

(a) Theprimarysuffix-ya,attestedinallperiodsofthelanguage,e.g.dvéṣ-ya-‘tobehated’,śrú-t-ya-‘tobeheard’,-kṛ-t-ya-‘tobedone’.

(b) The secondary suffix -ay-ya, documented about a dozen times andrestricted to theṚgveda, e.g.dakṣ-ayya- ‘tobe conciliated’,pan-ayya- ‘tobeadmired’,vid-ayya-‘tobefound’.

(c) Thesecondarysuffix-én-ya,documentedaboutadozentimes,e.g. īkṣ-énya‘worthytobeseen’,dṛś-énya-‘worthytobeseen’,yudh-énya-‘worthyofbeingfoughtwith’.

(d) The suffix -tva; gerundives in -tva aredocumented about adozen timesand only in the Ṛgveda. They are probably derived from an infinitivalnounin-tubyaddingthesuffix-a,e.g.kár-tva- ‘tobedone’,jé-tva- ‘tobewon’,vák-tva-‘tobesaid’.

Additionally,therearetwosuffixeswhicharenotdocumentedintheṚgvedabutappearonlyinlatertexts.

13

(e) The suffix -tavyà; gerundives in -tavyà begin to appear only in theAtharvaveda.Theyareprobablyderivedwiththesuffix-ya-addedtoaninfinitival noun in -tu, e.g. jan-i-tavyà- ‘to be born’, hiṃs-i-tavyà- ‘to beinjured’.

(f) The suffix -anīya; gerundives in -anīya also begin to appear in theAtharvaveda.Theyarederivedwiththesuffix-īyaaddedtoanominalizedverb in -ana, e.g. ā-mantr-aṇīya- ‘worthy to be addressed’ ā-mántr-aṇa‘theactofaddressingsomeone’

Although theGER+(NOM+)DATconstruction ispassive-like in the sense that it showsorientationtowardsthepatient—thegerundiveagreeswiththepatient,nottheagent,oftheevent—Hock(1985–1986)arguesthatitisinexacttoassumeapassivevalueforthe early Vedic gerundive, and we indeed concur with Hock. In the Ṛgveda thegerundive is patient-oriented with transitive verbs, while with intransitive verbsgerundivesshownopassivesyntax,andarereferredtoas“activegerundives”byHock(1985–1986)whoalsoobserves thatbothconstructionscontinue tobeuseduntil thetimeofPāṇini,althoughthepatient-orientationpatternhadbecomemoreregular.

Turningnowto the“agent”of theeventexpressedby thegerundive, itmayberealized in the dative, instrumental, or genitive. The dative with the gerundive isattestedonlyintheṚgveda.Thisdativeargumenthasbeendescribedasoneof“agency”intheliterature(Delbrück1888:396ff,Brugmann-Delbrück1893–1911:II/2,558–559,Green1913:52ff).Classicalexamplesofthisusagearegivenin(19).(19) a. údyata-sruce bhavasi śravayyaḥ

raised-ladle.DAT be.PRES.2SG invoke.GER.NOM.SG.M‘You(Agni)aretobeinvokedbyhimwholiftstheladle’(RVI31,5b)

b. dakṣayyo yó dasvate

satisfy.GER.NOM.SG.M REL.NOM.SG.M sacrificer.DATdáma ahouse.LOC PTC‘Hewhoistobesatisfiedbythesacrificerinhishouse’(RVII4,3d)

Withrespecttothedistributionofthe“agent”cases,Hettrich(2014)hasrecentlyshownthattheinstrumentalistypicallyusedintherealismood,thedativeispreferredinothermoods, and the genitive occurs typically with participles without being particularlyassociated with deontic modality. It is well known from Indo-Aryan studies that thegenitive invades the territory of the dative in both Sanskrit and Avestan (cf. Joseph2012,2013),andthatthegenitivetakesoverthefunctionsofthedativeinditransitiveconstructions,aswellasinobliquesubjectconstructions.Whetherornotthevariationbetween the dative and genitive in GER constructions is a part of this generaldevelopmentcannotberuledout.

Turning to passive clauses, the instrumental is themost frequently used casemarkeroftheagentinpassivesinSanskrit,althoughthegenitive,andinafewcasestheablative, are also found (see Jamison 1979, Andersen 1986, Luraghi 1986, Hettrich1990,Oberlies2003,Lühr2004).

14

(20) a. uṣaucchántī ribhyate vásiṣṭhaiḥ Uṣas.NOMshining.PART.PRS.F.SGrasp.to.PRES.PASS.3SGVasiṣṭha.INS.PL

‘The shining Uṣas is being rasped to by the Vasiṣṭhas’ (RV VII 76, 7b,translationafterJamison2000andKulikov2012:498–499)

b. patyuḥ krītā husband.GEN.SG.M buy.PPP.NOM.SG.F ‘boughtbyherhusband’(MSI.10.11;Jamison1979:133) c. tvat-tovātavavāmātyairbhidyate jātu

you-ABLoryouror.ministers.INSkeep.secret.PRES.PASS.3SG alwaysmantritamadvice.NOM‘Is[your]advicealwayskeptsecretbyyourselforbyyourministers?’

(MBh2.5.14,citedfromOberlies2003:103)

However, a dative encoding of the agent of passives is unattested in the Ṛgveda:Delbrück(1893/I:300)notesthathehasbeenunabletofindthe“dativeofagent”witha finitepassiveverb.Theonlyevidenceforadativeofagent inVedicarecaseswherethe agent is realized by a clitic personal form, as in example (21); in such cases, thedativeandgenitivearesyncreticallyencodedandthusthe formisnotunambiguouslydative.(21) mátsy ápāyi te

rejoice.AOR.2SG.IMPER drink.AOR.PASS.3SGyou.DAT/GENmádaḥexhilarating.drink.NOM‘Enjoy!Youhavedrunktheexhilaratingdrink’(RVI1751a;Gaedicke1880:134)

Other instances of the “dative of agent” can rather be interpreted as a benefactivedative, as in example (22), where the clitic pronoun me ‘me’, already ambiguousbetweendativeandgenitive,canbeinterpretedeitherasanexperiencer/beneficiary,asin Geldner’s translation, or as an agent (Havers 1911: 10). Gelder’s translation alsocaptures the fact thatmorphological passives of verbsof perceptionnormally changetheir meaning to anticausative, instead of the expected passive meaning (cf. Kulikov2011).(22) prá me pánthā deva-yanā adṛśrann

PTC me.DAT/GENpath.NOM.PL god-going see.AOR.PASS.3PL ‘For/tomethepathsleadingtotheGodshavebecomevisible’(benefactive,after Geldner1952–1957),or:‘BymethepathsleadingtotheGodswereseen’(agent, afterHavers1911:10)(RVVII76,2)Likewise,in(23),thedativemánuṣāya‘man(kind)’isinterpretedasadativeofinterestby Delbrück (1888: 145), while Gaedicke (1880: 134) admits to the possibility ofreadingmánuṣāya as a dative agent: ‘die Götter sind gepriesen dem Menschen = espreistderMenschdieGötter’.

15

(23) deva stavante mánuṣāya sūráyaḥgod.NOM.PL praise.PRES.3PL.MID man.DAT lords/benefactors.NOM‘Godsarepraisedaslords/benefactorsformen’or:‘Godsarepraisedbymenaslords’(RV10,65,4d)

Insum,thecanonicalcasesfortheexpressionofagentsofpassivesarethegenitiveandthe instrumental, although some examples of ablatives are also found. The dative,however, is confined to gerundives, i.e. the GER+(NOM+)DAT construction, and thisconstructionisdocumentedonlyintheṚgveda.Inlatertextsitisunattested.Indeed,thecombination of the gerundivewith the dative, alongwith the fact that the gerundiveitselfisneitheractivenorpassive,suggeststhattheconstructionshouldnotbeanalyzedasapassive construction.Thedifferences indistributionof casemarkers for “agents”acrosstheGER+(NOM+)DATconstructionandthepassiveisgiveninTable3.

Table3:TheDistributionofCaseMarkersfor“Agents”intheGER+(NOM+)DATconstructionandthepassiveconstructioninSanskrit. GerundivesPassives

Sanskrit

DAT(onlyinRV)INSGEN

INSGEN(ABL)

Gerundives share with passives the fact that they are patient-oriented, namely, theymark the patient of the event (or the goal in the case of motion verbs) with thenominative,buttheydifferfrompassivesinthatactivegerundivesareattestedatleastintheṚgveda.Furthermore,gerundivescannotbeconsideredfunctionally identical topassivesinthattheypossessaspecificsemanticvalueofnecessityandobligation,whichisabsentfromordinarypassives(Hock1983).

Tosummarizethecontentofthissectionsofar:● Sanskritgerundiveshaveamodalmeaningofobligationand/ornecessity;● Sanskrithasaspecificdevice,i.e.thedativecase(inadditiontothegenitiveand

instrumental), for marking the protagonist of the event denoted by thesegerundives;

● ThedativeisnotageneraldeviceforexpressingtheagentinSanskrit.

Toconclude,exactlyasinLatinandAncientGreek,theGER+(NOM+)DATconstructionisnotinterchangeablewithapassiveconstructioninSanskrit.2.4 AvestanThe Avestan situation is, perhaps unsurprisingly, quite similar to the one in VedicSanskrit. Avestan gerundives are primarily formedwith the suffix -θa- (yuxθa- ‘to beharnessed’), -θβa- (staoθβa- ‘to be praised’), or -iia- (yesniia- ‘worthy of sacrifice’)(Skjærvø2003:130–131).Thesesuffixesconveythefollowingrelatedmeanings:a)theeventexpressedbytheverbmustbecarriedout,b)isallowedtooccur,orc)isworthy

16

of being carried out (Kanga 1891: 267). Avestan gerundives are patient-oriented butalso neutral with respect to voice; they can be interpreted as active or passivedepending on the context. The deontic modality reading, however, is unassailable insuchexamples(seebelow).

As inLatin,Greek,andVedicSanskrit, theprotagonistof theGER+(NOM+)DATconstruction is regularly expressed in the dative, as show in (24) below (cf. Skjærvø2003:132):(24) a. yesniiąm. Aŋuhe. astuuaite.

to.be.adored.ACC existence.DAT bony.DAT‘(ArdviSuraAnahita)istobeadoredbytheBonyExistence’(Yt.5.1)

b. tāca. Vohū… yā.…

these.NEUT goods.NEUT which.NEUTfrāiiaštuua. naire. asaone.to.be.worshipped.NEUT.PL man.DAT faithful.DAT‘(Wesacrificeto)allthesegoodthingswhicharetobeworshippedbythe faithfulman’(Yt.13.153)

c. aguštā. vaca. səṇghāmahī.

not.to.be.heard.NEUT.PL words.NEUT.PL proclaim.PRES.1PLaēibiiō.those.DAT‘Weproclaimwordstonotbeheardbythose’(Y.31.1)

The agent of passives, however, is consistently expressed with the instrumental inAvestan,asshownin(25)below(cf.Jamison1979:129ff.):(25) a. yāiš. gərəhmā. asāt.

REL.INS.PL Grehma.and.his.people.NOM.PL Right.ABL.SGvaratā.prefer.AO.INJ.M.3SG‘BywhomGrehmawithhispeoplewaspreferredtotheRight’(Y.32,12)

b. parštəm. zī. ϑβā.1

asked.PPP.NEUT. PTC you.INS‘(Thequestion)askedbyyou’(Y.43.10)

Theagentmayalsobeexpressedwith thegenitive,as inexample (26); thisoccurs inparticularwiththeverbalparticiple in -ta- (cf.Reichelt1909:259, Jamison1979:129ff.). 1 The Avestan form ϑβā is syncretic with the accusative. Jamison (1979: 138), however, reports thatHumbachtranslatesthisasatrueinstrumentalagent,whileInsler(1975)suggests“askedofyoubyus”.We followHumbach’s and Jamison’s analysis ofϑβā as instrumental, as it is unclear to us how a pastparticiple could select for an accusative object or how such an accusative should be interpretedsemantically.

17

(26) a. kahe. nō. iδa. nąma. PR.INT/INDEF.DAT.SG us.GEN here name.NOMāγairiiāt.praise.3SG.INJ.P‘Bywhomwillournamebepraised’(Yt.13,50;Reichelt1911:117)

b. fraŋrasiiānəm… pairišxvaxtəm. aiiaŋhahe.

Frangrasyan.ACC surround.PPP.ACC.M.SG iron.GEN‘Frangrasyanwhowassurroundedbyiron’(Y.11,7;Reichelt1911:171)

The use of the dative, however, appears to be restricted to the gerundive. There aresporadicanddoubtfuloccurrencesofthedativewiththepastpassiveparticiplein-ta-,as inexample (27).However, the functionof thedative isactuallyambiguoushere. Itcouldbereadasa“dativeofagent”(‘conceivedbyanotherman’),butitismorelikelyadativeofinterest(‘conceivedto/foranotherman’).(27) yā aom puϑrəm. baraiti.

REL.NOM.F.SG there child:ACC bring.PRES.3SGaniiahmāi. aršānāi. varštəm.another.DAT man.DAT produce.PPP.ACC.M.SG‘Shewhobringsthechildconceivedbyanotherman’or:‘conceivedforanotherman)’(Yt.17.58)

Thecombinationofthegerundivewithdative,togetherwiththefactthatthegerundiveitselfisneitheractivenorpassive,suggeststhattheAvestanexamplein(27)shouldnotbeanalyzedasapassiveconstruction.Gerundivessharewithpassivesthefactthattheyarepatient-oriented,buttheydifferfrompassivesinthattheyareneutralwithrespecttovoiceand, inaddition,possessaspecificsemanticvalueofnecessityandobligation,whichisnotaninherentpartofpassives.

The differences in the distribution of case markers for “agents” across theGER+(NOM+)DATconstructionandthepassiveisgiveninTable4.

Table4:TheDistributionofCaseMarkersfor“Agents”intheGER+(NOM+)DATconstructionandthepassiveconstructioninAvestan. GerundivesPassives

Avestan

DAT

INSGEN

ThissectionhasillustratedtheuseofgerundivesinAvestan,showingthat:

● Avestangerundiveshaveamodalmeaningofobligationand/ornecessity;● Avestanhasaspecificdevice,i.e.thedativecase,formarkingtheprotagonistof

theeventdenotedbythesegerundives;● ThedativeisnotageneraldeviceforexpressingtheagentinAvestan.

Insum,exactlylikeinLatin,AncientGreek,andSanskrit,thecombinationofthedativeandthegerundiveisnotinterchangeablewithapassiveconstruction.

18

2.5 TocharianInbothTocharianAandB,anl-suffixisusedtoderivegerundives,eitherfrompresentstems (Gerundive I) or from subjunctive stems (Gerundive II). Gerundive I expressesnecessity,whileGerundiveIIexpressespossibility.Bothcanbeemployedattributivelyorpredicatively. In the latter case, thecopula in thepresentcanbeomitted (Bubenik1997:139).GerundiveIcorresponds inmanyrespectstoLatinandGreekgerundives,but Gerundive II is less comparablewith the other Indo-European gerundives: it is asecondaryderivation(fromthesubjunctivestem)andspecificallygrammaticalizesthemeaningofpotentiality.

ForGerundiveIinTocharian,theverbalargumentthatreferstotheprotagonistisassignedgenitivecase,as in theexamples in (28)below.This isunsurprisinggiventhe fact that in Tocharian the genitive has taken over the functions of the dative (cf.Krause&Thomas1960:82,Adams2011).

(28) a. ku se wesäñ tanneṃ yamaṣälle

what.NOM.M=N we.GEN thereby do.GER.NOM.M.SG‘Therebywhatshouldwedo?’(Toch.B:Š102.6a5;Werner1951:20)

b. saṅkästeryāñce träṅkӓl

Saṅghasthavirī.GEN.F.SG speak.GER.NOM.M.SG‘Saṅghasthavirīmustspeak’(Toch.A:Nr.414a1;Werner1951:19)

In Gerundive II, like in Gerundive I, the protagonist of the event is marked in thegenitivecase:(29) a. wsālu yetweyntuwaṣlaṃ ṣeñc-äm

clothes.NOM.PLdecoration.NOM.PL.Fcover.GERII.NOM.PL.F them.GEN‘Theycanwearclothesanddecorations’

(Toch.A;Krause&Thomas1960/I:187)

b. ---wrasaśśimā kälpāl naṣman.GEN.PL not reach.GERII.NOM.M.SG be.PRES.3SG‘---mencannotreach/---isnotreachablebymen’

(Toch.A:Nr.14b3;Werner1951:30)The genitive is one of the cases used for the agent in Tocharian, although this use isrestrictedtoparticularlinguisticscontexts,mainlyfoundwithnon-finiteverbforms,i.e.preterite participles, gerundives, and infinitives, and very rarely with finite passiveforms(Krause&Thomas1960:82–83).Hence,thegenitiveisusedtomarktheagentinpassive-likeconstructionsinTocharian,butnotinthepassiveproper.Furthermore,theinfinitive is almost interchangeable with the gerundive of necessity (Bubenik 1997:137),cf.example(30)below:(30) te päkṣälle ṣӓlype lipātsi

this.NOM cook.GER.NOM fat.NOM remain.INF‘Thishastobecooked,thefat(must)remain’(Krause&Thomas1960:184)

19

Asforpreteriteparticiples,sincetheyareunmarkedforvoice,anactiveinterpretationhasbeenproposedinthosecaseswheretheagentisexpressed(Bubenik1997:136),asin(31)below,althoughotherreadingsarealsoplausible:(31) ñi se pilko ste

I.GEN this.NOM.SG.M sight.NOM.SG.M view.NOM.SG.Mprākr=eṅkube.PRES.3SG=firmly+seize.PART.PRET.NOM.M.SG‘I have firmly adopted this view’ (Bubenik1997:136)or ‘This viewofmine isfirmlyadopted’or‘Asforme,thisviewisfirmlyadopted’

Thegenitivethereforehasa limiteddistribution inTocharianconstructions,whilethecasescommonlyemployedtomarktheagentinpassiveconstructionsaretheperlativeand the instrumental. Tocharian B uses the perlative in all contexts since it lacks aninstrumental case,while TocharianA uses the perlativewith animate agents and theinstrumentalwithinanimateagents,asin(32)below:(32) mā poryo tskāṃsaṃtär…

not fire.INS.SG burn.PRES.MID/PASS.3PLmā lāñcsā pärtsi yāteñcnot king.PERL.PL take.away.INF can.PRES.3PL

‘Theycannotbeburntbythefire…theycannotbetakenawaybythekings’ (Krause&Thomas1960:82–83,Bubenik2006:319)ItisclearfromthisoverviewthatTocharianisquitesimilartotheotherIndo-Europeanlanguagesexamined.Theonlydifferenceisthatthe“agent”ofgerundivesinTocharianisnotmarkedwiththedativebutwiththegenitive.However,sincethedativeandthegenitivemergedinTocharianandthegenitiveformreplacedthedativeform,itappearsreasonabletoassumethatthe“genitiveofagent”continuesanearlier“dativeofagent”construction.Thedifferencesinthedistributionofcasemarkersfor“agents”acrosstheGER+(NOM+)DATconstructionandthepassivearegiveninTable5.

Table5:TheDistributionofCaseMarkersfor“Agents”intheGER+(NOM+)DATconstructionandthepassiveconstructioninTocharian. GerundivesPassives

Tocharian

GEN

PERINS (only in TocharianA)

Insum,Tochariangerundiveshaveamodalmeaningofobligationand/ornecessityandseemtobeneutralwithrespecttovoice(theirinterpretationdependsonthecontext).Tocharianhasaspecificdevice, i.e.thegenitive,forexpressingtheso-called“agent”ofthesegerundives.Thegenitive,asameansofexpressingtheagent inTocharian,hasalimited distribution, as the canonical marking of the agent in passive constructionsinvolvestheperlativeandtheinstrumental.

ThissectionhasillustratedtheuseofgerundivesinTocharian,showingthat:

20

● Tochariangerundiveshaveamodalmeaningofobligationand/ornecessity;● Tocharianhasaspecificdevice,i.e.thegenitivecase,formarkingtheprotagonist

oftheeventdenotedbythesegerundives;● ThegenitiveisnotageneraldeviceforexpressingtheagentinTocharian;● Thegenitivehasalsotakenoverthefunctionsofthedative,whichdisappearedin

Tocharian.Exactly like in Latin, Ancient Greek, Sanskrit and Avestan, the combination of thegenitive and the gerundive is not interchangeable with a passive construction inTocharian.

2.6 LithuanianLithuanian grammars list the “participle of necessity” among its past participles(Kurschat 1876: 286, Schleicher 1896: 100, Leskien 1919: 201, Senn 1966: 380,Ambrazas 1997: 328). This participle is often compared to the Latin and Greekgerundive(e.g.Leskien1919:189;Senn1966:380),andtheyturnouttobeverysimilarindeed. The Lithuanian participle of necessity is a verbal adjective derived from theinfinitival stem with the suffix -tinas. It occurs with both transitive and intransitiveverbs,expressesnecessity,andcanbeemployedattributivelyorpredicatively.

The protagonist, the participant obliged to carry out the event denoted by thegerundive,isassigneddativecase,cf.(33).(33) a. tiẽ mẽtai yrà visíems

that.NOM.PL.M year.NOM.PL.M be.PRES.3SG all.DAT.PLatmintinìremember.GER.NOM.PL.M‘Everybodyshouldrememberthoseyears’

(Senn1966:416,Bauer2000:217)

b. šìtas dalỹkas ir táuthis.NOM.SG.M thing.NOM.SG.M and you.DAT.SGžinótinasknow.GER.NOM.SG.M‘Andyoushouldrememberthisthing’(Senn1966:416,Bauer2000:217)

In Lithuanian, as in the languages examined earlier, the case used for the “agent” ofgerundives/participlesofnecessityisnotthecasethatisnormallyusedforexpressingtheagentofregularpassives.Theoppositionbetweentheactivevoiceandthepassivevoiceisexpressedmainlybyparticiples.Bothperiphrasticformswithactiveparticiplesandsimplefiniteverbformssignifyactivevoice,whilethepassivevoiceisexpressedbyperiphrastic passive forms containing the passive participles and the copulabuti ‘be’.The example in (34a) shows a transitive active sentence,with (34b) representing itspassivevariant(Ambrazas1997:276).(34) a. mótina mylejo dùkterį

mother.NOM love.PAST.3SG daughter.ACC.SG‘Themotherlovedherdaughter’

21

b. dukte bùvo (mótinos) daughter.NOM.SG be.PAST.3SG mother.GEN.SGmylimàlove.PPP.NOM.SG.F‘Thedaughterwasloved(byhermother)’

In passive constructions, the agent is usually marked in the genitive, as in example(34b).Intheoldesttextsagentsarealsofoundgovernedbyprepositionalphrases, forinstance per together with an accusative and nuog (= nuõ) together with a genitive’(Senn1966:376),exemplifiedin(35a–b),respectively.(35) a. pastatitas nuog Pona Christusa

establish.PPP.NOM.SG.M by Lord.GEN.SG.M Christ.GEN.SG.M‘EstablishedbyChrist,theLord’

(OldLithuanianexample;Schmalstieg1987:180)

b. paraschits per Daktara Martinawrite.PPP.NOM.SG by doctor.ACC.SG Martin.ACC.SGLuteraLuther.ACC.SG‘writtenbyDr.MartinLuther’

(Vilentas’Catechism,titlepage,6;Schmalstieg1987:288)Although the agent of passive constructions is usually marked in the genitive or isgoverned by prepositional phrases, the dative is also found, both in the modernlanguageandinoldertexts,withpresentpassiveparticiples.Itisremarkable,however,that this combination of the dative with a present passive participle is possible onlywhentheparticipleexpressesthemeaningofpossibilityassociatedwithageneralizedevent(Schmalstieg1987:24–25,Ambrazas1997:355),cf.example(36).(36) buday pikti regimi wisiemus

manners.NOM.PL evil.NOM.PL see.PART.PRES.PASS.NOM.PLall.DAT.PL‘Evilmannersvisibletoall’

(Sirvydas’PunktaysakimuI25329,Schmalstieg1987:25)Thedativeisalsofoundinconstructionswithaninfinitive,againtoexpressthenotionof possibility, necessity, or obligation (Senn 1966: 469, Schmalstieg 1978: 224,Ambrazas1997:374),asintheexamplesin(37)below:(37) a. ką mán darýti?

what.ACC.SG I.DAT do.INF‘WhatshallIdo?(Senn1966:469)

b. nesutráukti mán jų

not+compress.INF I.DAT they.GEN‘Icannotcompressthem’(Senn1966:469)

The differences in the distribution of case markers for “agents” across theGER+(NOM+)DATconstructionandpassivesisgiveninTable6.

22

Table6:TheDistributionofCaseMarkersfor“Agents”intheGER+(NOM+)DATconstructionandthepassiveconstructioninLithuanian GerundivesPassives

Lithuanian

DAT

GENper +ACC(inoldesttexts)nuõ+GEN(inoldesttexts)DAT(onlywithpresent

passiveparticiplesexpressingpossibility)

Itisthusclearthatthedistributionofthe“dativeofagent”inLithuanianisrestrictedtoonlymodal contexts, involving necessity or possibility. These facts suggest that thecombination of the dative and the gerundive is not interchangeable with a passiveconstruction.Thedativedoesnotexpressanagentbutissimilartothedativesubjectofmodal predicates like reiketi ‘need’, as in (38) below,where the protagonist is in thedative,whiletheobjectneededisinthegenitive.(38) man reĩkia kambario

I.DAT need.PRES.3SG room.GEN‘Ineedaroom’(Schmalstieg1987:220)

ThissectionhasillustratedtheuseofgerundivesinLithuanian,showingthat:

● Lithuaniangerundiveshaveamodalmeaningofobligationand/ornecessity;● Lithuanianhasaspecificdevice,i.e.thedativecase,formarkingtheprotagonist

oftheeventdenotedbythesegerundives;● ThedativeisnotageneraldeviceforexpressingtheagentinLithuanian.

Insum,exactlylikeinLatin,AncientGreek,Sanskrit,Avestan,andTocharianabove,thecombination of the dative and the gerundive is not interchangeable with a passiveconstruction.3. AgainstalternativeaccountsInthefollowing,wediscussthreeearlieranalysesoftheDATintheGER+(NOM+)DATconstruction: a) the traditionalist (passive) account,b) thepossessiveaccount, and c)the benefactive account.We reject all three and suggest instead an account based onDanesi’s (2013) analysis where the GER+(NOM+)DAT construction is regarded as asubconstruction of the ordinary Oblique Subject Construction found across the Indo-Europeanlanguages,ancient,medieval,andmodern.3.1 Thetraditionalist(passive)accountOnthetraditionalaccountoftheGER+(NOM+)DATconstruction,itisassumedthatthemodal reading accompanying the construction is obtained from a passive-like formcombinedwithanovertlyexpresseddemotedagent(Green1913,Hettrich1990,George

23

2005). Even the traditional grammar definition of the gerundive as a “future passiveparticiple”highlightstheperceivedpassivityoftheverbaladjective.Thegerundiveisingeneralregardedaspassiveinnature:itrequiresonlyoneargumentwhichispatient-ortheme-like,exactlylikewithpassives,whiletheprotagonistmaybeleftunexpressed(ornot,asinGER+(NOM+)DAT).

ItisimplicitinthepassiveanalysisoftheGERconstructionthatitmuststandinasystematicrelationtothecorrespondingconstructionwithoutthedative(see1above).This is indeed reminiscent of passives and the non-compulsory occurrence of thedemotedagent.However,considertheexamplesin(39–41)below:(39) English a. Ilackfood

b. Foodislacking.(40) Latvian(Berg-Olsen2009:185,187)

a. Viņaitrūkstpieredze. she.DATlacksexperience.NOM ‘Shelacksexperience.’

b. …ka kaut kas pietrūkst. thatsome thing.NOM lacks ‘...thatsomethingislacking’(41) Icelandic a. Hennier þettaleyfilegt. she.DATis this.NOMallowed ‘Sheisallowedthis.’

b. Þetta erleyfilegt. this.NOMisallowed

‘Thisisallowed.’TheexamplesabovearefromModernEnglish,ModernLatvian,andModernIcelandic,respectively, and they show clearly that two-place argument structure constructionsmay systematically alternate with corresponding one-place argument structureconstruction, where the object of the transitive corresponds with the subject of theintransitive variant,without theneed for apassive analysis. Instead, it iswell knownfrom the study of argument structure that argument structure constructions mayappear in a systematic relation toother argument structure constructions, dependingonhowtheeventisconstrued(cf.Croft1998,2012,Barðdal2001,2015).In(40–41)asystematic alternation between DAT-NOM and NOM is presented from Latvian andIcelandic, respectively. The occurrence of the gerundive with or without the dativeargument does therefore not necessarily call for a passive analysis with an optionaldemotedagent. Furthermore,we have established the following facts for Latin, Ancient Greek,Sanskrit,Avestan,Tocharian,andLithuanian: ● Gerundiveshaveamodalmeaningofobligationand/ornecessity;● Each language has a specific device, i.e. the dative or a functionally equivalent

case,formarkingtheprotagonistoftheeventdenotedbythesegerundives;

24

● Thisspecificdeviceisnotageneraldeviceforexpressingtheagentinpassivesineachofthesesixlanguages.

Given these facts, a passive analysis of theGER+(NOM+)DAT construction is far fromsatisfying for the following reasons. First, the “dative of agent” is in a differentmorphological case than the usual demoted agent of passives in all the languagesinvestigated above. Second, it has been shown that when passives exhibit modalproperties crosslinguistically, the relevant modality is usually potentiality (Narrog2010). As is made very clear above, however, the GER+(NOM+)DAT constructionexpressesnecessityandobligation,butnotpotentiality.

In conclusion, the GER+(NOM+)DAT construction is structurally andsemanticallyverydifferentfromapassiveconstruction. Infact, thepassiveanalysisoftheGER+(NOM+)DATconstructionappearsquitesimplisticandnotproperlyrootedintheactualpropertiesoftheconstructionitself.Thepassiveanalysisshouldthereforebeabandoned as such, and following this, the traditional description of the dativeargumentasbeing“dativeofagent”. An attempt to rescue the passive account could perhaps bemade through ananalysisof theGER+(NOM+)DATconstructionasaspecial typeofpassive.Onsuchananalysisthepassiveandthemodalsemanticsmightbetakentobederiveddirectlyfromthe gerundive suffix, since -ndus in Latin, for instance, clearly does not belong to theordinarypassiveparadigm.Thereis,however,onemajorproblemwithsuchanaccount.SincetherelevantgerundivesuffixesareconfinedtotheGER+(NOM+)DATconstructionandarenotfoundoutsideofit,thisparticularproperty,namelypassive+modalreading,cannotbederivedfromanyotheraspectofthelanguageassuchandmustthereforebeassumed to be specific for this construction. In essence, this means that this specialpassivehastobestipulatedandaccountedforassuchinthegrammaroftherelevantlanguages,whichbasicallyamountstoanon-compositionalandconstructionalanalysisofthetypethatwesuggestinSection4below.

3.2 Thepossessiveaccount

As an alternate hypothesis of the origin of the GER+(NOM+)DAT construction, Bauer(2000)observesthatdifferentverbalforms,includinggerundivesbutalsoparticiplesin-to-, invariant participles, and infinitives, in constructions selecting for a dativeargumentmayberelatedtotheDativePossessiveConstruction.Considerexample(42)belowoftheDativePossessiveConstruction,wherethedativedesignatesthepossessorandthenominativethepossessed.(42) ubi tempus tibi erit

wheretime.NOM you.DAT be.3SG.FUT‘whereyouwillhavetime’(Ter.Eun.485,fromBauer2000:180)

Bauer (2000) points to the strong structural parallels between the two constructionsand argues that they should both be considered “impersonal” constructions asexpressions of transitive-independent syntax. She further claims that theseconstructions are residues of an earlier language stage at which transitivity wasconveyedbynominalagreement.

25

While we recognize the formal and functional similarities between the DativePossessiveConstruction and theGER+(NOM+)DAT construction,we argue that one isnothistoricallyderivedfromtheother.It iscertainlytruethatpossessivesmaybethesource to somemodal constructions (Narrog 2010), but such a scenario presupposesstructural identity which is not found here, as there is a gerundive in theGER+(NOM+)DATconstruction,butBE+NOUNinthepossessiveconstruction.

ThereareevenfurtherproblemswiththeassumptionthattheGER+(NOM+)DAThasdeveloped from theBE+NOUNconstruction, asBauerherself recognizes—namelythatstructuralsimilaritydoesnotimplyidentityoffunctions.Thetwodativessharethesame formbut not the samemeaning: in one case the dative is the possessor, in theotherthedativeisthepotentialagent,orinourterminology,theprotagonist.Therefore,any claims about the Dative Possessive Construction being the source of theGER+(NOM+)DATconstructionrequiresmoreevidenceandelaborationinordertobeconvincing.

3.3 ThebenefactiveaccountAnotherrecentanalysisissuggestedbyLuraghi(2016),whoargues,followingHettrich(1990) that the “agentive” use of the dative developed from beneficiary datives but,contraHettrich, that this development has takenplace independently in the differentdaughter languages. This last claim is based on three facts: a) that Sanskrit showsvariationbetweendative, instrumental, and genitive, b) that the interpretation of thedative in theGER+(NOM)+DAT construction is context-dependent (cf.Hettrich1990),suggestingalowdegreeofconventionalizationaccordingtoLuraghi(2016:20–21),andc) that there are no examples of the GER+(NOM)+DAT construction with -tva in theṚgveda.Wenowdiscusseachinturn:

Startingwith the first itemizedpoint above, namely the variation between thedative, instrumental,andgenitive inSanskrit,wehavealreadyreported inSection2.3aboveonHettrich’sfindings(alsocitedbyLuraghi2016)thattheinstrumentalseemstobe typicallyused todenoteevents in the realis, thedativewithothermoods, and thegenitive with participles. Hence, the distribution of the dative, instrumental, and thegenitivewithgerundives inSanskrit isclearly functionallyandstructurallymotivated,underminingLuraghi’sclaimthattheGER+(NOM)+DATstructureisnotaconstructionofitsowninSanskrit.

Turning to the second itemized point above, namely the alleged lack ofconventionalityofthe“dativeofagent”constructioninSanskrit,thisclaimisbasedonthefactthatcontextisneededtodisambiguatebetweendifferentreadingsofthedativewithgerundives.Insomeinstances,thedativeisabeneficiary,whileinothercases,itisthereferenttakingontheobligationexpressedbytheevent,ourprotagonist.Thereis,however,amajorproblemwiththeclaimthattheneedforcontextfordisambiguationsignalslackofconventionalization.Thereasonforthisissimple,namelythatfunctionalambiguitydoesnotequatelowdegreeofconventionalization,itonlyspeaksforaone-to-many mappings of a morphosyntactic form. Such one-to-many mappings may bearoundinlanguagesforcenturies,whichinturnshowsthatone-to-manymappingsdonotnecessarilysignalthatadevelopmentmustberecent.

Itiswellknownfromlexicologyandlexicalsemantics,forinstance,thatcontextisneeded todisambiguatewordsenses thatareotherwiseambiguous (Kilgariff1997,Swaab,Brown&Hagoort2003,Nash2008);thisisirrespectiveofwhetherthedifferent

26

senses have arisen recently orwhether they have existed for a long time. InModernEnglishthepastparticlewithapresenttense‘be’,asin“thefishisfried”isambiguousbetween a resultative and a passive reading, again demanding context to decidebetween the two readings. The periphrastic passive can be traced at least to Proto-Germanic, including the ambiguity resulting from theuse of thepresent tenseof ‘be’.This ambiguity must thus have existed since the periphrastic passive arose duringprehistorictimes.

In Modern Swedish, moreover, the verbal suffix -s is ambiguous between apassive and a middle reading, requiring context to disambiguate between the two(Barðdal & Molnár 2003). This verbal suffix originally developed from the reflexivepronoun sik ‘self’, first to a clitic and then later to a suffix (Ottósson 1992). Thisdevelopmentwascompletedaround1200 inOldSwedishand it isgenerallyassumedthat the middle meaning is derived from its reflexive/reciprocal meaning. The firstunambigousexamplesofthes-formwithapassivereadingarefoundinOldSwedishlawtexts(Wessén1965:167ff.),forinstanceÖstgötalagenwhichdateto1290.This,inturn,meansthatthetworeadings,themiddleandthepassivereadings,haveexistedsidebysideformorethansevencenturiesinthehistoryofSwedish.

In none of these instances can it be argued that these constructions are non-conventionalized and the same is true forword senses.Hence, these examples showthatonemorphosyntacticformcanbemappedontomanyfunctionsandsuchmappingsmayexistnotonlyforcenturies(-sinSwedish)butalsomillennia(periphrasticpassivein Germanic). Therefore, the claim that the GER+(NOM)+DAT construction is non-conventionalized due to the need for context for disambiguation in Sanskrit isabsolutelywithoutanymeritwhatsoever.

The relevance of the third itemized point above, i.e. the fact that there are noexamplesoftheGER+(NOM)+DATconstructionwith-tvaintheṚgveda,isuncleargiventhatSanskrit -ya-, cognate to theAvestan -iia-, is found in theṚgveda, i.e. theearlieststagesof Indo-Aryan,with adative.Thus, the sameor a cognate form is found in theṚgveda,aswellasintwodaughterlanguagesoftheIndo-Aryanbranch,namelyClassicalSanskritandAvestan.Thisdistributionclearlyspeaks for inheritance.Taken together,Luraghi’s assumption that the “dative of agent” construction is an independentdevelopmentinthedaughterlanguagesappearsasillfounded.

It isfurthermoreassumedinLuraghi’sanalysis(2016:26)thatthedativeitselfexists as an independent category in the early and ancient Indo-European languages,irrespectiveoftheconstructionsitoccursin.Thisviewcanbetracedbacktoatleastthestructuralists,cf.Hjelmslev(1935)andJakobson(1936),whereforinstancethedativeis studied across different contexts and an abstractmeaning is assigned to it on thebasisofitsmeaningindifferentconstructions.

Theproblemwiththistraditionalistandstructuralistapproachis thatdifferentusesofthedativearelumpedtogether,includingtheusesofthedativeindifferentandevenunrelated constructions (for a general criticismof sucha LUMPINGapproach, seeCroft2001:65–75andparticularlywithregardtocasemarkingandsemanticroles,seeBarðdal2003,Barðdaletal.2013).Inordertoillustratethis,considertheexamplesin(43)below,all instantiatingdativepersonalpronounsindifferentIndo-Europeancaselanguages:

(43) moi (AncientGreek) mihi (Latin) tebe (OldRussian)

27

táu (Lithuanian) mér (Icelandic dir (German)In isolation,thesedativeformshavenomeaning.Theyareonly imbuedwithmeaningthroughsyntacticcontext,ascanbeseenbythefactthatthesameformcaninstantiatemultiple functions, e.g. experience, recipiency, possession, benefaction, malefaction,absolutive (for Gothic), etc.While certain functions of the dative could be viewed asprimary, this is not because these represent any default meaning of the dative, butratherbecausethatsyntacticcontextofthedativeismoresalientormorefrequentinalanguage.

Hence, we believe that the meaning of datives comes directly from theconstructions they instantiate, and that this meaning does not exist independent ofconstructions. This implies that the alleged benefactivemeaning of the dative comesfrom theditransitive constructionwhere the indirectobject is in thedative casewithverbs like ‘give’, ‘bring’, ‘send’, ‘say’, etc. In the same vein, the alleged experiencermeaning of the dative comes directly from dative subject verbs which often expressemotion, such as ‘like’, ‘feel good/bad’, ‘be bored’, among others. Thus, on our view,therecannotbeanyindependent“benefactive”or“experiencer”meaningofthedativeinthelanguagesunderinvestigationfromwhichthedativeintheGERconstructionhasdevelopedfrom.

Luraghi (2016: 134) further lays out how the connection between thebeneficiarydative and thedativeof theprotagonist inmodal constructionsmayhavearisen, namely through a metaphorical extension where a) beneficiaries have beenassociatedwithrecipients(mostlikelythroughmetonymy,weinfer),andb)obligationsare perceived of as transferred objects. Luraghi argues that this last assumption isrooted in a metaphor OBLIGATIONS ARE OBJECTS TRANSFERRED. An immediateproblem is that the term beneficiary implies that the relevant participant receivessomethinggood,i.e.benefitsfromthetransfer,whichisfarfromalwaysthecasewithanobligation.

Moreover,itisunclearwhattheroleofthisallegedmetaphoris,especiallygiventhat it does not seem to be a conceptualmetaphor in the sense of Lakoff & Johnson(1980),nordoesitseemtobeattestedoutsideofthisparticularlinguisticdomain.Thatis,noindependentmotivationseemstoexistforthestipulationofthismetaphor;rather,thisallegedmetaphorisassumedtoexistonthebasisoftheverydataitisintendedtoexplain,andassuchthisargumentationiscircular,tosaytheleast.Asimplersolution,we believe, takes the GER+(NOM+)DAT construction to be a subconstruction of theordinary Oblique Subject Construction, found more widely across Indo-Europeanlanguages.WelayouttheargumentsforthispositioninSection4below.

Finally, Luraghi’s analysis assumes that the gerundive construction was anindependent parallel development in each language. That is, the same metaphoricalextensionwasinstantiatedinfivesubbranchesandsixdifferentdaughterlanguagesofIndo-European.Suchaclaimisnotimpossible,butgiventhestructural,semantic,and,insomecases,etymologicalsimilaritiesbetweentheGER+(NOM+)DATconstructioninthe languagesdetailedabove,we findthisclaimtobe inherently lesssimple thanonethatassumesacommonIndo-Europeanorigin.Whilewedonotproposetoreconstructthe “dative of agent” as Hettrich (1990) suggested, we find the fact that allGER+(NOM+)DATconstructionshavethesamemodalmeaning,thesametypeofverbal

28

adjective,andthesamecase(dativeorafunctionallyequivalentone)todesignatetheprotagonistrequires—evendemands—aProto-Indo-Europeansource.

To conclude, the GER+(NOM+)DAT construction is indeed quite puzzling; thecombinationofagerundivebeingusedfinitelywithadativeargument isnotatypicaltype of finite clause. Second, there is a clear modal meaning associated with theconstruction,whichcannotbeattributedtoanymodalelementintheclause,neitheramodal verbnor amodalparticle.Aswehave suggestedelsewhere (Barðdal&Danesi2014), thismodalmeaningmust be directly attributed to the gerundive constructionitself. In the following section, we provide our analysis and reconstruction of theGER+(NOM+)DATconstructionforProto-Indo-European.4. ReconstructionBelow we analyze the GER+(NOM+)DAT structure as a construction in the sense ofConstruction Grammar, i.e. as a form–meaning correspondence (Fillmore, Kay, &O’Connor 1988, Jackendoff 1997, Kay & Fillmore 1999; Goldberg 1995, 2006, Croft2001,interalia).Thisconstruction,webelieve,isofanIndo-Europeanorigin,belongingto thebroader category ofOblique Subject Constructions in Indo-European, and is assuchreconstructableforaproto-stage.

On a constructional approach to language, constructions are form–meaningpairings larger than words and as such they constitute the basic building blocks oflanguage. In contrast to other current syntactic frameworks, Construction Grammardoesnotassumeastrictseparationbetweenthesyntaxandthelexicon.Constructionsdiffer fromeachotherwith regard to their complexityandschematicity, varyingonascale fromsimpletocomplexandsubstantivetoschematic(Croft2001,Croft&Cruse2004,Barðdal2001,2008,2013).Togiveanexample,theworddogismorphologicallysimpleandsubstantivewhiletheditransitiveconstructioniscomplexandschematicinthe sense that it can be filled with all verbs that satisfy the constraints of theconstruction(seeGoldberg1995,Croft2003,Barðdal2007,Barðdaletal.2011forananalysisoftheditransitiveconstructionindifferentlanguages).

Since Construction Grammar takes constructions to be the basic units oflanguageandsinceconstructionscanalsobecomplexsyntacticstructures,ConstructionGrammar is the ideal framework for reconstructing syntax (cf. Barðdal& Eythórsson2012a–b, 2016, Barðdal 2013, 2015, Barðdal et al. 2013); Construction Grammar isparticularlyusefulas thecomparanda in the Indo-European languagesunder scrutinyarenotsimplelexicalitemsbutrathercompositionalandschematicexpressions.4.1 Formandmeaning:AnIndo-Europeanconstruction Thepresenceof theGER+(NOM+)DATconstruction in fivedifferentbranchesof Indo-Europeanandacrosswidelydivergentgeographicareas,spanningtheEast-Westdivide,clearly suggests that this combination continues a construction of common Indo-Europeanorigin(Hettrich1990:64ff.,Danesi2013).Thecombinationofadativeandagerundive also shows many similarities at different linguistic levels across thelanguagesexamined,whichare toogreat tobeattributed toa simple coincidencebutmustinsteadbeassumedtoderivefromacommonorigin.Wenowdiscusseachofthesesimilaritiesinturn:

29

At the morphological level, in all the languages taken into account except forTocharian, thegerundive isderivedviaasuffixaddedtoaverbalrootorstem.At thesyntactic level, inall the languagesunder investigation, thisgerundiveselects for twoarguments,anominativeandadative,bothofwhichareoptional,eventhoughourfocusabovehasbeenonthevariantwiththedative.Atthesemanticlevel,inallthelanguagesexamined,thegerundiveconstructionexpressesthesamemodalmeaningofobligationornecessity.Atanetymologicallevel,SanskritandAvestansharethesamesuffixes:theSanskrit -ya-, -tva- and Avestan -iia-, -θβa- are clearly etymologically related. Anetymological connection between Sanskrit -tva-, -tavya- and Greek -τεο- <*-τεϝo-(Brugmann1886–1893:424)isalsopossible. Beforewecontinuetoourreconstruction,letusaddresstheissueofwhetherthemodalmeaningcanbetakentobederivedfromthesuffixitselforwhetheritmustbeassumedtobeaninherentpartoftheconstructionasawhole.SincethesuffixdoesnotoccuroutsidetheGERconstruction,itisclearthatthemodalmeaningisnotpredictablefrom any aspect of the language outside this construction, and it would only bepredictable fromthegerundivesuffixon the linguist’sanalysis that thesuffixhas thismeaning.Suchananalysiswouldnotbeindependentlymotivated,butwouldbebasedstrictly on themeaningof the gerundive construction itself.As such, this becomes ananalytical distinction. The suffix, in our view, is an inherent part of the deonticconstructionasisevidentfromourreconstructionbelow(Figure1).

We suggest the correspondence set given in Table 7, which shows four casemarkers for the protagonist, the dative, the accusative, the genitive, and theinstrumental, on thebasisof thedatapresented inSection2above.Theaccusative isonly found in Ancient Greek, the instrumental is confined to Sanskrit, the genitive isfound in the two Indo-European branches where the dative and the genitive havemerged,namelyTocharianandIndo-Aryan,while thedative is found in fivebranches,i.e. all branches except for Tocharian. The GER+(NOM+)DAT construction in VedicSanskrit showsvariationbetween thedative, thegenitiveand the instrumental,whileonlythegenitiveandtheinstrumentalareattestedinClassicalSanskrit,suggestinganongoingchangewith theoriginaldativebeing lostand its functions takenoverby thegenitive.Table7:CorrespondencesetfortheGER+(NOM+)DATconstruction. Alt1 Alt2 Alt3 Alt4

LatinAncientGreekVedicSanskritSanskritAvestanTocharianLithuanian

GER-‘be’-(Nom)-DatGER-‘be’-(Nom)-DatGER-‘be’-(Nom)-DatGER-‘be’-(Nom)-DatGER-‘be’-(Nom)-Dat

GER-‘be’-(Nom)-Acc

GER-‘be’-(Nom)-GenGER-‘be’-(Nom)-GenGER-‘be’-(Nom)-Gen

GER-‘be’-(Nom)-InsGER-‘be’-(Nom)-Ins

SincethepatternswithaccusativeandinstrumentalareeachconfinedtoonlyoneIndo-European subbranch, Greek and Indo-Aryan, respectively, and since the genitive is asecondarydevelopmentinbothTocharianandIndo-Aryan,continuinganearlierIndo-Europeandative,noneofthesequalifyascandidatesforreconstruction.Incontrast,theconstructionwith the dative can be reconstructed on the basis of occurrences in five

30

Indo-European subbranches, six with Tocharian included, given that the TochariangenitivehastakenoverthefunctionsoftheIEdative.

Thus, our reconstruction of theGER+(NOM+)DAT construction for Proto-Indo-Europeanisbasedonthefollowingcriteria:

● thefunctionoftheverbaladjectiveasgerundiveinallsixbranches● theunexpectedpredicativeuseofthegerundiveacrossallsixbranches● thegerundive’sunexpectedsubcategorizationforadativeargument(orfora

casethatcontinuestheIndo-Europeandative)acrossallsixbranches● thealreadyestablishedinternaletymologiesofdativeendingsanddative

pronounsintheIndo-Europeanlanguages● thealmostidenticalsemanticsoftheGER+(NOM+)DATconstructionas

conveyingobligationornecessityacrossallsixbranches● thepartialetymologicalrelationsfoundacrossSanskrit,Avestan,and

(potentially)GreekThelastbulletedpointinvolvingthepartialetymologicalrelationsonlyaddssupporttothe reconstruction; strictly speaking, they are not necessary for proposing an Indo-Europeanreconstruction.Thereasonisthatmorphologicalmaterial,exactlylikelexicalmaterial,canbereplacedovertime,whilethemoreschematiccategoryitselfstaysthesame. Indeed, the lack of cognate derivational suffixes may suggest that thisconstructionshouldevenbereconstructedforamuchdeepertimespanthanonlybacktolateProto-Indo-European. Inordertoexpandonthedetailsofourreconstruction,weemploytheformalismof Sign-Based Construction Grammar (Sag 2012, Michaelis 2009, 2013), which isparticularlyadequateinmodelingthegrammarofsynchroniclanguagesasitfleshesoutall the relevant details of a given construction, including information about form andmeaning.This formalism involves three levels forargumentstructureconstructions,aFORMfield,aSYNfield,andaSEMfield.TheFORMfielddescribesthemorphosyntacticproperties of the construction, the SYN field specifies case marking and argumentstructure,whiletheSEMfieldexpressestherelevantsemanticproperties.

* Argumentstructurecxt

FORM <gerundive‘be’>

SYN <NP-DATi,NP-NOMj>

SEM

FRAMES

Obligation_frameRESPONSIBLEPARTYiDUTYj

Figure1.ReconstructionoftheGER+(NOM+)DATconstructionforProto-Indo-

European

31

Our reconstruction of the GER+(NOM+)DAT construction for Proto-Indo-European isgiveninFigure1.Sincethisreconstructioninvolvesaschematicconstructionandnotalexical construction, there is no phonological material in the FORM field, but only aspecificationofthetypeofmorphologicalcategorythepredicateconsistsof,inthiscasethegerundivetogetherwiththeverb ‘be’(whichcanbeomittedincertaincontexts inthelaterstagesofsomeofthedaughterlanguages).Whilethegerundivesuffixisnotthesame across all six languages, the gerundive is formally one and the same in eachlanguage, in that it is a verbal adjective that can occur in a predicative position andconveysamodalmeaning.

TheSYNfieldspecifiesthetwoarguments,onebeinginthedativeandtheotherin the nominative. The SEM field renders the semantics of the construction throughsemanticframes,inthiscasetheobligation_frame(seetherelevantentryinFrameNet),wheretheprotagonistislabeledthe“ResponsibleParty”andthenominativeislabeled“Duty”.2Thetwoparticipantrolesareeachindexedwithanior j,which, inturn, linkstheparticipantroleswiththedativeandthenominativeargumentfromtheSYNfield.Thisishowcasemarkingandparticipantrolesaremappedtoeachotherinthistypeofrepresentationalformalism.

A formalization of the reconstruction in Figure 1 not only licenses allinstantiations of the construction in the daughter languages, thus emphasizing thecognacy of the construction across the daughters, what is more, it also models theaspects of the grammar of Proto-Indo-European that are relevant for theGER+(NOM+)DAT construction. Our goal is to present a full reconstruction of theconstruction,inwhichallrelevantdetailsareincluded.Onlythroughsuchanelaborateformalismisacompletereconstructionachieved.

The reconstruction in Figure 1 above fleshes out the details of theGER+(NOM+)DAT construction for Proto-Indo-European, which is only oneconstructional variant of the more schematic GER+(NOM)+(DAT) construction. Themoreschematicconstructionhasthreevariants, theonewiththedative, theonewiththe nominative and the one with both the dative and the nominative present. Thedeonticreadingisfoundwithallthreeconstructionalvariants,astheyarealltiedtothepresenceof thegerundive incombinationwithexpressedorunexpressedparticipantsin the obligation frame. Whether the protagonist, the object needed or both areexpressedboilsdowntothechoicesofthespeakerandthepragmaticcontext.4.2 Modality,transitivity,andnon-canonicalcasemarking

We take the GER+(NOM+)DAT construction to be part and parcel of a larger set ofoblique subject constructions found throughout the Indo-European languages (cf. theoverview in Barðdal et al. 2012). What characterizes this construction is a generalreductionintransitivity(cf.Barðdal&Eythórsson2009,Danesi2014).Inthesamevein,modal meaning is usually associated with a reduction of transitivity, which in turninvolves a number of componentswhich all relate to the effectiveness withwhich aneventtakesplace.

In clauses with gerundives, there is an event which is expected to take place(evennecessitatedbythecircumstancessurroundingtheevent),buthasnothappenedyet.AsHopper&Thompson(1980:252)observe,aneventwhichhasnottakenplaceis

2 https://framenet2.icsi.berkeley.edu/fnReports/data/frameIndex.xml?frame=Obligation_scenario

32

“less effective” thanonewhich ispresentedas a real event.Effectiveness is implicitlydefined(pp.251–253)asthetransferofanactionfromanagenttoapatient;themoreaffectedthepatientisbytheaction,themoreeffectiveandalsotransitivetheclauseis—and vice versa. Hence, a constructionwith amodalmeaning is less effective and lesstransitive,astheeventhasnottakenplace.

Turningtothe“dativeofagent”,this“agent”isnorealagentatall,asshownforeachofthelanguagesdiscussedinSection2above,butratheraparticipantoftheeventwhoisconceptualizedashavingnocontrolorvolition,asthisparticipantisundertheobligation to perform a specific duty. In this respect, the protagonist is similar to apatient in being affected by his/her obligation, which is externally imposed. It isgenerally well known in the literature that non-canonical casemarking is connectedwitha lowdegreeof transitivity, cf.Shibatani (1985),Tsunoda(1985),Onishi (2001),Barðdal(2004),Barðdal&Eythórsson(2009),Narrog(2010),andDanesi(2014).

More particularly, as pointed out by Langacker (1991: 409–413), andemphasized by Smith (2001) and Barðdal (2004), since accusatives and datives aretypically used to mark objects of transitive verbs, these case markers denoteaffectedness to amuchhigher degree thannominatives.Hence, the use of an obliquecasetomarktheprotagonistoftheGER+(NOM+)DATconstructionismotivatedbythemodalmeaningof theconstruction, the lackofagentivepropertiesof theprotagonist,andtheconstruction’slowdegreeoftransitivity.

From a typological perspective, moreover, non-canonically case-markedsubjects, especially in dative subject constructions, may be associated with a modalmeaning.Insuchconstructions,thesubject-likeargument,theprotagonist,ismarkedinthe dative case, rather than in the nominative (Onishi 2001, Narrog 2010, Danesi,Johnson&Barðdal2016).However,fromatypologicalperspective,itismorecommonto assume that dative subject constructions are experiencer constructions and notmodalconstructions.Another incongruitybetweentheGER+(NOM+)DATconstructionin the Indo-European languages and this “typologically-attested dative-subjectconstruction” is the fact that an inherently non-finite structure, in this case thegerundive, isused finitely across fivedifferent subbranchesof Indo-European.This isnot a general property of constructions exhibiting dative subjects cross-linguistically.Hence, there is no doubt that this is an idiosyncrasy, specific to the Indo-Europeanlanguages, and must as such be inherited from an earlier proto-stage. For furtherarguments illustrating the incompatibility between the “typological dative” and thedativeintheIndo-Europeanlanguages,seeBarðdaletal.(2012).

ThedativesubjectconstructioninseveralIndo-Europeanlanguageshasrecentlybeenthesubjectofalarge-scalecomparison,carriedoutbyBarðdal(2004,2006,2008),and Barðdal et al. (2012, 2013, 2016). One of their major findings is that that thepredicates instantiating this construction are not confined to experiencer subjects, asthe “typological dative” is assumed to be, but covers instead five different semanticfields.Thesearethesemanticfieldsofexperienceandhappenstanceevents,inadditionto modality, evidentiality, and possession. The two major fields of experience andhappenstanceeventscaneachbefurtherdividedintosmallersubfields,asschematizedinFigure2.Notethatthebenefactiveusesofthedativearetypicallyfoundwithverbsexpressinggainandperhapssuccess,inadditiontoditransitiveswhichfalloutsidethescopeofdativesubjectconstructions.

33

Figure 2. Reconstruction of the semantics of the Dative Subject Construction for acommonproto-stage(Barðdaletal.2012:529).

Furthermore,thereisnodoubtthatthedativeoftheGER+(NOM+)DATconstructionissemanticallysimilartothedativeexpressingthesubjectofmodalpredicatesinsimilarobliquesubjectconstructionsacrossdifferentIndo-Europeanlanguages.Someoftheseare given in (44) below from three of the languages discussed above, namely Latin,AncientGreek,andLithuanian.

(44) a. Latin

Huius nobis exempla permulta opusthis.GEN us.DATexamples.NOM very.many.NOM.PL needsuntbe.PRES.3PL‘Weneedverymanyexamplesofthis’(Cic.Inv.2.19.57)

b. Greek

θεοῖσι προσβαλεῖν χθονὶ ἄλλην gods.DAT put.to.INF hearth.ACC other.ACC.SGδεήσει γαῖανbe.needful.FUT.3SG world.ACC‘Thegodswillhavetoaddanotherearthtoourworld’(Eur.Hipp.941)

c. Lithuanian

Man reĩkia kambarioI.DAT need.PRES.3SG room.GEN‘Ineedaroom’(Schmalstieg1987:220)

These examples show strong structural parallels with the GER+(NOM+)DATconstruction:Eachoftheseexamplesispatient-orientedinthesensethatthepatientismarkedinthenominativecaseandthesubject-likeprotagonistinthedativecase.

Weargue that the existenceof theGER+(NOM+)DAT construction in the Indo-European languages is licensed by the existence of a more general dative subjectconstruction;inthissense,theGER+(NOM+)DATconstructionisasubconstructionofa

34

larger family of constructions with non-canonical case-marking. With this family ofconstructions it shares the formandoneof itsmeanings,namely themodalmeaning.The GER+(NOM+)DAT construction, moreover, is mostly schematic; part of the verbphrase is lexically filled with the verb ‘be’, but the gerundive suffix can in principlecombinewithanylexicalverbfittingtheconstruction.5. SummaryandConclusionsIn the present article we have analyzed the syntactic and semantic properties ofgerundivesthatselectfordativearguments,comparingtheirbehavioracrosstheIndo-European languages where they are documented, namely Latin, Greek, Sanskrit,Avestan, Tocharian, and Lithuanian—languages that represent five separatesubbranches of Indo-European and span the East-West divide. On the basis of thiscomparisonwehavereconstructed forProto-Indo-Europeanaschematicconstructionwith a gerundive predicate (consisting of an optional form of the verb ‘be’ and thederived gerundive itself) and a dative argument. This construction conveys a modalmeaningofobligationand/ornecessityinallsixlanguages,andthisisthemeaningthatwesuggestinourproposedreconstruction.

Furthermore, the protagonist is expressed in the dative (with the exception ofTocharian and Indo-Aryan where the genitive and the dative havemerged), and theconstructionispatient-orientedinthesensethatthereferentwiththesemanticroleofpatientisexpressedinthenominative,theusual“unmarked”case.Sincethispatientisassigned nominative case, the GER+(NOM+)DAT construction resembles passivestructures and has frequently been analyzed as such. However, the GER+DATconstructiondoesnothavethesameformorthesamemeaningasthepassiveinanyofthe languages examined. Crucially, in all of the languages, the agent of “true” passivestructuresisnotexpressedwithadativebutwithothercasesorprepositionalphrases.Inaddition,theGER+(NOM+)DATconstructionisnotfunctionallyequivalenttopassivessinceithasaspecificmeaningofobligation.

Since the dative is frequently used in Indo-European languages to express apossessor,wehavealsoexamined thehypothesis,proposed in the literature, that theGER+(NOM+)DATconstructionshouldberegardedasapossessiveconstruction,andwehaveshownthatthisinterpretationistoosimplistic.Eventhoughthetwoconstructionsexhibitsomestructuralparallels,oneisclearlynotderivedfromtheother;insteadthetwoarebetter interpretedas twosubconstructionsofa larger familyofconstructionswherethesubject-likeargumentismarkedinthedativecase.

Wehave also argued against a recent analysis in the literaturewhich assumesthatthe“dativeofagent”isnotreconstructableforProto-Indo-European,butisinsteadan independent development in the daughter languages, based on the beneficiarymeaningofthedative,allegedlyinheritedfromProto-Indo-European.Wehaveprovidedseveralargumentsagainstsuchananalysis,includingpointingoutthatanybeneficiarymeaning of the dative does not exist irrespective of the constructions that the dativeoccurs in. This means that the beneficiary meaning of the dative derives fromditransitivesandasmall subsetofdativesubject constructionexpressinggain. In thissense, there cannotbe any abstract independentbeneficiarydative in the early Indo-European languages, and, consequently, there can be no development from any suchallegedabstractbeneficiarydativetothe“dativeofagent”construction.

35

Instead of a) the traditionalist analysis, b) the possessive analysis, and c) thebenefactiveanalysis,wehavearguedabovethatthenon-canonicalcase-markingofthesubject of the GER+(NOM+)DAT construction is directly associated with the lowertransitivity of modal constructions, as is the case with other oblique subjectconstructions in the early Indo-European languages. Thus, the GER+(NOM+)DATconstructionisbestanalyzedasasubconstructionofthemoregeneralobliquesubjectconstruction in Indo-European which has been reconstructed exhibiting five majorsemantic fields, namely those of experience, happenstance, possession, modality andevidentiality.

To conclude, theworkpresented in this article offers an important exercise insyntactic reconstruction. The similarities between the GER+(NOM+)DAT constructionacrossthesixIndo-Europeanlanguagesdiscussedabovearetoogreattobeamatterofchance, but rather represent a continuation of an Indo-European construction. Thisconstructionisitself,giventhesimilaritiestootherconstructionswithnon-nominativesubjects,asubconstructionoftheobliquesubjectconstructionthatisfoundevenmorewidely across Indo-European and for which reconstructions have already beenproposedintherecentliterature. References

Adams,DouglasQ.2011.ALateAlternativetotheMihiEstConstructioninTocharianB.TocharianandIndo-EuropeanStudies12.53–56.

Allen, Joseph Henry & James Bradstreet Greenough. 1903. New Latin Grammar.Boston/London:Ginn&Company.

Ambrazas,Vytautas.1997.LithuanianGrammar.Vilnius:BaltosLankos.Andersen,PaulKent.1986.ThegenitiveagentinRigvedicpassiveconstructions.CollectanealinguisticainhonoremAdamiHeinz,edbyF.Slawski,9–13.Wrocław:ZakładNarodowyim.Ossolińskich.

Barðdal,Jóhanna.2001.ThePerplexityofDat-NomVerbsinIcelandic.NordicJournalofLinguistics24:47–70.

Barðdal,Jóhanna.2003.Reviewof“CognitiveLinguisticsandthePolishDative”byEwaDabrowska.CognitiveLinguistics14(4):379–386.

Barðdal,Jóhanna.2004.TheSemanticsoftheImpersonalConstructioninIcelandic,GermanandFaroese.Beyondthematicroles.FocusonGermanicTypology,ed.byWernerAbraham,105–137.Berlin:AkademieVerlag.

Barðdal,Jóhanna.2007.TheSemanticandLexicalRangeoftheDitransitiveConstructionintheHistoryof(North)Germanic.FunctionsofLanguage14(1):9–30.

Barðdal,Jóhanna.2008.Productivity:EvidencefromCaseandArgumentStructureinIcelandic.Amsterdam:JohnBenjamins.

Barðdal,Jóhanna.2011.TheRiseofDativeSubstitutionintheHistoryofIcelandic:ADiachronicConstructionGrammarApproach.Lingua121(1):60–79.

Barðdal, Jóhanna. 2013. Construction-Based Historical–Comparative Reconstruction.TheOxfordHandbookofConstructionGrammar,ed.byGreameTrousdale&ThomasHoffmann,438–457.Oxford:OxfordUniversityPress.

Barðdal, Jóhanna.2015.SyntaxandSyntacticReconstruction.TheRoutledgeHandbookofHistorical Linguistics, ed. by ClaireBowern&BethwynEvans, 343–373. London:Routledge.

36

Barðdal, Jóhanna, CarleeArnett, StephenMark Carey, Thórhallur Eythórsson, GardB.Jenset, Guus Kroonen & Adam Oberlin. 2016. Dative Subjects in Germanic: AComputational Analysis of Lexical Semantic Verb Classes Across Time and Space.STUF:LanguageTypologyandUniversals69(1):49–84.

Barðdal,Jóhanna,ValgerðurBjarnadóttir,SerenaDanesi,TonyaKimDewey,ThórhallurEythórsson,ChiaraFedriani&ThomasSmitherman.2013.TheStoryof'Woe'.JournalofIndo-EuropeanStudies41(3–4):321–377.

Barðdal,Jóhanna&SerenaDanesi.2014.ConstructionGrammarandGreek.EncyclopediaofAncientGreekLanguageandLinguistics(EAGLL),ed.byG.K.Giakanis,375–379.Brill:Leiden.

Barðdal,Jóhanna&ThórhallurEythórsson.2009.TheOriginoftheObliqueSubjectConstruction:AnIndo-EuropeanComparison.GrammaticalChangeinIndo-EuropeanLanguages,ed.byVitBubenik,JohnHewson&SarahRose,179–193.Amsterdam:JohnBenjamins.

Barðdal,Jóhanna&ThórhallurEythórsson.2012a.ReconstructingSyntax:ConstructionGrammarandtheComparativeMethod.Sign-BasedConstructionGrammar,ed.byHansC.Boas&IvanA.Sag,257–308.Stanford:CSLIPublications.

Barðdal,Jóhanna&ThórhallurEythórsson.2012b."HungeringandLustingforWomenand Fleshly Delicacies": Reconstructing Grammatical Relations for Proto-Germanic.TransactionsofthePhilologicalSociety110(3):363–393.

Barðdal,Jóhanna&ThórhallurEythórsson.2017.HowtoIdentifyCognatesinSyntax:TakingWatkins’LegacyOneStepFurther.ToappearinReconstructingSyntax,ed.byEugenioR.Lujan,JóhannaBarðdal&SpikeGildea.Leiden:Brill.

Barðdal,Jóhanna,KristianE.Kristoffersen&AndreasSveen.2011.WestScandinavianDitransitivesasaFamilyofConstructions:WithaSpecialAttentiontotheNorwegianV-REFL-NPConstruction.Linguistics49(1):53–104.

Barðdal,Jóhanna&ValériaMolnár.2003.ThePassiveinIcelandic–ComparedtoMainlandScandinavian.StructuresofFocusandGrammaticalRelations,ed.byJorunnHetland&ValériaMolnár,231–260.Tübingen:MaxNiemeyerVerlag.

Barðdal, Jóhanna & Thomas Smitherman. 2013. The Quest for Cognates: AReconstructionofObliqueSubjectConstructions inProto-Indo-European.LanguageDynamicsandChange3(1):28–67.

Barðdal,Jóhanna,ThomasSmitherman,ValgerðurBjarnadóttir,SerenaDanesi,GardB.Jenset&BarbaraMcGillivray.2012.ReconstructingConstructionalSemantics:TheDativeSubjectConstructioninOldNorse-Icelandic,Latin,AncientGreek,OldRussianandOldLithuanian.StudiesinLanguage36(3):511–547.

Bauer,BrigitteL.M.2000.ArchaicSyntaxinIndo-European:TheSpreadofTransitivityinLatinandFrench.Berlin:MoutondeGruyter.

Bennett,CharlesE.1914.SyntaxofEarlyLatin.II.TheCases.Boston:AllynandBacon.Berg-Olsen,Sturla.2009.LackinginLatvian:CaseVariationfromaCognitiveandConstructionalPerspective.TheRoleofSemantic,PragmaticandDiscourseFactorsintheDevelopmentofCase,ed.byJóhannaBarðdal&ShobhanaL.Chelliah,181–202.Amsterdam:JohnBenjamins.

Bowern,Claire.2008.SyntacticChangeandSyntacticReconstructioninGenerativeGrammar.PrinciplesofSyntacticReconstruction,ed.byGiselaFerraresi&MariaGoldbach,187–216.Amsterdam:JohnBenjamins.

Brugmann,Karl.1886–1893.GrundrissdervergleichendenGrammatikderindogermanischenSprachen.KurzgefassteDarstellungderGeschichtedesAltindischen,Altiranischen(AvestischenundAltpersischen),Lateinischen,Umbrisch-Samnitischen,

37

Altirischen,Gotischen,Althochdeutschen,LitauischenundAltkirchenslavischen.Vol.2,Part1:Wortbildungslehre(Stammbildungs-undFexionslehre).Strassburg,Trübner.

Brugmann,Karl&Delbrück,Bertold.1900.GrundrißdervergleichendenGrammatikderindogermanischenSprachen,Band2,Teil2.Straßburg:Trübner.

Bubenik,Vit.1997.TheVerbalSystemofTocharian.TenseandAspectinIndo-EuropeanLanguages:Theory,typology,diachrony,edbyJohnHewson&VitBubenik,125–142.Amsterdam:JohnBenjamins.

Bubenik,Vit.2006.Tocharian:FromCasetoAdposition.ThedevelopmentofconfigurationalsyntaxinIndo-Europeanlanguages,ed.byJohnHewson&VitBubenik,317–333.Amsterdam:JohnBenjamins.

Conti,Luz.1999.LaexpresióndelaCausaenHomeroconreferenteshumanos.Emerita67.2,295–313.

Conti,Luz.2009.WeitereszumGenitivalsSemisubjektimAltgriechischen:AnalysedesKasusbeiimpersonalenKonstruktionen.HistorischeSprachforschung122.182–207.

Cotticelli Kurras, Paula & Alfredo Rizza. 2013. Reconstructing Proto-Indo-EuropeanCategories:TheReflexiveandtheMiddleinHittiteandtheProto-Language.JournalofHistoricalLinguistics3(1):7–27.

Croft,William.1998.Eventstructureinargumentlinking.Theprojectionofarguments:lexicalandcompositionalfactors,ed.MiriamButt&WilhelmGeuder,1–43.Stanford:CenterfortheStudyofLanguageandInformation.

Croft,WilliamA.2001.RadicalConstructionGrammar:SyntacticTheoryinTypologicalPerspective.Oxford:OxfordUniversityPress.

Croft,William.2012.Verbs:AspectandCausalStructure.Oxford:OxfordUniversityPress.

Croft,William.2003.LexicalRulesvs.Constructions:AFalseDichotomy.MotivationinLanguage:StudiesinHonourofGünterRadden,ed.byH.B.Cuyckens,Th.Berg,R.Dirven,andKl.-U.Panther,49–68.Amsterdam:JohnBenjamins.

Danesi,Serena,CynthiaA.Johnson&JóhannaBarðdal.2017.WhereDoestheModalityof Ancient Greek Modal Verbs Come From? The Relation between Modality andObliqueCaseMarking.UnderrevisionforJournalofGreekLinguistics.

Daues,Alexandra.2006.OntheFunctionoftheMediopassivePerfectintheHomericEpics.WordClassesandRelatedTopicsintheAncientGreek,ProceedingsoftheConferenceon'GreekSyntaxandWordClasses'heldinMadridon18–21,June2003,ed.byE.Crespo,J.DelaVilla&A.R.Revuelta,257–272.Louvain-La-Neuve:Peeters.

Dahl,Eystein&ChiaraFedriani.2012.TheArgumentStructureofExperience:ExperientialConstructionsinEarlyVedic,HomericGreekandEarlyLatin.TransactionsofthePhilologicalSociety110(3).342–362.

Danesi,Serena.2013.BetweentheHistoricalLanguagesandtheReconstructedLanguage:The‘VerbalAdjective+Dative’Construction.Atalkpresentedatthe21stInternationalConferenceofHistoricalLinguistics,Oslo,5–9August.

Danesi,Serena.2014.AccusativeSubjectsinAvestan:‘Errors’orNoncanonicallyMarkedArguments.Indo-IranianJournal57(3).223–260.

Danesi,Serena,CynthiaA.Johnson&JóhannaBarðdal.2016.WhereDoestheModalityofAncientGreekModalVerbsComeFrom?TheRelationbetweenModalityandObliqueCaseMarking.SubmittedtoJournalofGreekLinguistics.

Dash,NiladriSekhar.2008.ContextandContextualWordMeaning.SKASEJournalofTheoreticalLinguistics5(2):21–31.

Delbrück,Berthold.1888.AltindischeSyntax.Hallea.S.:VerlagderBuchhandlungdesWaisenhauses.

38

Eythórsson,Thórhallur&JóhannaBarðdal.2011.DieKonstruktionsgrammatikunddiekomparativeMethode.IndogermanistikundLinguistikimDialog:AktenderXIII.FachtagungderIndogermanischenGesellschaftvom21.bis27.September2008inSalzburg,ed.byThomasKrisch&ThomasLindner,148–156.Wiesbaden:ReichertVerlag.

Eythórsson,Thórhallur&JóhannaBarðdal.2016.SyntacticReconstructioninIndo-European:TheStateoftheArt.FranzBoppandhisComparativeGrammarModel(1816–2016),ed.byJ.Gorrochategui,C.GarcíaCastillero&J.M.Vallejo.AspecialmonographicvolumeinVeleia33:83–102.

Fedriani,Chiara.2014.ExperientialConstructionsinLatin.Brill:Leiden.Fillmore,Charles,PaulKay&CatherineO'Connor.1988.RegularityandIdiomaticityinGrammaticalConstructions:TheCaseofletalone.Language64:501–538.

Gaedicke,Carl.1880.DerAccusativimVeda.Breslau:W.Koebner.Garrett,Andrew.1990.TheSyntaxofAnatolianPronominalClitics.HarvardUniversityPhD.Dissertation.

Geldner,KarlFriedrich.1952–1957.DerRig-Veda:ausdemSanskritinsDeutscheÜbersetztundmiteinemlaufendenKommentarversehen.HarvardOrientalSeries33–36,Bd.1–3.Cambridge,Massachusetts:HarvardUniversityPress.Availableonlineathttp://www.sanskritweb.net/rigveda/rigveda.pdf/.

Gildea,Spike.1998.OnReconstructingGrammar:ComparativeCaribanMorphosyntax.Oxford:OxfordUniversityPress.

Gildersleeve,BasilL.andLodge,Gonzalez.1894.Gildersleeve'sLatingrammar.3ded.NewYork:BostonUniversityPublishingCompany.

Goldberg,Adele.1995.Constructions:AConstructionGrammarApproachtoArgumentStructure.Chicago:UniversityofChicagoPress.

Goldberg,Adele.2006.ConstructionsatWork:TheNatureofGeneralizationinLanguage.Oxford:OxfordUniversityPress.

Goodwin,WilliamG.1900.GreekGrammar.Boston:GinnandCompany.Green,Alexander.1913.TheDativeofAgency:AChapterofIndo-EuropeanCase-Syntax.NewYork:ColumbiaUniversityPress.

George,CoulterH.2005.ExpressionsofAgencyinAncientGreek.Cambridge:CambridgeUniversityPress.

Hale, Mark. 1987a. Studies in the comparative syntax of the oldest Indo-Iranianlanguages,Ph.D.Dissertation,HarvardUniversity.Cambridge,MA.

Hale, Mark. 1987b. Notes on Wackernagel’s Law in the language of the Rigveda. InStudiesinmemoryofWarrenCowgill(1929–1985):PapersfromtheFourthEastCoastIndo-EuropeanConference,ed.CalvertWatkins,38–50.Berlin–NewYork:WalterdeGruyter.

Harris,AliceC.&LyleCampbell.1995.HistoricalSyntaxinCross-LinguisticPerspective.Cambridge:CambridgeUniversityPress.

Harrison,ShellyP.2003.OntheLimitsoftheComparativeMethod.TheHandbookofHistoricalLinguistics,ed.byBrianD.Joseph&RichardD.Janda,343–368.Oxford:Blackwell.

Havers,Wilhelm.1911.UntersuchungenzurKasussyntaxderindoeuropäischenSprachen.Strassburg:Trübner.

Hettrich,Heinrich.1990.DerAgensinpassivischenSätzenaltindogermanischerSprachen.Göttingen:VandenhoeckundRuprecht.

39

Hettrich,Heinrich.2014.ZumAgensinpassivischenSätzendesRigveda.Na-wa/i-VIR.ZI/AMAGNUS.SCRIBA–FestschriftfürHelmutNowickizum70.Geburtstag,ed.byCyrilBrosch&AnnickPayne,107–112.Wiesbaden:Harrassowitz.

Hjelmslev,Louis.1935.Lacatégoriedescas:Etudedegrammairegénérale(ActaJutlandica:AarsskriftforAarhusUniversitet9.3).Copenhagen:Munksgaard

Hock,HansHeinrich.1985–1986.Voice,mood,andthegerundive(kṛtya)inSanskrit.IndologicaTaurinensia13:81–102.

Hock,HansHenrich.2013.Reconstruction,Typology,Validation.JournalofHistoricalLinguistics3(2):49–76.

Hopper,PaulJ.&SandraA.Thompson.1980.TransitivityinGrammarandDiscourse.Language56:251–299.

Insler,Stanley.1975.TheGāthāsofZarathustra.Leiden:Brill.Jackendoff,Ray.1997.TheArchitectureoftheLanguageFaculty.Cambridge,MA:MITPress.

Jakobson,Roman.1936.BeitragzurallgemeinenKasuslehre:GesamtbedeutungenderrussischenKasus.TravauxduCercleLinguistiquedePrague6:240–299.

Jamison,StephanieW.1979.TheCaseoftheAgentinIndo-European.DieSprache15.129–143.

Jamison,StephanieW.2000.OnTranslatingtheRigVeda:Threequestions.Proceedingsofthe11thAnnualUCLAIndo-EuropeanConference.LosAngeles,June4–51999,ed.byKarleneJones-Bley,MartinE.Huld,&AngelaDellaVolpe,1–20.Washington:InstitutefortheStudyofMan.

Jasanoff, JayH.2006.TheOriginof theLatinGerundandGerundive:ANewProposal.HarvardUkrainianStudies28(1–4):195–208).

Jeffers,RobertJ.1976.SyntacticChangeandSyntacticReconstruction.CurrentProgressinHistoricalLinguistics:ProceedingsoftheSecondInternationalConferenceonHistoricalLinguistics,ed.byWilliamM.Jr.Christie,WilliamM.Jr.,1–15.Amsterdam.

Joseph,BrianD.2012.AVariationistSolutiontoApparentCopyingAcrossRelatedLanguages.Copiesvs.CognatesinBoundMorphology,ed.byLarsJohanson&MartineRobbeets,151–164.Leiden:Brill.

Joseph,BrianD.2013.DemystifyingDrift:AVariationistAccount.SharedGrammaticalization:WithspecialfocusontheTranseurasiannnlanguages,ed.byMartineRobbeets&HerbertCuyckens,43–66.Amsterdam:JohnBenjamins.

Kanga,KavasjiEdulji.1891.APracticalGrammaroftheAvestaLanguage.Bombay:EducationSoc.Press.

Kay,PaulandFillmore,CharlesJ.1999.GrammaticalConstructionsandLinguisticGeneralizations:theWhat’sXDoingY?Construction.Language75(1):1–33.

Keydana,Götz.2013.InfinitiveimRigveda:Formen,Funktion,Diachronie.Leiden:Brill.Kikusawa, Ritsuko. 2002. Proto Central Pacific Ergativity: Its Reconstruction andDevelopment in the Fijian, Rotuman and Polynesian languages. Canberra: PacificLinguistics.

Kilgarriff,Adam.1997. IDon'tBelieve inWordSenses.Computersand theHumanities31(2):91–113.

Klein,Jared.2010.Reviewof“PrinciplesofReconstruction”,ed.byGiselaFerraresiandMariaGoldbach2008).Language86:720–726.

Krause,Wolfgang&WernerThomas.1960.TocharischesElementarbuch.Heidelberg:C.Winter.

Kühner,Raphael&Gerth,Bernhard.1955.AusführlicheGrammatikdergriechischenSprache:Satzlehre.Hannover:Hahn.

40

Kühner,RaphaelandStegmann,Carl.1955.AusführlicheGrammatikderlateinischenSprache:Satzlehre.3.Aufl.Leverkusen:Gottschalk.

Kulikov,Leonid.2011.Passivetoanticausativethroughimpersonalization:ThecaseofVedicandIndo-European.Impersonalconstructions.Across-linguisticperspective,edbyAnnaSiewierska&AndrejMalchukov,229–254.Amsterdam:JohnBenjamins.

Kulikov,Leonid.2012.TheVedic-ya-presents:PassivesandintransitivityinOldIndo-Aryan.Amsterdam:Rodopi.

Kulikov,Leonid&NikolaosLavidas.2013.ReconstructingPassiveandVoice inProto-Indo-European.JournalofHistoricalLinguistics3(2):101–124.

Kurschat,Friedrich.1986.Grammatikderlittauischensprache.Halle:VerlagderBuchhandlungdesWaisenhauses.

Lakoff,George&MarkJohnson1980.MetaphorsWeLiveBy.Chicago:UniversityofChicagoPress.

Langacker,RonaldW.1991.FoundationsofCognitiveGrammarII:DescriptiveApplication.Stanford:StanfordUniversityPress.

Leskien,A.1919.LitauischesLesebuchmitGrammatikundWörterbuch.IndogermanischeBibliothek.,ErsteReihe,Lehr-undHandbücher;12.Bd.Heidelberg:C.Winter.

Leumann,Manu,JohanBaptistHofmann,&AntonSzantyr.1972.LateinischeSyntaxundStilistik,Part2,Volume2.Reprint.Munich:C.H.Beck.

Lühr,Rosemarie.2004.ThematischeRollenundKasus:ZuAgensundPatiensbeimPassivimAltindischen.LinguistischeArbeitsberichte81,99–126.

Lühr, Rosemarie. 2008. Competitive Indo-European Syntax. In Principles of SyntacticReconstruction. Eds.GiselaFerraresi&MariaGoldbach,121–159.Amsterdam: JohnBenjamins.

Luraghi,Silvia.1986.Onthedistributionofinstrumentalandagentmarkersforhumanandnon-humanagentsofpassiveverbsinsomeIndo-Europeanlanguages.IndogermanischeForschungen91,48–66.

Luraghi,Silvia.1995.PrototypicalityandagenthoodinIndo-European.HistoricalLinguistics1993,ed.byHenningAndersen,259–268.Amsterdam:JohnBenjamins.

Luraghi,Silvia.2000.SpatialmetaphorsandagenthoodinAncientGreek.125JahreIndogermanistikinGraz,ed.byCh.Zinko&M.Offisch,283–209.Leykam:Graz.

Luraghi,Silvia.2010.ExperiencerpredicatesinHittite.ExAnatolialux:AnatolianandIndoEuropeanStudiesinHonorofH.CraigMelchertontheOccasionofhisSixty-FifthBirthday,ed.byRonaldKim,NorbertOettinger,ElisabethRiecken&MichaelJ.Weiss,249–264.AnnArbor:BeechStavePress.

Luraghi,Silvia.2016.TheDativeofAgentinIndo-EuropeanLanguages.STUF–LanguageTypologyandUniversals69(1):15–47.

Macdonnell, Arthur Anthony. 1916. A Vedic Grammar for Students. Oxford: OxfordUniversityPress.

Matasović,Ranko.2013.Latinpaenitetme,miseretme,pudetmeandactiveclausealignmentinPIE.IndogermanischeForschungen118.93–110.

Menge,Hermann.2000.LehrbuchderlateinischenSyntaxundSemantik:VölligneubearbeitetvonThorstenBurkardundMarkusSchauer.Darmstadt:WissenschaftlicheBuchgesellschaft.

Michaelis,LauraA.2009.Sign-BasedConstructionGrammar.TheOxfordHandbookofLinguisticAnalysis,ed.byB.HeineandH.Narrog,155–176.Oxford:OxfordUniversityPress.

41

Michaelis,LauraA.2013.Sign-BasedConstructionGrammar.TheOxfordHandbookofConstructionGrammar,ed.byT.HoffmanandG.Trousdale,133–152.Oxford:OxfordUniversityPress.

Narrog,Heiko.2010.Voiceandnon-canonicalcasemarkingintheexpressionofevent-orientedmodality.LinguisticTypology14,71–126.

Oberlies,Thomas.2003.AGrammarofEpicSanskrit.Berlin:deGruyter.Onishi,Masayuki.2001.Non-canonicallymarkedsubjectsandobjects:Parametersandproperties.Non-CanonicalMarkingofSubjectsandObjects,ed.byAlexandraY.Aikhenvald,R.M.W.Dixon&MasayukiOnishi,1–51.Amsterdam:JohnBenjamins.

Ottósson,KjartanG.1992.TheMiddleVoiceinIcelandic.Ph.D.dissertation,LundUniversity.

Palmer,FrankR.2001.MoodandModality.2ndEdition.Cambridge:CambridgeUniversityPress.

Reichelt,Hans.1909.AwestischesElementarbuch.Heidelberg:C.Winter.Reichelt,Hans.1911.Avestareader:texts,notes,glossaryandindex.Strassburg:VerlagvonKarlJ.Trübner.

Sag,IvanA.2012.Sign-basedConstructionGrammar:Aninformalsynopsis.Sign-basedConstructionGrammar,ed.byHansC.Boas&IvanSag,69–202.Stanford:CSLIPublications.

Schleicher,August.1856.Handbuchderlitauischensprache.Prag:J.G.Calve.Schmalstieg,WilliamR.1987.ALithuanianHistoricalSyntax.Columbus,Ohio:SlavicaPublishers.

Senn,AlfredErich.1966.HandbuchderlitauischenSprache.1,Grammatik.Heidelberg:C.Winter.

Shibatani,Masayoshi.1985.Passivesandrelatedconstructions:Aprototypeanalysis.Language61.821–848.

Skjærvø,ProdsOktor.2003.An Introduction toYoungAvestan, rechecked4thversion.Unpublished.Availableonlineathttps://www.fas.harvard.edu/~iranian/Avesta/.

Smith,MichaelB.2001.WhyQuirkyCaseReally Isn’tQuirky(OrHowtoTreatDativeSickness in Icelandic).Polysemy in Cognitive Linguistics, ed. by Hubert Cuyckens &BrittaZawada,115–159.Amsterdam:JohnBenjamins.

Smyth,HerbertWeir.1963.Greekgrammar(revisedbyGordonMMessing).Cambridge:HarvardUniversityPress.

SwaabTamara,ColinBrown&PeterHagoort.2003.UnderstandingWordsinSentenceContexts:TheTimeCourseofAmbiguityResolution.BrainandLanguage86(2):326–343.

Tsunoda,Tasaku.1985.RemarksonTransitivity.JournalofLinguistics21.385–396.VanHoecke,Willy. 2011. The LatinDative.TheDative: Volume 1:Descriptive studies.Amsterdam:JohnBenjamins.

Viti, Carlotta. 2014. Reconstructing Syntactic Variation in Proto-Indo-European. Indo-EuropeanLinguistics2:73–111.

Viti,Carlotta.2016.ArealdistributionofargumentmarkingofIndo-Europeanexperiencepredicates.JournalofIndo-EuropeanStudies44:1–84.

Walkden, George. 2014. Syntactic Reconstruction and Proto-Germanic. Oxford: OxfordUniversityPress.

Watkins,Calvert.1976.“TowardProto-Europeansyntax:Problemsandpseudo-problems.”InSteever,S.,Walker,C.&Mufwene,S.(eds.),305–326.

Werner,Thomas.1951.DietocharischenVerbaladjektiveauf1:EinesyntaktischeUntersuchung.[Diss.].Series:Berlin.DeutscheAkademiederWissenschaften.Institut

42

fürOrientforschung.Veröffentlichung,9.Göttingen:DeutscheAkademiederWissenschaften.InstitutfürOrientforschung.

Wessén,Elias.1965.SvenskspråkhistoraIII.Grundlinjertillenhistorisksyntax[SwedishLanguageHistoryIII:TheMainFeaturesofHistoricalSyntax].Stockholm:AlmqvistandWiksell.

Willis,David.2011.ReconstructingLastWeek’sWeather:SyntacticReconstructionandBrythonicFreeRelatives.JournalofLinguistics47(2):407–446.

Winter,Werner.1984.ReconstructionalComparativeLinguisticsandtheReconstructionoftheSyntaxofUndocumentedStagesintheDevelopmentofLanguagesandLanguageFamilies.HistoricalSyntax,ed.byJ.Fisiak,613–625.TheHague:Mouton.