Possessor Raising and Slavic clitics

Possessor Raising and Slavic Clitics1 Anton Zimmerling Moscow State University for the Humanities (MSUH) meinmat@yahoo.com The paper discusses syntactic features of Slavic possessive clitics and Slavic constructions with socalled ―Possessor Raising‖. I prove that only a minority of Slavic languages have true phrase-level (NP-level or DP-level) possessive clitics and arguing against a generalized syntactic account of all Slavic constructions with possessive operators. The weak aspects of the PR hypothesis are that it takes the mapping of syntax and possessive semantics to be iconic and the rules/principles of extracting a NP/DP-level possessive operator out of the NP/DP to be trivial. The analysis has shown that these assumptions are poorly justified. Slavic languages typically apply different case-marking for nonagreeing phrase-level possessive operators and non-agreeing clause-level possessive operators. A group of languages including Modern Russian lacks phrase-level possessive clitics. For this group the PR hypothesis cannot be retained. Most cases where the PR hypothesis has been proposed in previous Slavic studies do not conform to the definition of Raising as a syntactic operation, since the identity of structures with a clause-level or phrase-level possessive operator cannot be established. Such cases must be reanalyzed in terms of Possessive Shift, i.e., alternation of a true possessive construction with a NP/DP-level possessive element and its quasi-synonym, a pseudo-possessive construction with a case-marked verbal argument. 1. Possessor Raising and Possessive Shift In descriptive terms ―PR‖ refers to a quasi-synonymic transformation where a phrase-level possessive operator located in a NP/DP and expressed by a clitic/free pronoun/NP is made an argument of the clausal predicate, cf. Szabolcsi (1983). PR and other instances of raising can be analyzed as stages in the derivation of a sentence, cf. Don-Won Lee (2004), Pei-Jung Kuo & Yi-An-Lin (2008). In Slavic, specifically, in Russian studies, a description of possessive constructions in terms of PR became standard after the publications of Alexander Kibrik, cf. (2000) and (2003), though there is no consensus whether underlying possessivity should treated as a purely semantic relation, cf. Seliverstova (1990) or as syntactic feature associated with some presumably prototypical possessive constructions, be it bare genitive NPs in all world‘s languages, cf. Kibrik (2003: 307) or language-specific constructions like the Russian construction with the preposition u + genitive NP, cf. Mel‘čuk and Iordanskaya (1995), Rakhilina (2000: 54). The term ―Possessive Shift‖ is introduced in this paper for a quasi-synonymic relation of sentences with a possessive operator that may be pragmatically equivalent in some contexts but have a different syntactic structure. I argue that different locations of a possessive element may either change syntactic structure or preserve it, depending on the value morphosyntactic parameters assume in a given language. In some Slavic languages phrase-level and clauselevel possessive operators are marked with different morphological cases. For instance, Russian phrase-level possessives are genitives while Russian clause-level possessives are datives. Pairs of sentences like a) Rus. Ona ne [NP doč’ Petrova-Gen/ego-Gen doč’] ‗She is not Petrov’s daughter/his daughter‘ ~ b) Ona emu-Dat ne doč’/ Petrovu-Dat ne doč’‘ should 1 The paper is written with financial support from the Russian Ministry of Education and Culture, project 20121.2.1-12-000-3004-9859 «Models of Case Grammar and Natural Languages». The author is grateful to the audience of SLS-6 conference (Aix-en-Provence 2011) and, personally, to Steven Franks, Frank Gladney and Grigory Kreidlin for the valuable comments. I am also grateful to the anonymous reviewers for their criticism. All responsibility for shortcomings is my own. 2 Anton Zimmerling be analyzed as Possessor Shift, i.e., as an alternation between different structures. In Modern Russian Possessive Shift cannot be analyzed as Raising since both the syntax and the argument marking in (a) and (b) are different. In Old Church Slavonic, Modern Bulgarian, and Macedonian Possessive Shift is bound to the use of pronominal clitics. These are marked by the same overt case (dative), both on the phrase-level and on the clause-level. For this group of languages, a PR analysis of clausal possessive forms remains possible. In Modern Serbo-Croatian, dative possessives in clausal 2P are marginally acceptable, according to Pennington (2010), but phrase-level dative possessives in SC are ungrammatical. A similar result can be shown for Modern Russian, a language lacking pronominal clitics. Here, clauselevel dative possessive pronouns are marginally acceptable, cf. (1a), while phrase-level dative possessives are ungrammatical, cf. (1b). (1) Russ a. Ja sebe ne vrag. I REFL.DAT. not enemy ‗I am not an enemy for myself‘. b. *Ja vstretil vraga sebe. I met enemy REFL.DAT Intended meaning: ‗I met my own enemy‘. The two oldest Slavic idioms—Old Church Slavonic (OCS) and Old Northern Russian (ONR)—exemplify two extremes: ONR completely lacked dative possessives, while in OCS they were common both in clausal-second position (2P)2 and on the phrase-level. The clausal 2P typically hosts Slavic argument and reflexive pronominal clitics, cf. DimitrovaVulchanova (1999), Franks & King (2000), Zimmerling (2008). Following Zimmerling et al. (2013), I claim that the majority of Slavic languages only have clause-level possessives and that these pattern with argument dative clitics. Modern Bulgarian and Macedonian give the best chances to check the Raising hypothesis, since these two languages both seem to have DP-level dative clitics (cf. Mišeska Tomić (2004), Franks et al. (2004)) and mechanisms allowing for extracting dative clitics out of DP. In Franks & King (2000) and Franks (2008) pronominal and auxiliary Bg clitics are analyzed as verb-adjacent proclitics in syntax but at as phonetic 2P enclitics that cannot stand clause-initially due to a presumably non-syntactic condition. Dimitrova-Vulchanova (1999), Zimmerling (2012) and Zimmerling & alii (2013) analyse all Bg clustering clitics as 2P elements. There are two competing accounts of Bulgarian possessive clitics. According to Schűrcks & Wunderlich (2003), Bulgarian dative possessives raise out of DP and a position overtly resembling clausal 2P3 where they cluster with other 2P clitics, such as Bg. yes-no particle li, in (2a) and (2b). (2) a. Bg. Tja nameri=li [DP užasni-te=si greški]? She found.PST3.SG. Q horrible-the REFL.DAT. mistakes I am adopting a traditional view that ‗clausal 2P‘ is a position or a block of adjacent positions that can be defined in terms of syntax, cf. Progovac (1996) or syntax-prosody interface, cf. Zaliznjak (2008) and do not take into account an alternative hypothesis raised by Agbayani & Golston (2010) who argue that ‗2P‘ is an epiphenomenal notion and claim that clitic hosts of 2P clitics always lie clause-external to them. Agbayani & Golston‘s analysis is difficult to apply to Slavic clustering clitics. 3 The exact definition of the position taken by Bulgarian dative possessives depends on the analysis of Bg pronominal and auxiliary clitics – whether they are explained as verb-adjacent elements in syntax or as 2P elements. I am adopting a 2P analysis to Bg clustering clitics and gloss them as enclitics (X=CL) in this paper. The problem of clitic hosts is however of minor importance for the analysis of examples like (1b), since clauselevel clustering clitics and clusters like =li=si in (1b) end up in verb-adjacent positions, as an anonymous reviewer justly points out. 2 Possessor Raising and Slavic Clitics 3 ‗Did she find her horrible mistakes?‘ b. Bg. Tja nameri=li=si [DP užasni-te___ greški]? She found.PST3.SG. Q REFL.DAT. horrible-the mistakes ‗Did she find her horrible mistakes?‘ According to Cinque & Krapova (2013) Bulgarian possessives do not raise in sentences denoting inalienable possession (3a-b) and are never extracted out of PPs, cf. (4). (3) (4) Bg. Bg. a. Toj mi= se= izkrjaska [PP v [DP uxoto ]]. he me.DAT.SG. REFL.ACC shouted.PST.3SG. in ear.the ‗He shouted in my ear‘. b. Toj se= izkrjaska [PP v [DP uxoto=mi ]]. he. REFL.ACC shouted.PST.3SG. in ear.the me.DAT.SG ‗He shouted in my ear‘. *Az ì= mislja I her.DAT.3SG.F. think.PRS.1SG. Intended: ‗I think of her eyes‘. [PP za [DP očite __]] for eyes.the I adopt Cinque & Krapova‘s view and treat Bulgarian examples with inalienable possession like (3a-b) as instances of Possessive Shift, i.e., as pairs of different syntactic structures with a possessive operator, along the same lines as Russian examples (1a-b). Bulgarian sentences with alienable possession, such as (4a-b), seem however to be compatible with a PR analysis. 2. Possessivity: Semantics and Syntax In this section, I render the basic facts concerning the semantics and syntax of possessive constructions. A seminal analysis of Russian existential and possessive predicates has been proposed in Arutyunova & Shiryaev (1983) who analyze both semantic and pragmatic / communicative features of Russian possessive constructions. An analysis in terms of PR has been proposed for Russian by Kibrik (2000), (2003) who takes for granted that the prototypical function of genitive phrases in Russian and beyond is ‗encoding of the possessive semantic relation‘.4 However, the interpretation of all phrases of the type [NP N° Ngen] as encoding the possessive relation is too broad and does not provide any basis for separating possessive relation of the type ‗X has/owns Y‘ from the WHOLE : PART relation, the characterization relation ‗X has feature Y‘ etc. which are analyzed usually differently since Arutyunova & Shiryaev (1983), cf. also Rakhilina (2000: 36-55). Therefore, some semantic and syntactic conditions must be imposed on the configuration of Possessors and Possessees. Mel‘čuk & Iordanskaja (1995) and Rakhilina (2000: 54-56) discuss the distribution of Russian constructions ‗Y X-a‘ and ‗u X-a est’ Y‘ at some length and arrive at the conclusion 4 The feature ascribed in Kibrik (2003:, 307) to all genitive phrases is more likely to be associated with a bulk of constructions called ‗weak genetives‘ in Grashchenkov (2007). Standard (i.e. ‗strong‘) genitive markers link together two arguments, expressed by two NPs, X and Y. The tag ‗weak genetives‘ is reserved for languages which apply the same markers both for linking two nominal expressions together as well as for adjoining relative clauses, adjectives, pronouns. Grashchenkov (2007: 44) claims that weak genitive markers tend to be phrasal affixes, not true case affixes. If this claim is true, the notion of ‗weak genitive construction‘ does not fit to most Slavic phrase-level possessive operators. 4 Anton Zimmerling that both constructions encode the whole-part relation, though in a different perspective: in a bare genitive construction the most communicatively salient component is WHOLE, while the u + gen construction brings about a PART-oriented perspective, cf. Rakhilina (2000: 54).5 Mel‘čuk & Iordanskaja (1995: 152) and Rakhilina (2000: 156) in this context claim that in the Russian u + gen construction, cf. U MašiGEN glazaNom.PL. golubyeADJ.NOM.PL., ‗Masha‘s eyes are blue‘ the possessor is always focalized. The term ‗focalization‘ seems to be infelicitous here, since under a standard reading cf. (5a), the possessor is topicalized (thematic), not focalized, while true focalization is Russian normally require a different word order and accent marking, cf. (5b). In the notation of (5a-b) I use accent tags ‗‘ for a standard Russian topic accent (a steep rise followed by a steep fall on posttonics if any) and ‗‘ for a standard Russian focus (rheme) accent; the lowercase tages ‗T‘ and ‗R‘ stand for Topic and Focus respectively, while the lowercase tag ‗R Proper‘ refers to the part of the Focus constituent which takes the focus accent. The lowercase tag ‗Contr‘ indicates that the corresponding communicative constituent is contrastive. Curly brackets stand for communicative constituents, square brackets stand for formal constituents. (5) Russ. a. b. {T [PP U Maši]} {R [NP glaza] {R Proper [PredP golubye}]}. By MashaGEN. eyesNOM.PL. blueNOM.PL ‗Masha‘s eyes are blue‘. { T CONTR [NPGlaza] [PredP golubye]} {R CONTR [ PP u Maši]}. EyesNOM.PL. blueNOM.PL by MashaGEN. ‗It is Masha (not someone else) whose eyes are blue‘. Kibrik (2003: 309) introduces another term for Mel‘čuk‘ & Iordanskaja‘s ‗focalized possessors‘ and claims that Possessor is ‗a local focus of empathy in a genitive phrase‘. This is a terminological improvement. However, it is clear that Kibrik‘s syntax-oriented approach to genitive NPs (and to other structures analyzed in his theory as derived from genitive phrases by virtue of PR and other transformations) is not directly compatible with Mel‘čuk‘ and Iordanskaja‘s semantics-oriented approach. Any WHOLE-oriented structures, including Russian bare genitive NPs (i.e. prototypical possessive phrases in Kibrik‘s theory) won‘t be acknowledges as conveying possessive semantics in Mel‘čuk & Iordanskaja‘s and Rakhilina‘s theories. A further problem with Mel‘čuk & Iordanskaja‘s and Rakhilina‘s description of Russian data is that it is not framework-neutral and strongly relies on principles of construction grammar and dependency syntax. In Mel‘čuk‘s dependency syntax where the notion of constituency does not play any role, the distribution of bare genitive phrases of the type ‗Y X-a‘ and preposition structure ‗u X-a Y‘ can be interpreted as a partial contrast of two Russian constructions each of which has its idiomatic semantics. This approach is illustrated by contexts where the segment ‗Y X-a‘ can be substituted with a segment ‗Y u X-a‘, cf. Russ. <Ščeki Maši> vspyxnuli ‗Maša‘s cheeks flushed slightly‘ vs <Ščeki u Maši> vspyxnuli ‗the same‘ without a clear semantic difference, cf. Mel‘čuk & Iordanskaja (1995: 147) as well as by contexts where such a substituted cannot be made and only one construction is possible, cf. Russ. <Glaza u Maši> golubye ‗Maša‘s eyes are blue‘ but not *<Glaza Maši> golubye ‗Maša‘s eyes are blue‘. A substitution procedure in contexts like <Ščeki Maši> vspyxnuli ~ <Ščeki u Maši> vspyxnuli brings about an illusion that expressions like ‗Y X-a‘ and expressions like ‗Y u X-a‘ are possessive operators of the same level. This is entirely This point defended by Mel‘čuk & Iordanskaja (1995) and Rakhilina‘s (2000) goes back to Kreidlin‘s analysis (1979), which is explicitly stated in (Rakhilina 2000: 54). 5 Possessor Raising and Slavic Clitics 5 misleading and will be proven wrong in any version of constituency analysis. Indeed, an adjacent position of the Possessor (X) and Possessee (Y) in sentences <Ščeki u Maši> vspyxnuli or <Nožka u stola> gnilaja ‗<The leg of a table> is rotten‘ results from a movement transformation, namely, extraction of a head noun out of an NP. This transformation has a clear semantic load. In example (6a) the whole VP is focal, with the focus accent on the NP ščeki (which has the role of Possessee). In the derived structure (6b) this NP is topicalized and made part of the Topic, while the verb vspyxnuli is in narrow focus. (6) Russ. a. b. {T [PP U  Maši]} { R [VP vspyxnuli { R PROPER [NP  ščeki]}}. By MashaGEN. flush.slightlyPST.PL. cheekNOM.PL. ‗Masha‘s cheeks flushed slightly‘. {T {T PROPER [NP ščeki]i} [PP u Maši]} {[VP vspyxnuli] ti }.6 cheekNOM.PL. by MashaGEN. flush.slightlyPST.PL. ‗the same‘. That the moved NP [NP ščeki] forms in the topicalized structure (6b) one and the same communicative constituent with the PP [PP u Maši] certainly does not prove that they form one formal constituent. It is evident that PPs of the type ‗u + gen‘ are not phrase-level, but clause-level possessive operators.7 Following Seliverstova (1990) and a bulk of preceding literature going back to Benveniste (1960), I specify that as a semantic relation, the possessive relation is strictly binary and assume that imposing a condition that the Possessor is animate gives the best chances for keeping apart possessive relation ‗X has/owns Y‘ from other semantic relations, including the WHOLE : PART relation ‗Y is a part of X‘ and the attributive relation ‗X has a feature Y‘. The Possessee is normally inanimate. In a typical communicative reading, the Possessor is a Topic/Theme, while the Possessee and the verb/auxiliaries are parts of the Focus/Rheme; cf. Arutyunova & Shiryaev (1983). (7) Russ. { T [ PP U Maši]} {R [VP est‘ By MashaGEN. bePRS ‗Masha has a flat in Lubercy‘. kvartira [PP v Lubercax]]}. flatNOM.SG.F. in LubercyLOC The possessive relation intersects with the WHOLE : PART relation, though not all sentences expressing the WHOLE : PART relation are possessive in the specified sense. The possessive relation is grammaticalized in a different way with alienable and inalienable possession, cf. Aikhenvald (1998: 93). In some languages this asymmetry also affects the whole : organic part relation, but most Slavic languages keep them apart. In Modern Russian the meaning ‗the leg of a table‘ can be expressed both by a bare genitive NP, cf. [NP nožka [NP stola]] and by a PP with a preposition ot ‗of‘, ‗from‘, cf. [NP nožka [PP ot [NP stola]],8 but with an animate Possessor the insertion of ot is no longer possible: rot Maši ‗Masha‘s mouth‘, but not *rot ot Maši. Claims raised elsewhere that Russian and probably other Slavic languages belong to the class of world‘s languages which allow inanimate possessors, cf. Herslund & Baron (2001) are based not on semantic considerations but on the observation that these languages Remarkably, in Kibrik‘s theory (2000: 308) instances with a topicalized Possessee are classified with PR (ekstrapozicija vneshnego possessora) not with Possessee extraction (ekstrapozicija obladaemogo). 7 This fact is acknowledged by Russian academic grammars as well, cf. Russkaja grammatika (1982: 149-151), where u + gen phrases are recognized as the so called ‗determinants‘, i.e. immediate daughters of S. 8 Selective restrictions on the use of Russ. ot are mentioned in Rakhilina (2000: 43). 6 6 Anton Zimmerling apply similar constructions for encoding possession and for expressing other semantic relations, cf. (8a-d) and (9a-d). Examples (8a) and (8c) where the Possessor is a Topic and the remaining part of the sentence is in Focus can be regarded as basic, while variants (8b) and (8d) show topicalization of the Possessee. (8) Russ. a. {T [PP U Maši]} {R [VP BE.PRS.IND. [PredP [NP gniloj zub]}. By MashaGEN. rottenNOM.SG.M. toothNOM.SG.M. ‗Masha has a bad (lit.: rotten) tooth‘. b. {{T [NP  zub] i [PP u Maši]} {R [VP BE.PRS.IND. [PredP gniloj ti ]}}. toothNOM.SG.M. by MashaGEN. rottenNOM.SG.M. ‗Masha‘s tooth is rotten‘. c. {T [PP U Maši]} { R [VP sgnil. [NP zub]}. By MashaGEN. rot.awayPST.SG.M. toothNOM.SG.M. ‗Masha‘s tooth rotted away‘. d. { T [NP  Zub] [PP u Maši]} { R [VP sgnil]}.9 ToothNOM.SG.M. by MashaGEN. rot.awayPST.SG.M. ‗Masha‘s tooth rotted away‘. In a similar way, in a construction encoding the WHOLE : PART relation, one gets two variants with a topical argument with the role ‗WHOLE‘ and the rest of the sentence in broad focus, cf. (9a) and (9c) and two variants with a topicalized argument with the role ‗PART‘, cf. (9b) and (9d). (9) Russ. a. {T [PP U stola]} {R [VP BE.PRS.IND. [PredP [NP gnilaja nožka]}. By tableGEN. rottenNOM.SG.F. legNOM.SG.F. ‗The table has a rotten leg‘. b. {{T [NP  nožka] i [PP u stola]} {R [VP BE.PRS.IND. [PredP gnilaja ti ]}}. legNOM.SG.F. by tableGEN. rottenNOM.SG.F. ‗The leg of the table is rotten‘. c. {T [PP U stola]} { R [VP sgnila. By tableGEN. rot.awayPST.SG.F. ‗The table‘s leg rotted away‘. d. { T [NP  Zub] [PP u Maši]} { R [VP sgnila]}.10 LegNOM.SG.F. by tableGEN. rot.awayPST.SG.M. ‗The leg of the table rotted away‘. [NP  nožka]}. legNOM.SG.M. The parallelism of (8a-d) and (9a-d) is evident, but it can be interpreted differently. A straightforward solution is to admit that Russian allows inanimate possessors. This kind of analysis, is however based on the assumption that all non-locative uses of the u + gen 9 The linear order in (8d) can also be linked with a different communicative reading where the extracted NP is focalized and the rest of the sentence is deaccented: {R [NP  Zub] [PP u Maši]} [VP sgnil]}. 10 The linear order in (9d) can also be linked with a different communicative reading where the extracted NP is focalized and the rest of the sentence is deaccented: {R [NP  nožka] [PP u stola]} [VP sgnila]}. Possessor Raising and Slavic Clitics 7 construction are possessive.11 The validity of this assumption is far from self-evident and it is reasonable to check an alternative hypothesis that there is no direct mapping between possessive semantics and the structure of the u + gen construction. Constructions expressing the possessive relation are labeled ―possessive‖, though they usually express non-possessive meanings in the same language as well. (10) Russ. a. b. U menja est’ By meGEN. bePRS. ‗I have a map‘. karta. mapNOM.SG.F. U menja est’ podozrenie čto P. By meGEN. bePRS. suspicionNOM.SG.N. that P. ‗I have a suspicion that P‘. Russian (10a) and its English equivalent are true possessive sentences, while Russian (10b) and its English equivalent may be called pseudo-possessive, cf. Zimmerling (2000). The notion of ―pseudo-possessivity‖ can also be applied to syntax if there is no real possessive relation between the predicate arguments on the surface level, but such a relation can be reconstructed by postulating a transformation, cf. Szabolcsi (1983). Along these lines, Russian (11a), meaning ‗X sewed a button on Y‘s coat‘, should allow for possessive structure (11b), with the intended meaning ‗Y‘s button‘. However, this analysis is on the wrong track, since the Russian sentence with the reconstructed NP pugovica rebenka is ill-formed, as shown by the ungrammaticality of (11c): (11) Russ. a. Ona [VP prišila rebenku pugovicu]. SheNOM.SG.F. sewPST.NOM.3SG.F childDAT.SG.M buttonACC.SG.F. ‗She sewed a button on the child<‘s clothes>‘. b. [NP pugovica [NP rebenka]. buttonNOM.SG.F. child GEN.SG.M. ‗The child‘s button‘. c. *Ona prišila [NP pugovicu [NP rebenka]]12. SheNOM.SG.F. sewPST.NOM.3SG.F buttonACC.SG.F. child GEN.SG.M.. Intended: ‗She sewed a button on the child‘s clothes‘. It is reasonable to restrict the notion of quasi-possessivity with those situations, where two different possessive constructions seem to be competing, as in (12a-b). (12) Russ. a.?On [vP grubym priemom [VP slomal [NP nogu Aršavina/ego nogu]]. He roughINSTR.SG.M modeINSTR.SG.M. breakPST.3SG.M. legACC.SG.M.A.GEN.SG/his leg ‗He broke Arshavin’s leg/his leg in a rough way‘. b.On [vP grubym priemom [VP slomal [NP Aršavinu/emu] [NP nogu]]. He roughINSTR.SG.M modeINSTR.SG.M. breakPST.3SG.M. legACC.SG.M.A.GEN.SG/his leg 11 It is evident that locative and non-locative uses of Russian u + gen phrases have different properties, cf. U nashego doma tri magazina ‗There are three shops by our house‘(locative u + gen phrase) vs U nashego doma tri dveri ‗Our house has three doors‘ (non-locative u + gen phrase). 12 Russian (11c) is well-formed only in the meaning ‗X made use of Y‘s button and sewed it on some place not necessarily related to Y‘s clothes‘. Such a reading however is unnatural and requires a special context. 8 Anton Zimmerling ‗the same‘, lit. ‗He broke the leg to Arshavin/to him’. Both (12a) and (12b) are well-formed, although standard Russian favors (8b). In Russian, complex NPs with a special possessive marker/possessive genitive to some extent alternate with constructions of other types, though there are exceptions where complex NPs with a bare genitive NP cannot be used; cf. (12c) above and (13b) below. (13) Russ. a. b. 3. [NP probka [PP ot [NP butylki]]]. plugpNOM.SG.F of bottleGEN.SG. ‗The cork from the bottle‘. ?? [NP probka [NP butylki]]. plugNOM.SG.F of bottleGEN.SG. Intended: ‗The cork from the bottle‘. PR and External Dative Possessor In Baker (1988) the term ―Possessor Raising‖ is reserved for the inversion of the arguments of a ditransitive verb, where the [+ Animate] argument takes the position of Direct Object, as in Eng. He gave flowers [to Kate] > he gave Kate flowers, while a construction where a possessive argument has been added to an intransitive sentence, as in Russ. ty u menja molodec ‗you did well‘, is an External Possessor Construction.13 This delimitation seems too strict for two reasons. First, argument inversion in transitive sentences does not hang on possessivity. Second, the opposition of transitive vs intransitive verbs is not always clear-cut. In Norwegian, one and the same verb allows transitive (14a) and non-transitive uses (14b). Instances of prepositional government (14b) can also be analyzed as combinations of phrasal verbs with a post-verbal particle. Sentences where the Possessor is Case-marked with prepositionless Accusative and the Possessor is a PP pattern with PR, as in (14c): (14) Norw. a. Hunden slikket [DP hånden hans]. Dog-the lickedPST hand-the hisPOSS. ‗The dog licked his hand‘. b. Hunden slikket [PP på [DP hånden hans]]. Dog-the lickedPST on hand-the hisPOSS. ‗The dog licked the hand on him‘, lit. ‗the dog licked on his hand‘. c. Hunden slikket ham [PP på [DP hånden]]. Dog-the lickedPST himOBL.SG.M. on hand-the lit. ‗The Dog licked him on the hand‘. Slavic languages provide a parallel to these Norwegian examples, since Slavic Aktionsart prefixes corresponding to Slavic prepositions have functions similar to those of Norwegian prepositives like på ‗on‘ which function as Aktionsart particles. However, this parallel is only partial. Modern Russian allows for structures without PR resembling (14a) and (14b), but not structures with PR like (14c). The term ‗External Possessor‘ (Russ. vnešnij possessor) is used in Kibrik (2000), (2003: 308-318) who, contrary to Baker, extends it to transitive predicates as well. 13 Possessor Raising and Slavic Clitics 9 (15) Russ. a. Sobaka lizala/liznula [NP ego ruku]. DogNOM.SG.F. lickPST.IMPF.3SG.F/PST.PRF.3SG.F. hisPOSS handACC.SG.F. ‗The dog was licking/licked once his hand‘ b. Sobaka polizala [NP ego ruku]. DogNOM.SG.F. ASP-lickPST.IMPF.3SG.F hisPOSS handACC.SG.F. ‗The dog licked his hand (several times)‘. c. *Sobaka liznula ego [PP po ruke]. DogNOM.SG.F. lickPST.PRF.3SG.F. himACC.SG.M. on handLOC.SG.F. Intended: ‗The dog licked his hand (once)‘. In (15a-b) the indeclinable possessive pronoun ego ‗3Sg.M.Poss‘ is NP-internal, just as the indeclinable possessive pronoun hans ‗3Sg.M.Poss‘ in Norwegian in examples (14a-b). In the Norwegian example (14b) one deals with a non-transitive or semi-transitive predicate slikke på handen lit. ‗to lick on one‘s hand‘, where the element på can be analyzed both as a preposition belonging to a PP [PP på handen] or as part of the phrasal verb slikke på ‗to lick on smth‘. Contrariwise, the Russian sentence (15b) is with the predicative polizala is transitive and does not involve any instance of preposition government. Finally, the Norwegian example (14c) demonstrates PR, where the oblique pronoun ham may be viewed as a raised external possessor, while the ungrammaticality of (15c) indicates that the construction with PR in this case is impossible. 4. Slavic Possessive Clitics Slavic pronominal Dative possessive clitics, have non-trivial features and may be used both as NP/DP-level clitics and as clause-level clitics across Slavic languages; cf. Franks & King (2000), Zaliznjak (2008: 35). Unambiguous DP-level clitics are attested in Bulgarian and Macedonian (cf. Franks et al. 2004) while the nature of the NP/DP-distinction in Polish remains controversial (cf. Rutkowsky 2002). There are two competing views of Bulgarian Dative possessive clitics. Schürсks & Wunderlich (2003), based on examples like (2) above, claim that PR is generalized in this language. On the contrary, Cinque & Krapova (2013) argue that PR is only possible in some sentences expressing alienable possession. Modern Russian has phrase-level Dative possessives but does not allow NP-level Dative possessives. Recall (1), repeated here as (16): (16) Russ. a. b. Ja sebe ne vrag. I REFL.DAT. not enemy ACC.SG.M. ‗I am not an enemy for myself‘. *Ja I vstretil metPST.1SG. vraga enemy ACC.SG.M. sebe. REFL.DAT In (16) the element sebe is a reflexive clitic. Grigory Kreidlin (p.c.) points out that in (16a) the form sebe is no longer used as a reflexive marker and is preferably analyzed as a discourse particle rather than as a pronoun. Irrespective of the validity of this claim, sebe in (16a) is an unambiguous clause-level element, and the ill-formedness of (12b) cannot be explained by the alleged pronoun/particle distinction. It is worth mentioning that in Slavic languages which make use of pronominal Dative possessive clitics, there is no obvious contrast in the syntax of non-agreeing possessive personal pronouns and non-agreeing possessive reflexives. 10 Anton Zimmerling Old Russian had both clause-level possessive clitics and NP-level possessive clitics attached to nominal heads. Clause-level dative possessive clitics normally merged in clausal 2P, while NP-level possessive clitics did not have a fixed position in a clause. Applying this criterion, one can easily establish that the second entry of 2p.Sg. ti in (17) is a NP-level element while the first entry of ti is a clause-level element, since it takes clausal 2P and is not adjacent to any element of a NP. (17) Old.Russ. а. Čto vozdamŭ=ti [PP protivou [NP blagodĕjaniju=ti]]?14 What render.PRS.1SG you.DAT.2SG. for benefaction youDAT.2SG. ‗What can I render for your benefaction?‘. Meanwhile, sentences like (18), where the Dative Possessive clitic splits the clause-initial NP and is placed after a noun, remain ambiguous, since we lack reliable criteria to establish whether we deal with a clause-level element related to the predicate or with a NP-level element. (18) Old. Russ. [NP brata=ti Romana] Bogŭ pojalŭ15. Brother.ACC.SG. you.DAT.2SG. Roman.ACC.SG. God took.PRF.3SG.M. ‗God took from you (your) brother Roman‘ OR ‗God took your brother Roman (from you)‘. Turning back to the Bulgarian data that motivated the contradicting syntactic accounts of PR and Cinque & Krapova‘s hypothesis that in Bulgarian PR is only possible with alienable possession, I restate the basic observations. First of all, Bg blocks extraction out of PPs headed by a lexical preposition, as in (4) repeated here (in a modified notation) as (19). (19) mislja *Az=ìi I her.DAT.3SG.F. think.PRS.1SG. Intended: ‗I think of her eyes‘. Bg. [PP za [DP očite __i]] for eyes.the For the second, in sentences like (3a) repeated here as (16b), the Dative clitic mi is a clauselevel element taking clausal 2P and not a raised DP-level possessive clitic, since such a derivation would involve extraction out of the PP [PP v [uxoto=mi]], in violation of island constraints, cf. Cinque & Krapova (2013) consequently, (20a) and (20b) have different syntactic structures and do not exhibit PR. (20) Bg. a. Toj mi= se= izkrjaska [PP v [DP uxoto ]].16 he me.DAT.SG. REFL.ACC shouted.PST.3SG. in ear.the ‗He shouted in my ear‘. The example is from the Ipatyevskaja chronicle [1199], list 244. The first =ti stands after the verb vozdamŭ, not after the first phonetic word, wh-word čto, since the latter could act as an optional Barrier triggering late clitic placement, cf. Zaliznjak (2008: 55). 15 The example is from the Ipatyevskaja chronicle [1180], list 217. All varieties of Old Russian allowed for inserting 2P clitics into fronted NPs, and in Old Novgorod Russian such clitic placement was obligatory, cf. Zaliznjak (2008), Zimmerling (2012). 16 On reasons specified above in Footn. 3, I gloss Bulgarian pronominal and auxiliary clitics as enclitics. These forms are usually glossed as verb-adjacent proclitics: toj mi=se= izkrjaska. 14 Possessor Raising and Slavic Clitics 11 b. Toj se= izkrjaska [PP v [DP uxoto=mi ]]. he me.DAT.SG. REFL.ACC shouted.PST.3SG. in ear.the ‗He shouted in my ear‘. I claim that (16a) and (16b) differ both syntactically and semantically, (16b) being a true possessive sentence containing a DP [DP uxoto=mi] and (16a) being a pseudo-possessive sentence where the Dative clitic mi is a verbal argument related to the main predicate izkrjaska ‗shouted‘. In other words, pairs of sentences like (16a-b) are examples of Possessive Shift, despite the morphological similarity of the possessive and the quasi-possessive Dative clitics. Cinque & Krapova seem to arrive at a similar conclusion, albeit in a different way. They argue that in spite of the fact that the clause-level dative clitic and the DP-level dative clitic bear the same surface case, they still have different underlying Cases. For them, in (16a) the clause-level clitic mi gets underlying Dative Case and the role of Addressee / Benefactor, while in (16b) the DP-level clitic mi gets underlying Genitive Case and the role of Possessor. This explanation in terms of underlying Cases is possible but it may turn out to be redundant if no decisive proof is found for the fact that Bulgarian has PR in other types of constructions. Indeed, the PR hypothesis can be applied to Bulgarian sentences that do not involve extraction out of a island PP and express alienable possession, as in (2) below as (21). (21) Bg. a. Tja nameri=li [DP užasni-te=si greški]? She found.PST3.SG. Q horrible-the REFL.DAT. mistakes ‗Did she find her horrible mistakes?‘ b. Tja nameri=li=sii [DP užasni-te ti greški]? She found.PST3.SG. Q REFL.DAT. horrible-the mistakes ‗Did she find her horrible mistakes?‘. In (21b) the dative reflexive si takes a position of where Bg clause-level clitics form a clusterclausal 2P and clusters there with another clause-level clitic, the question particle li. In Slavic languages, only clause-level clitics cluster, cf. Zimmerling (2000), Zimmerling et al. (2013),17 but this important fact unfortunately does not provide an independent basis for the verification of the Raising analysis in Bulgarian. Si in (21b) could equally well be a basegenerated element merged in 2P or a raised element extracted out of a DP.18 Consequently, one cannot prove whether the alternation (21a-b) exhibits PR or Possessive Shift. Conclusions Slavic data do not refute the hypothesis of Possessor Raising in Universal Grammar, but this hypothesis has a limited applicability in Slavic languages. The weak aspects of the PR hypothesis are that it takes the mapping of syntax and possessive semantics to be iconic and that the rules/principles of extracting a NP/DP-level possessive operator out of the NP/DP are 17 The verification of this claim depends on the analysis of Bg and Mac DPs containing combinations of a definite article and possessive clitics, cf. užasni-te=si greški in (21a), If we impose a condition that clusters are not occasional sequences of adjacent clitics but fixed combinations of clitics of the same level, combinations like Bg. [[užasni-te]=si ] greški ‗one‘s awful mistakes‘ , Bg. [[mlada-ta]=mu] žena ‗his young wife‘ can be ruled out, since the definite article is likely merged at an earlier stage than the dative possessive pronoun. 18 Unlike Serbo-Croatian, Burgenland Croatian, Slovene, Vojvodina Rusin, Czech and Slovak, Bulgarian lacks Clitic Climbing. Anonymous reviewer points out that the absence of Clitic Climbing in Bg is due to the fact that Bg lacks infinitives and suggests that Possessor Raising and Clitic Climbing instantiate one and the same mechanism. This observation sets an interesting perspective, though Clitic Climbing (i.e. extraction of clitics out of embedded non-finite clauses) finds a much better empirical support than PR. 12 Anton Zimmerling trivial. The analysis has shown that these assumptions are poorly justified. Universal semantic relations may correspond to similar yet not identical syntactic patterns. Constructions expressing alienable and inalienable possession often express other predicative meanings as well. Moreover, sentences with possessive predicates can be pragmatically equivalent to sentences with pseudo-possessive constructions. Slavic languages typically apply different case-marking for non-agreeing phrase-level possessive operators and non-agreeing clauselevel possessive operators. A group of languages including Modern Russian lacks phraselevel possessive clitics: both Russian possessive datives and Russian possessive phrases of the type u +gen are clause-level operators. For this group the PR hypothesis cannot be retained, unless we advance a special syntactic theory, for instance, postulate a mismatch of underlying and surface cases and rules transforming bare genitive NPs into preposition genitive phrases or dative NPs. As attested in modern South Slavic languages, Old Church Slavonic and Old Russian, dative pronominal and reflexive clitics are a special means of marking the possessive semantics. However, true NP/DP-level dative possessive clitics are found in an absolute minority of Slavic languages, notably Bulgarian, Macedonian and Old Church Slavonic, while clause-level dative possessive elements are a more common option. Most cases where the PR hypothesis has been proposed in previous Slavic studies do not conform to the definition of Raising as a syntactic operation, since the identity of structures with a clause-level or phraselevel possessive operator cannot be established. Such cases must be reanalyzed in terms of Possessive Shift, i.e. alternation of a true possessive construction with a NP/DP-level possessive element and its quasi-synonym, a pseudo-possessive construction with a casemarked verbal argument. The PR hypothesis is still possible for a subclass of Bulgarian constructions with alienable possession and a dative clitic, but the alternative hypothesis of Possessive Shift may be applied here as well, since there is no independent verification that the extraction of a possessive clitic has taken place. The most general explanation of the fact that Slavic languages for the most part block for PR as a syntactic operation is that Slavic constructions with non-agreeing NP/DP-level dative possessive clitics seem to be secondary and less common than Slavic constructions with clause-level dative clitics. It is wrong to assume that a Slavic clausal clitic must/may be a raised NP/DP-level element just because the sentence has possessive semantics. A verification procedure is needed. References Agbayani, B. & Golston, C. 2010. Second-position is first-position. Wackernagel‘s law and the role of clausal conjunction, Indogermanische Forschungen 115. Aikhenvald, A. 1998. Warekena. In Handbook of Amazonian Languages, Vol. 4. D.C. Derbyshire & G.K. Pullum (Eds), 225–433. Berlin, New York: Mouton de Gruyter. Arutyunova, N. D. & Širyaev, Yu.N. 1983. Russkoe predloženie. Bytijnyj tip, Moscow. Baker, M. 1988. Incorporation: A theory of grammatical function changing, Chicago: The Chicago University Press. Benveniste, E. 1960. ―Être‖ et ―avoir‖ dans leurs fonctions linguistiques, Bulletin de la Société de linguistique de Paris 55(1): 113-134. Browne, W. 2007. Word Order in Burgenland Croatian: clitics. Talk at the Third Southeast European Studies Association Conference, April 26-28, 2007. Cardinaletti, A. & Starke, M. 1999. The typology of structural deficiency. In Clitics in the languages of Europe. H. van Riemsdijk (Ed.), 245-233. Berlin, New York: Mouton de Gruyter. Cinque, G. & Krapova, I. 2013. The Case for Genitive Case in Bulgarian. In Studies in Generative Grammar. Nominal Constructions. Slavic and beyond. L. Schürcks, U. Etxeberria, A. Giannakidou & P. Kosta (Eds). Berlin, London: Mouton de Gruyter. Possessor Raising and Slavic Clitics 13 Dimitrova-Vulchanova, M. 1999. Clitics in the Slavic languages. In Clitics in the languages of Europe, H. van Riemsdijk (Ed.), 83-121. Berlin, New York: Mouton de Gruyter. Don-Won Lee. 2004. Possessor-Raising in Existential Constructions, Studies in Generative Grammar 14(2): 235-242. Dybo, V.A. 1975. Zakon Vassilieva-Dolobko v drevnerusskom jazyke na materiale Čudovskogo Zaveta, International journal of linguistics and poetics 18(1): 7-81. Franks, S. 2008. Clitic placement, Prosody and the Bulgarian verbal complex, Journal of Slavic linguistics 16(1): 91-137. Franks, S., Junghanns, U. & Law, P. 2004. Pronominal Clitics in Slavic, Journal of Slavic Linguistics 12(1-2): 3-36. Franks, S. & King, T. 2000. A handbook of Slavic clitics. New York: Oxford University Press. Grashchenkov, P.V. 2007. Tipologia possessivnyx konstrukcij, Voprosy jazykoznanija 3: 2554. Herslund, M. & Baron, I. 2001. Introduction: dimensions of possession. In Dimensions of Possession. I. Baron, M. Herslund & F. Sørensen (Eds), 1-25. Amstersdam – Philhadelphia: Benjamins. Kibrik, A.E. 2000. Vnešnij possessor kak rezul‘tat rasščeplenija valentnostej. In Slovo v tekste i slovare. Sbornik statej k 75-letiju Ju.D. Apresjana. L.L. Iomdin & L.P. Krysin (Eds), 434446. Moscow: Jazyki russkoj kultury. Kibrik, A. E. 2003. Konstanty i peremennye jazyka. Sankt-Petersburg: Aletheia. Kreidlin, G.E. 1979. Služebnye slova v russkom jazyke (semantičeskie i sintaksičeskie aspekty ix izučenija. PhD. Moscow. Mel‘čuk, I. & Iordanskaja, L. 1995. *Glaza Maši golubye vs Glaza u Maši golubye: Two Russian Constructions. In I. Mel‘čuk, Russkij jazyk v modeli „Smysl  Tekst”, 135-164. Moscow – Wien: Jazyki russkoj kultury. Mišeska Tomić, O. 2004. The South Slavic Pronominal Clitics, Journal of Slavic linguistics 12(1-2): 213-248. Pennington, J.J. 2010. Kombinovanje objekta adnominalnog posesivnog dativa s dopunama glagola u jednoj klauzi u bosansko-hrvatsko-srpsko-crnogorskom. University of Ohio. Ms. Pei-Jung Kuo & Yi-An-Lin 2008. Possessor Raising, Resumptive Pronouns, and Phases. Paper presented at The 39th annual meeting of the North East Linguistic Society (NELS39), Ithaca, November 7, 2008. Progovac, L. 1996. Clitics in Serbian/Croatian: Comp as the second position. In Approaching second: Second position clitics and related phenomena. A.L. Halpern & A.M. Zwicky (Eds), 411-428. Stanford. Rakhilina, E.V. 2000. Kognitivnyj analiz predmetnyx imen: Semantika i sočetaemost‘. Moscow: Russkie slovari. Russkaja grammatika 1982: Russkaja grammatika, vol. 2. Sintaksis, N.Yu. Švedova (Ed.). Moscow: Nauka. Rutkowski, P. 2002. Noun/pronoun asymmetries: Evidence in support of the DP hypothesis in Polish, Jezikoslovje 3(1-2): 159-170. Seliverstova, O.N. 1990. Kontrastivnaja sintaksičeskaja semantika. Moscow: Nauka. Szabolcsi, A. 1983. The Possessor that Ran away from Home, The Linguistic Review 3: 89102. Schűrcks, L. & Wunderlich, D. 2003. Determiner-Possessor Relation in the Bulgarian DP. In From NP to DP. Vol. 2: The expression of possession in noun phrases. M. Coene & Y. D'hulst (Eds), 121-139. Amsterdam: Benjamins. Zaliznjak, A.A. 2008. Drevnerusskie enklitiki, Moscow: Jazyki slavjanskoj kul‘tury. 14 Anton Zimmerling Zimmerling, A. 2000. Obladat‘ i byt‘ rjadom. In Logičeskij analiz jazyka. Jazyki prostranstv. N.D. Arutjunova & I.B.Levontina (Eds), 179-188. Moscow: Indrik. Zimmerling, A. 2006. Encoding Strategies in Word Order: the Evidence of Slavic Languages. The 1st meeting of the Slavic Linguistic Society, Bloomington, 8-10 September, 2006. Zimmerling, A. 2008. Porjadok slov v slavyanskix, germanskix i romanskix jazykax. In Ot Imen k Faktam (Slavjano-germanskie issledovanija, vol. 3). A.A. Gugnin & A.V. Zimmerling (Eds)., 165-239. Sankt-Peterburg: Aleteia. Zimmerling, A. 2009. ‚Klitiki v prostranstve drevnerusskogo jazyka‗. Review of Zaliznjak (2008), Russkij jazyk v nauchnom osvesjchenii 1(17): 259-277. Zimmerling, A. 2012. Sistemy porjadka slov s klitikami v tipologicekom aspekte, Voprosy jazykoznanija 4: 3-38. Zimmerling, A. et al. 2013. Slavic Clitic Systems in a Typological Perspective. In Studies in Generative Grammar. Nominal Constructions. Slavic and beyond. L. Schürcks, U. Etxeberria, A. Giannakidou & P. Kosta (Eds). Berlin, London: Mouton de Gruyter. Zwicky A.M. 1977. On Clitics. Bloomington: Indiana University Linguistics Club.