(PDF) Verb second languages and clause structure

Verb second languages and clause structure1 Zimmerling, Anton (

) Institute for Modern Linguistic Research SMSUH / Institute of Linguistics, Russian Academy of Sciences, Moscow, Russia Lyutikova, Ekaterina (

) Lomonosov Moscow State University / Institute for Modern Linguistic Research SMSUH, Moscow, Russia Abstract: In this paper, we discuss verb second (V2) languages as an object of the parametric typology. V2 languages are a small group of syntactically uniform languages sharing a number of parameters constraining the clausal architecture and the finite verb placement. We identify the constituting properties of V2 languages, as well as parameters of the cross-linguistic variation among them. We argue that whereas verb movement to the dedicated intraclausal position is indeed involved in the derivation of V2, the verb movement per se is not sufficient to produce the classic V2 profile and must be reinforced by the ―bottleneck condition‖ on the first clausal constituent. Key words: verb-second languages, movement, syntax, parametric typology, information structure, parsing 1. V2 across languages and frameworks. The term ‗Verb-Second Language‘ (V2 languages) refers to a relatively small class of world‘s languages where finite verbs have a fixed position in some types of declarative clauses. Following (Zimmerling 2002), we distinguish between ‗Strict V2 languages‘, e.g. German, Dutch, Afrikaans, Danish, Swedish, Norwegian, where the verb invariably takes the clause- second position, and ‗V1/V2 languages‘, e.g. Old Icelandic, Modern Icelandic, Faroese, Middle Norwegian, Old High German, Yiddish, where both second and first positions (but not the third, the forth, etc.) are licensed. Almost all strict V2 and V1/V2 languages belong to the Germanic group of Indo- European languages, the notable exceptions being Kashmiri and Raeto-Romance, cf. Kaiser (2002). V2 is a grammaticalized formal constraint which seems to be independent from information structure (IS) and prosodic issues. V2 languages share many parameter settings in word order, which can be implemented in rule-based NLP parsing. We assume that the verb placement in every V2 language can be described by a strictly finite set of rules, whereas the operationality of linearization constraints postulated in formal models of language 2 can be checked empirically. Different approaches to V2 parsing are possible. Typologically valid models of e.g. German V2 can with some adjustments be extended to other V2 languages with similar parameter settings, e.g. Afrikaans or Kashmiri, cf. Bhatt (1999). One can also define a set of word order 1 This research has been supported by the Russian Science Foundation, project RFH 14-18-03270 ‗Word order typology, communicative-syntactic interface and information structure in world‘s languages‘. The authors are grateful to the anonymous reviewer for the valuable critical comments. 2 In this paper, we only outline possible implications between parametric typology and rule-based parsers of V2 languages and do not discuss psycholinguistic aspects of parsing. 1 constraints for all V2 languages and adjust it to particular types of V2, which has been suggested by Vikner (1995), Roberts (2012) and Wolfe (2015). The list of parameters of clausal architecture bearing on V2 is given below in section 5. Framework-oriented accounts of V2, e.g. den Besten (1983), Holmberg & Platzack (1995) and Holmberg (2015) work out the insight that verb-second phenomenon is not a primitive feature of a language but a superficial generalization on the clause structure that is triggered by more general mechanisms, notably — verb movement to a position in the left periphery (clausal C-domain) and accompanying movement of phrasal categories to a specifier position preceding the target position of the moved verb. As a consequence, broader definitions of V2 have been proposed that generalize about various instances of clause-internal verb movement. In this paper we argue that although V2 indeed involves verb movement and thus linking V2 to movement is a correct analysis, other conditions have to be met in a strict V2 language, and, therefore, V2 is more complex than a side effect of verb movement. Broad definitions of V2 languages as word order systems where verb movement is characteristic of some declarative clauses are hence counter-productive. 2. Fundamental properties of V2. We use the tag ‗V2 language‘ for syntactic systems fitting two requirements: a) finite verb forms (Vfin) take second position in some type of declaratives, usually – in root (i.e. independent) clauses, cf. Den Besten (1983), but sometimes also in embedded declaratives and in some interrogative clauses, cf. Bhatt (1999); b) the clause-initial position in the diagnostic type of clauses is filled by exactly one constituent — the so called bottleneck condition, in terms of Holmberg (2015). On economic reasons a) and b), as correctly suggested by the anonymous reviewer, can be restated as one general condition: if the second position of a finite verb is generalized in some of type of clauses, there can be exactly one preverbal (i.e. first) constituent in this type of clauses. In the classic structuralist account of V2, cf. Diderichsen (1976), requirements a) and b) are viewed as two sides of one and the same basic condition on clausal architecture, a similar analysis is proposed in Zimmerling (2002:221). However, the current practice in Chomsky‘s framework is keeping a) and b) apart, since verb movement and accompanying movement of phrasal categories to the preverbal position have different motivation. In addition, the exact number of clause-internal preverbal constituents is not always clear, since tentative V2 languages can have topicalization constructions, cf. section 2.3. below. Moreover, some languages which ban V > 3 in a diagnostic type of declarative clauses at the same time license V1 orders in the same type of clauses, whereby the preverbal position is not filled by any overt sentence material, cf. section 2.5. below. Recent generative approaches to V2 work out the insight that the requirements a) and b) are partly independent (Roberts 2012; Holmberg 2015). Bech & Salvesen (2014) and Wolfe (2015) claim that the verb can reach its dedicated position in the clausal left periphery even if the bottleneck condition is not satisfied in any type of clauses: such word order systems are called ‗residual V2 languages‘, cf. the discussion in sections 3.2. and 3.3. below. It is unclear whether the class of non-restrictive residual V2 languages is an extension of the class of restrictive V2 languages. In the next sections, we show that topicalization constructions, V1 orders and multiple XP-fronting can be explained without giving the bottleneck condition. 2 2.1. The bottleneck condition. A descriptive schema of V2 needs three symbols — a symbol of clausal (left) border (#), a symbol of the preverbal constituent (XP) and a symbol for the finite verb (Vfin). (i) #XP — Vfin, *#XP — Y — Vfin The bottleneck condition is crucial for V2 diagnostics. It predicts two features of V2 syntax: a) that a combination of two (or more) phrasal categories X, Y preceding the finite verb in a V2 language should be ungrammatical; b) that the XP-position in the diagnostic type of V2 declaratives is not reserved for any particular syntactic category (e.g. noun phrase) and does not express any particular grammatical relation (e.g. subject).3 In other words, the XP-position in a V2-language can be filled by any element in an OR- expression {Cat1  Cat2  … Catn}, but simultaneous spell-out of two or more hierarchically independent categories in XP is blocked: *#{XP Cat1 & Cat2 &… Catn} — Vfin. Parsing of well-formed V2 structures is licensed by a combination of an OR-expression filter, which lists sentence categories that fill XP in language L, and an &-expression filter which determines which types of expressions count as single constituents when filling XP in language L, and which do not. 2.2. Multiple fronting. All V2 languages are sensitive to constituency rules. Parsers applied to V2 languages must include rules for XP-movement, since the prediction that if α is a constituent and can be inserted post-verbally, it also can be inserted pre-verbally and fill XP, is not always borne out.4 In other words, the regular constraints on the phrasal movement are further adjusted by additional requirements specific for this construction. Thus, Danish strongly favours the PP split in clause-first XP promotion, being more tolerate in other instances of phrasal movement (see, e.g., Zimmerling 2002: 234-6): (1) a. Han har ikke nok [DP stor tillid [PP til [DP min mor]]]. he have.PRS NEG enough big confidence to my mother ‗He surely does not have a big confidence to my mother.‘ b. [XP Min mor] har han ikke nok [DP stor tillid [PP til min mor]]. c. *[XP Til min mor] har han ikke nok [DP stor tillid til min mor]. — Danish, strict V2 Another well-known obstacle to dropping special rules for XP-movement is that XP in some V2 languages hosts not only whole constituents (maximal projections), but also parts of NP/DP or VP etc. Example (1) shows extraction of a head element from an NP/DP, while example (2) shows left branch extraction (LBE).5 (2) a. [XP Útfall] var [DP útfall sjávarinnar]. 3 Roberts (2012) refers to the (b) property of V2 languages as {-EPP}, which is to some extent misleading, because the [EPP] feature in minimalist syntax is responsible for attraction of a constituent to the specifier of a head. As V2 languages in effect attract an XP to the first position of the clause, the incongruity arises. The insight behind Roberts‘ notation is that in a language like English, where T is characterized as having an [EPP] feature, it attracts the subject exclusively, which is not the case with the first clause position in V2 languages. Thus, the label {-EPP} is rather a shortcut of an empirical generalization that the preverbal position is not dedicated for subjects. To avoid confusion we will use Roberts‘ terms in braces. 4 (i)  [# [XP…] —...Vfin…. α  #[XP α] — Vfin… α ] The formal notation of this rule is given in (i): 5 An alternative analysis of (2) involves movement of maximal projections and discontinous spellout of the moved constituent (cf. Nunes 1995, Pereltsvaig 2008) 3 flood.NOM.SG. be.3SG.PRT sea.GEN.SG.DEF ‗The was a flood of tide‘ — Old Icelandic, V1/V2 b Þeirrar skal=tu [DP þeirrar konu] biðja. this.GEN.SG.F shall.2SG=CL.2SG. woman woo ‗You shall woo that woman‘ — Old Icelandic, V1/V2 Finally, in some cases it is impossible to determine whether a sequence Cat 1 & Cat2 forms a single constituent or not without checking its capacity to fill XP. So Norwegian, which is considered a strict V2 language, occasionally licenses sequences of several adverbials in XP, cf. (3). (3) [XP [AdvP I byen] [AdvP i dag]] trefte jeg Marit. [ XP [AdvP In the town] [AdvP today]] met I Marit. ‗Today, I met Marit in the town‘. — Norwegian, strict V2 Both adverbials in (3) have the same IS status (correspond to a Theme), and there are no grounds to believe that any of them is extraclausal. Therefore, one must assume that at least those native speakers who accept (3) generate/parse a single adverbial phrase there. Let us call the option allowing for ‗XP- ensembles‘, i.e. generation/parsing of a single constituent in XP consisting of seemingly independent phrases, ‗multiple XP-movement‘ parameter6. Multiple XP-movement is characteristic for a minority of V2 languages including Modern Icelandic, Faroese and Old Swedish, cf. the discussion in Zimmerling (2002: 291; 501). A similar or identical parameter licenses optional multiple wh-fronting in Kashmiri interrogatives, cf. (4). (4) a. [WhP [WhP Kus] [WhP kemyis] [WhP kyaa]] dii? who whom what give-FUT ‗Who will give what to whom?‘ — Kashmiri, V1/V2 b. [WhP [WhP Kus] [WhP kyaa]] dii [WhP kemyis]? c. [WhP [WhP Kus]] dii [WhP kemyis] [WhP kyaa] ? Other V2 languages, e.g. German, ban structures like (3) and (4a), so multiple-XP-movement and multiple wh-fronting parameters may take different values in V2 languages. Multiple XP-movement/XP- fronting is compatible with V2, insofar the possibility of making a single constituent out of hierarchically independent phrases is restricted to special types of phrases and to contexts where all components of an ensemble have the same IS value – that of a Theme/Topic in (3) or a Focus/wh-word in (4a).7 2.3. Extraclausal topics. A group of V2 languages licenses constructions with a left-dislocated topical element coindexed with a main clause element; this is illustrated in (5), where the dislocated DP su LaRk ‗that boy‘ is coindexed with the resumptive pronoun temis ‗him‘. (5) [DP Su LaRk]i, Rameshan vuch temisi tsuur karaan. 6 A similar parameter allowing XP-ensembles in Clitic-Second languages is discussed in Zimmerling & Kosta (2013:194), Zimmerling (2013: 374-380). 7 XP-ensembles in Verb-Second languages and Clitic-Second languages instantiate multiple movement to a single target position. They could be explained by Sidewards Movement in the model of Bailyn (2012). Note however that XP-ensembles do not necessarily show Superiority Effects, i.e. fixed ordering. E.g. Kashmiri wh-phrases in (4) and Norwegian adverbials in (3) can be placed in whatever order. The same holds for XP-ensembles in Clitic-Second languages like Czech, cf. Zimmerling (2013: 379). 4 that boy.NOM Ramesh.ERG saw he.DAT theft do.N.PERF ‗As for that boy, it is Ramesh who saw him stealing‘. — Kashmiri, V1/V2 Some V2 languages also license structures with a dislocated VP-fragment, like that in (6a), where the resumptive non-referential pronoun det ‗it‘ is coindexed with the whole dislocated phrase läser boken ‗read the book‘, while the resumptive verb gör ‗do‘ is coindexed with the verbal head läser ‗(X) reads‘. (6) a. [VP‘ Läserj boken] i det i görj han nu. read.3SG.PRS book.SG.DEF it do.3SG.PRS he.NOM now ‗He is reading the book, that is what he is doing now‘. — Swedish, strict V2. b. *[VP‘ Läserj boken]i görj han nu. read.3SG.PRES. book.SG.DEF do.3SG.PRS he.NOM now A sentence like (6b) with a topicalized verb phrase in XP and a resumptive verb in V2, is ungrammatical in Swedish. Therefore, it is clear that the initial phrase in (6a) is extraclausal (left-dislocated). Structures like (5) and (6a) are compatible with V2, since formal criteria are met confirming that dislocated topics are extraclausal, and since there is an exact mapping between a marked IS semantics and a choice of the construction in question. 2.4. Verb movement, finiteness and complemetizers. Framework-internal minimalist accounts of V2 elaborate on the idea that the bottleneck condition is impossible without verb movement to a position in the left periphery. In early versions of the Chomskyan framework this domain has been identified as C, since it was believed that Comp and Vfin always have a complementary distribution in V2 languages and compete for one and the same slot, cf. den Besten (1983), Holmberg & Platzack (1995). This seemingly aprioristic claim is based on two empirical generalizations: (ii) In V2 languages finite and non-finite verbs (e.g. infinitives, participles, gerunds etc.) take different positions. (iii) In allegedly prototypical V2 languages (Modern German, Dutch, Danish, Swedish, Norwegian) there is root-subordinate clause asymmetry: in the presence of an overt complementizer, finite verbs do not take V2 and are either placed clause-finally — the West-Germanic option, (7a-b), or one step further to the right, after negation/negative phrases/sententional adverbs — the Mainland Scandinavian option, (8a-b). (7) a. Der Hans hat dem Peter keine Instruktionen gegeben. ‗Hans has not given any instructions to Peter.‘ b. Ich glaube, [CP daß {der Hans dem Peter keine Instruktionen gegeben} hat]. ‗I believe that Hans has not given any instructions to Peter.‘ — German, strict V2. (8) a. Jens har ikke givet instruktioner til Peter. ‗Jens has not given instructions to Peter.‘ b. Jeg tror, [CP at Jens {Neg ikke} har givet instruktioner til Peter]. ‗I believe that Jens has not given instructions to Peter.‘ — Danish, strict V2. The generalization (ii) predicts that verb movement correlates with finiteness feature and verb morphology (TAM markers and inflectional properties). In both subtypes, finite verbs must move in diagnostic declarative clauses, and non-finite verbs must not, and this prediction is borne out. There are 5 no {-EPP} languages such that their clause-second position is occupied by a base-generated finiteness marker, while their clause-first position is not reserved for a specific syntactic category. Thus, many Mande languages have basic word order S AUX O V and overtly resemble to V2 systems. However, this similarity is superficial, since the clause-initial position in those Mande languages we are aware of is invariably reserved for the grammatical subject. The generalization (iii) is more of a technical issue. It states which type of declarative clauses is diagnostic for a V2 language. In the form given above, (iii) is falsifiable, since there are languages where V2 order comes up not only in root declaratives, but also in some subordinate clauses, e.g. Kashmiri (Bhatt 1999), Icelandic (Zimmerling 2002: 303) and Afrikaans. Other V2 languages, including Danish, Swedish and Norwegian, have numerous deviations from (iii) too, therefore some linguists prefer to speak not of the root vs. subordinate clause asymmetry regarding V2, but of ‗subordinate clauses with a subordinate clause word order‘, where the verb does not move, vs. ‗subordinate clauses with a main clause word order‘, cf. Vikner (1995). However, if an overt complementizer like that in (9) does not block V2-orders automatically, such tags are of no explanatory power since they simply restate in other terms the input information ‗this clause has V2 order‘ or ‗this subordinator can license/must license V2 order‘. (9) XP Vfin S O Vinf а. raath dyut laRkan tswaTh daar-yith. yesterday gave boy.ERG waste.NOM throw-out ‗Yesterday the boy threw out waste‘. XP Vfin Comp XP Vfin S O Vinf b. tem dop [CP ki raath duyt laRkan tswaTh daar-yith]. ‗He said that yesterday the boy threw out waste.‘ — Kashmiri, V1/V2 2.5. V1 and V>2 orders. Different positions of finite verbs in other types of clauses, i.e. V1 orders in interrogatives and conditionals without an overt complementizer, are often considered as an additional proof that finite verbs move in V2 languages. However, V1 orders in interrogatives, imperatives and conditionals lacking an overt complementizer, as well as subject-verb inversion and adjacent verb-subject orders are not diagnostic for V2 languages, contrary to the claims made in Salvesen & Bech (2014). As stated by Kaiser (2002), Kaiser & Zimmermann (2015), these features are widely attested in non-V2 languages that do not fit the ‗bottleneck‘ condition in declaratives, e.g. in Spanish, Italian, French, Middle Romance languages, Basque, Estonian etc. They have different triggers: thus, VS orders in Spanish or Italian are not bound to the presence of a preverbal non-subject constituent, there are varieties of Germanic V2 languages that lack V2 orders in wh-questions, etc. Verb-subject adjacency of postverbal subject DPs #... [V-S]… is not diagnostic either: as noted in Bhatt (1999) and Zimmerling (2002: 490; 2013: 188-195), many V2 languages (German, Dutch, Kashmiri) with scrambling in the middle field (i.e. between Vfin and Vinf) lack fixed slots for a postverbal subject DP. 6 2.5.1. V1 orders. V1 orders in yes-no questions, imperatives and marginally acceptable V1 declaratives in strict V2 languages8 are usually explained in generative literature by postulating invisible operators or silent topic elements in XP, cf. Platzack (2008). These are framework-internal explanations characteristic of theories that crucially rely on the assumption that V2 languages always have an overt or silent syntactic category in front of the moved verb. For parametric typology such stipulations are redundant, if one explicitly specifies that in each V2 language V2 orders are restricted to some diagnostic group of clauses, and that in certain clauses the verb moves higher than (the target position of) V2. The functional motivation for this proposal is that V1 clauses have a different illocutionary force than V2 declaratives and it is preferable not to masque this fact by claiming that overt V1 and overt V2 have the same underlying structure. As for V1 declaratives in V1/V2 languages, we raise a stronger claim (iv): (iv) V1 declaratives in V1/V2 languages are IS-marked and formally derived variants of V2 declaratives. An analysis of Old Icelandic, Middle Norwegian, Modern Icelandic put forward in Zimmerling (2002: 363-366) shows that V1 declaratives in such V1/V2 systems are found in a wide variety of different contexts, and the tag ‗narrative inversion‘ is just a descriptive convention, since V1 items can be: a) thetic sentences, b) structures with a dislocated focus, c) structures with a focalized verb, d) structures with a topicalized verb. This approach to V1 and V2 orders can also be extended to non-V2 languages with verb fronting, like Russian, cf. Yanko (2001), Zimmerling (2013: 280-283), and Ossetic, cf. Lyutikova & Tatevosov (2009). To put it more correctly, the analysis of scrambling patterns in {- EPP} languages which has already become standard for languages like Russian, cf. Kovtunova (1976), Bailyn (2004), can be easily extended to V1/V2 languages, since there is no evidence that a ban on V>2 declaratives has any impact on IS-motivated derivation of V1 orders. 2.5.2. V>2 orders. We are aware of three types of well-formed structures with a ‗late placement‘ of the finite verb in a diagnostic group of declarative clauses. 1) Constructions with a dislocated extraclausal topical constituent, cf. examples (5) and (6a), have been discussed in 2.3. They do not involve V2 violation, since the dislocated constituent is extraclausal, so in point of fact, the finite verb is in the clause-second position. 2) Constructions with an initial ensemble, cf. examples (3) and (4), have been discussed in 2.2. They do not involve V2 violation either, since the initial complex constituent is formed by multiple XP-movement / multiple wh-fronting, which obey special rules of the syntax – information structure interface. The third type of V>2 constructions emerges when the target position of verb movement can be reached by some other sentence category in root clauses. This is a rare option, but it is attested as well. Thus, in Swedish, the modal adverb kanske ‗maybe‘ takes the same slot as the tensed verb and competes with it for C/V2, cf. Platzack (2008); very similar Danish (Diderichsen 1976) and Norwegian (Faarlund et al. 1997) word order systems lack this option. (10) a. Nu kanske Johan inte vill komma. 8 Cf. Ger. #Kenn ich nicht ‗I don‘t know <him, her, them>‘, Sw. <Var är tidningen?> #Så jag nyss på bordet. ‗<Where is the newspaper?> I have just seen it on the table‘. We are not sure whether all such V1 clauses in strict V2 languages are real declaratives and not indirect speech acts. However, some V1 clauses that are marginally possible in lively narrative style seem to instantiate the so called narrative inversion (or, in other terms – focus dislocation). Cf. Ger. ‗#Ging ich neulich zu den Müllers ‗<Listen> So I went recently to the Müllers‘ – quoted after Kaiser & Zimmermann (2015: 4). 7 now MAYBE John not FUT come.INF ʻJohn probably won‘t come now‘. b. Johan kanske inte vill komma. John MAYBE not FUT come.INF ‗the same.‘ c. Vem kanske inte vill komma? Who MAYBE not FUT come.INF ‗Who won‘t probably come?‘ A close parallel to this pattern is found in some Clitic-Second languages, where V2 orders come up in derived structures with a so called Barrier constituent9. With the default word order XP-CL, the clausal- second position is filled by clustering clitics and is of course not available for the verb, cf. (11a). But if the initial topical constituent has Barrier properties, the clitics normally do not attach to it, and the vacant target position is filled by the verb in clauses like (11b), which gives rise to Verb-Second and Clitic-Third orders; see Zimmerling & Kosta (2013: 197-199) and Zimmerling (2013: 445-464) for discussion and further examples. (11) a. [PP Poslije toga] =su dobili pozive u reprezentaciju. after that CL.AUX3.PL. get.3.PL.PERF calls to national. team ‗After that, they have been summoned to the national team.‘ b. [BARRIER [PPPoslije svega toga]] bilo =mi =je After all that AUX.3SG.N.PERF CL.1SG.DAT CL.AUX.3SG. potrebno samo ležati na pijesku. necessary.ADJ.SG.N. only lie.INF on sand ‗After all that, everything I needed was to lie on sand.‘ — Serbian-Croatian- 10 Bosnian, strict CL2, V2 orders in derived sentences with clitics All these options are language-specific and subject to microvariation in genetically and areally related V2 idioms. There is, however, one general conclusion we would like to draw. It is not verb movement itself, but the requirement that target position 2P attracting verbs AND/OR some other sentence category must be filled in a diagnostic group of clauses that is crucial for V2 syntax. 3. Verb movement accounts. In this section, we present framework-internal and framework-external accounts of verb movement and argue that while classic accounts of verb movement and narrow definitions of V2 are operational, combining movement analysis with the so called sentence cartography causes problems. We also argue that all definitions of the so called ‗partial‘ or ‗residual‘ V2 languages, where the ‗bottleneck‘ condition is not imposed on any type of clauses, are not operational, and these tags are misleading. 9 Note however that complementizers and modal words like Sw. kanske ‗maybe‘ are merged (i.e. base-generated) in the clause-second position, while verbs and clustering clitics reach the clause-second by movement. For the sake of space we don‘t discuss here the hypothesis that (parts of) clitic cluster can be base-generated directly in 2P. 10 The examples (11a-b) are both from one and the same Croatian idiom. For the details about this idiom and its parameter settings see Zimmerling (2013: 457-462). 8 3.1. Feature-driven movement to C. In the Minimalist Program and its mathematical formalization, Stablerian Minimalist Grammars (Stabler 1997), movement is feature-driven. Early (up to mid-1990s) versions of Chomsky‘s framework tag the left periphery as a C(omplementizer)P. Functional heads like C or, in later notation, Fin(iteness), to which tensed verbs move in V2 languages, are defined as having an uninterpretable Tense feature (uT-feature) or a set of uninterpretable inflectional agreement features (uPhi-features) attracting the pivotal category – tensed verbs in V2 languages and second position clitics in CL2 languages, cf. Roberts (2012) and Zimmerling & Kosta (2013). If such categories (verbs, clitics etc.) get a fixed position with respect to the clausal left border, this means that verb movement/clitic movement is obligatory, at least within the diagnostic group of clauses — with a stipulation that verb/clitic movement does not take place if the target position is already filled by some other category, as in examples (10a-c) and (11b) above. XP-movement is triggered by an active Edge Feature [+EF], which attracts phrasal categories to the clause-initial position (SpecCP or SpecFinP in the later notation). The moved phrasal category and the moved verb/clitic head (V⁰, cl⁰) form a required Spec-Head configuration for feature agreement, which is a well-formedness condition in the early minimalist syntax. This analysis captures correctly three basic facts about V2 languages: a) verb movement to 2P is obligatory in the diagnostic group of clauses, b) all categories that can fill XP lie clause-internally, c) head movement to 2P and phrasal movement to SpecTP in V2 and V1/V2 languages have grammar-internal motivation and do not depend on IS/prosody, while marked constructions with V>2 orders have IS-triggers. 3.2. V>2 orders and cartography. The classic version of deriving V2 by movement, as we demonstrated, is merely an explanation of clause structure constraints which can be captured in a different framework without movement, e.g. in an early structuralist template analysis of Danish V2 (Diderichsen 1976), in Optimality Theory (Legendre 2001), in LFG, etc. The revival of the researchers‘ interest to V2 is due to the cartography hypothesis which suggests a universal template of multiple functional projections arranged in a fixed order common for all languages. According to Rizzi 1997, the left- periphery of the clause has a fine structure like Force > Topic1 > Interrogative > Topic2 > Focus … Topicn > Finiteness > TP, with a higher projection ForceP marking the clause type/illocutionary force11 and the projection FinP hosting some complementizers like Italian de. The specialized Interrogative and Focus positions attracting fronted wh-words and fronted focal expressions, respectively, are motivated by the fact that fronted wh-word(s) and moved focal expressions often fill the preverbal positions in European languages. Multiple topical positions Top1…Topn are needed partly because topics of different kinds (Hanging Topics, Frame-Setters, Familiar Topics etc.) combine with each other, and partly because most languages have both Top > Foc orders and Foc > Top orders with fronted focal expressions, cf. Frascarelli & Hinterhölzl (2007: 112), Bech & Salvesen (2014: 252). Of course, we are not able to evaluate the cartographic approach here; instead, we concentrate on the implications of a cartographic 11 A cartographic account of subordinate clause V2 orders is that complementizers first merge in Fin, but then move to Force except for some languages like Yiddish or Icelandic which do not show root/embedded clause asymmetry (Roberts 2012). This stipulation simply restates the empirical generalization like ‗complementizer that in language L does not block V2 order‘. 9 analysis for V2 syntax. The obvious question is, therefore, which one of the multiple projections of a finer-structured left periphery succeeds the non-split single C of the previous analyses in being the locus of V2 phenomena. It seems that Fin is generally acknowledged as the successor of C; so, Holmberg (2015), Bech & Salvesen (2014), Wolfe (2015) argue that Fin attracts tensed verbs in V2 languages and presumably in a broader class of languages, the so called ‗partial‘ or ‗residual‘ V2 languages, where verb movement to FinP is not generalized in any group of declaratives. As FinP is dominated by other functional projections which can in principle host multiple XPs, the crucial question is whether cartographic theories retain the restrictive ‗bottleneck‘ condition or give it up. If references to cartography, as in Bhatt (1999: 112), are made just to specify that elements filling XP in an OR-expression {Cat1  Cat2  … Catn} take different slots in the left periphery, some of them being topical, some of them being focal, little if anything changes, except for the claim that XP is a descriptive tag, while exact definitions of its syntactic position come from cartography. The same holds for hybrid accounts of XP-movement, which are based on the idea that only a part of phrasal categories reach the left periphery by movement, while other categories are base-generated there. For instance, Mathieu (2006) argues that only non-subject DPs are moved to the preverbal position in Old French – the language he describes as V2, while subject DPs do not move out of TP12. Beninca & Poletto (2004) make a general claim that only focus elements move, while topic elements are base-generated in the left periphery. If, however, cartographic theories include a claim that both V2 languages and non-V2 languages where the verb moves to the left periphery, but two or more clause-internal categories can precede it, have the same syntactic build-up, multiple issues arise. Sentences with a topical constituent in front of XP are also known in V2 languages, but there, as shown above in 2.3 and 2.5.2, the topical constituent is extraclausal, so examples like (5) and (6a), strictly speaking, show not V>2 orders, but V2 orders with a preceding dislocated phrase. For languages like Old English and Old French — which, according to Bech & Salvesen (2014), both have V2 — there is no independent verification that any of the preverbal constituents is extraclausal, since the ‗bottleneck‘ condition on a single preverbal phrase is violated in all clause types, cf. <S Adv1 Adv2 O V> order in (12) and <S O1 O2 V O3 > order in (13). (12) [DP Goliath] [AdvP par quarante jurs] [AdvP le matin é le vesper] [PP a l‘ost de Israel] vint (Bible 1170) ‗For forty days, Goliath came to Israel‘s army in the morning and in the evening‘. — Old French. (13) 7 [hy] [him] [PP æfter ðæm grimme] forguldon And they them after that cruelty repaid-3PL.PRT 12 Note that the argumentation in favor of a hybrid XP/V2 analysis is somewhat circular, if a linguist attempts at combining the claim that subject DPs in language L do not move with the claim that L is a V2 language. E.g. Old French example (i) overtly shows a verb-final order and a V4 order (if one counts the object proclitic l‘=). The claim that the verb atendi really moves is supported by the observation that there is a pre-verbal non-subject constituents set jours. The claim that the order S Adv CL=V does not violate V2 in Old French is only supported by the stipulation that subject DPs do not move in Old French, but are base-generated [Mathieu 2006]. (i) É [DP Saül] (1) [AdvP set jours] (2) l‘= (3) atendi (Bible 1170) ‗And Saul (1) awaited him (3) seven days (2)‘. 10 [DP þone wigcræft [CP þe [hy] [PP æt him]] leornodon]. (Or. 22) that art.of.war that they at him learn-3PL.PRT ‗And after that, they bitterly repaid him for the art of war that they learned from him‘. — Old English. There are two ways to save the idea that languages like Old French and Old English have V2. The first alternative is to stipulate ad hoc that though inconsistent V2 languages license V>2 orders, only one preverbal constituent lies clause-internally, while other elements are extraclausal. This stipulation largely amounts to the claim that if the verb moves out of TP and is placed clause-internally, the language has V2 syntax: a step in this direction is made in Wolfe (2015). The second alternative is to stipulate that inconsistent V2 systems have verb movement to the clausal left periphery only in some declarative clauses. This conjecture has been put forward for Old English in Bech & Salvesen (2014: 257). Neither alternative provides us with effective and reliable criteria for identifying V2. The main problem is that if a language has scrambling of preverbal elements and can place them in whatever order — S Adv V ~ Adv S V, S O V ~ O S V etc., there are no formal markers indicating which category is extraclausal. A better and more natural solution is to conclude that the entire perspective set out by a cartographic revision of V2 is misleading. ‗Partial‘ or ‗residual‘ V2 languages do not exist, non-restrictive word order systems with verb movement to clause-internal positions cannot be extensions of restrictive V2 or V1/V2 systems. 3.3. CFin-V2 languages and CForce-V2 languages. Most cartographic theories of V2 languages share the view that tensed verbs move to FinP. Following the ideas of Roberts (2012) that any functional head in the left periphery can be the locus of V2 effects, Wolfe (2015) recently proposed a classification of Middle Romance systems with a distinction of CForce-V2 languages, where the verb moves to Force, vs. CFin-V2 languages, where the verb moves to Fin. The first group includes later Old French, Old Spanish and Old Venetian, the second one – Old Sicilian and Old Occitan. The author admits that V2 is not obeyed in any of these languages, but tries to find restrictions on V1 and V3 and correlations between them. His data suggest that in CForce-V2 languages, V1 orders are more restricted, and V3 orders mainly include a Frame-Setting topic, while in CFin-V2 languages, V1 and V3 orders express a wider amount of different IS options. Since the number of languages in each group is very small, it is unclear to us whether Wolfe‘s correlations are typologically valid. The idea of including the availability of V1 in the definition of V2-systems is promising, though we cannot characterize any of the languages in Wolfe‘ selection as a restrictive V2 system13. Wolfe also mentions that the group with freer word order, CFin-V2 languages, allows V4, which allegedly conforms to a cartographic pattern Frame-Setting Topic — Standard Topic — Focus — Verb. If the preverbal phrase pir virtuti divina ‗through divine virtue‘ in (14) is a moved focal element, then Old Sicilian patterns with languages like the non-V2 Modern Russian, which allows Left Focus Movement of VP-elements across the verbal head14. 13 The language with the most rigid order, Old French, still has 24, 53% V3 clauses, according to Wolf‘s own data. 14 On Left Focus Movement and its realization in Modern Russian cf. Zimmerling (2013: 281). Typically, Left Focus Movement in Russian brings about a reinforced focus accent marked below as ‗↘↘‘. Cf. Russ. Zatem, posle peredači partnera, Vasja [vP [FocusProper za sčet ↘↘driblinga]i prošel dvux sopernikov t i ].‗Then, after the a pass from a team partner, Vasja went by two adversary players owing to his dribbling skills.‘. 11 (14) A [AdvP tamen] [PP poy di la morti loru] [DP li ossa loru] And then after from the death of.them the bones their {FocusProper[PP pir virtuti divina]}i operannu⁰ [DP miraculi] ti. (Gregoriu, 262) through virtue divine perform-3PL.PRS miracles ‗Then, after their death, their bones perform miracles through divine virtue‘ — Old Sicilian15. 4. Verb movement in non-V2 languages. In this section we briefly examine two languages which are definitely not V2 in the classic sense (that we share and advocate for here), but still have some properties of ‗non-strict‘ / ‗partial‘ / ‗residual‘ V2. The aim of the discussion below is to show that clause-internal verb movement per se is not sufficient to produce the whole range of phenomena associated with V2 and can actually give rise to quite different and sometimes very peculiar systems. 4.1. Ossetic. The enigmatic clause structure in Ossetic has been a challenge for many formal linguists who attempted to make it fit into the system based on more or less reasonable assumptions about what functional projection constitute the clause and how they are placed with respect to each other (cf. Lyutikova&Tatevosov 2009, Ershler&Volk 2009, Gareyshina et al. 2011, Ershler 2012a,b, Belyaev 2013, 2014, Zimmerling&Lyutikova 2015). The most striking characteristic of the Ossetian clause is that constituents normally found at the left edge of the clause (that is, complementizers and wh-phrases) are located in the preverbal position, that is, clause-internally. Together with negative particle and negative XPs, they form a rigid preverbal complex that cannot be separated from the verb by any argumental XP.16 At the same time, the linear order of other constituents of the clause is – at least superficially – free, so that they can precede or follow the verb giving rise to various IS-interpretations. Thus, in (15) a complex sentence embedding finite complement clause is demonstrated; note that the complementizer kæj ‗that‘ can only occur preverbally, whereas argument XPs can be positioned to the left or to the right of the complementizer and verb. (15) a. æž žon-ın I know-PRS.1SG madinæ jæ fırt-ı kæj arvıšt-a goræt-mæ. M. 3SG.GEN son-ACC that send.PST-TR.3SG city-LAT ‗I know that Madina sent her son to the city.‘ b. ...madinæ jæ fırtı kæj arvıšta gorætmæ. c. ...madinæ jæ fırtı gorætmæ kæj arvıšta. d. ...madinæ kæj arvıšta jæ fırtı gorætmæ. e. ...kæj arvıšta madinæ jæ fırtı gorætmæ. f. *... kæj madinæ jæ fırtı gorætmæ arvıšta. 15 Wolfe takes (14) for an example of V4, though it should rather be tagged V5 on a descriptive level, since the initial adverb tamen hardly belongs to the subsequent PP. 16 The preverbal complex can be splitted by 2P clitics if it happens to start the clause; some adverbials like æftæ ‗so‘ and comparatives also intervene between the preverbal constituents and the verb. 12 g. *...madinæ jæ fırtı kæj gorætmæ arvıšta. h. *... kæj madinæ arvıšta jæ fırtı gorætmæ. i. *... arvıšta madinæ kæj jæ fırtı gorætmæ. (16a-b) show the preverbal complex consisting of an interrogative wh-XP and a negative XP; crucially, the subject XP occupies different positions wrt the finite verb: it is strictly adjacent to the verb when being a negative pronoun and precedes the negative oblique complement when being an interrogative pronoun. The preverbal position of the subject XP in (16a-b) is thus a result of its movement to the preverbal complex in virtue of belonging to the category of NPI/wh-XPs, and not in virtue of being a subject. This reasoning is further supported by (16c) where the regular subject XP occurs postverbally. (16) a. æž žon-ın kæj-mæ niči azırd-ta. I know-PRS.1SG who-COMIT nobody speak-PST.3SG ‗I know whom nobody spoke to‘. b. æž žon-ın či nikæj-mæ azırd-ta. I know-PRS.1SG who nobody-COMIT speak-PST.3SG ‗I know who spoke to nobody‘. c. æž žon-ın kæj nikæj-mæ azırd-ta zalinæ. I know-PRS.1SG that nobody-COMIT speak-PST.3SG Z. ‗I know that Zalina spoke to nobody‘. It follows from (15) and (16) that the verb must undergo a clause-internal movement in order to reach the structural position adjacent to the preverbal complex, thus leaving behind all the constituents c- commanding it in the VP and surfacing in some higher functional projection FP, as in (17). In Lyutikova&Tatevosov (2009 and elsewhere) it is argued that this functional projection is the T(ense)P, but nothing in our current argumentation crucially depends on this particular assumption. Whatever projection the verb moves in Ossetic, it shall contain no [EPP] feature, which otherwise would attract the dedicated XP (i.e. the subject of the clause if FP is indeed the TP) and create an intervenor between the verb and the preverbal complex. (17) FinP NegP Preverbal FP complex … VP Subj V 13 To sum up, the clausal architecture of Ossetic calls for clause-internal verb movement and against grammatical feature-driven subject movement, thus {-EPP} in Roberts‘ (2012) taxonomy. At the same time, Ossetic is hardly a V2 language in the classic sense, as it does not meet the crucial ‗bottleneck‘ condition for (V1/)V2. Needless to say that verb movement does not obligatorily result in complementary distribution of finite verbs and subordinators, even if they show strong positional interactions. Interestingly, Ossetic allows us to make an even stronger claim that multiple XPs preceding the verb (and the preverbal complex) are neither extraclausal nor form an initial ‗ensemble‘. Ossetic 2P clitics clearly show that the preverbal subject is intraclausal in (18a) and does not form an ‗ensemble‘ with other preverbal components in (18c). (18) a. madinæ=mın nisavær zırd žaxta. Madina=CL.1SG.DAT no word said ‗Madina said no word to me.‘ b. nisavær zırd=mın žaxta Madinæ. no word=CL. 1SG.DAT said M. ‗Madina said no word to me.‘ c. *madinæ nisavær zırd=mın žaxta. Madina no word=CL. 1SG.DAT said ‗Madina said no word to me.‘ 4.2. Russian. Russian is an example of a language where word order patterns sometimes mimic the true V2 languages‘ template. Thus, at least since Kovtunova (1976) the pairs of sentences like (18a-b) and (19a-b) are considered as derivationally related, so that the preverbal position is occupied by the topical constituent, and the verb forms a rhematic (wide focus) IS-constituent with the postverbal material. Elena Paducheva (2008 and elsewhere) dubs the (b) examples as involving the subject inversion that she considers as a postsyntactic LA-transformation (examples are from Paducheva 2008). (18) a. [Th Sobaka] [Rh ukusila mal‘čika]. dog.NOM bit boy.ACC ‗The dog bit a/the boy.‘ b. [Th Mal‘čika] [Rh ukusila sobaka]. boy.ACC bit dog.NOM ‗The boy was bitten by a/the dog.‘ (19) a. [Th Lodka] [Rh ležala na beregu]. boat.NOM lay on shore ‗The boat lay on the shore.‘ b. [Th Na beregu] [Rh ležala lodka]. on shore lay boat.NOM ‗On the shore, there was a boat lying.‘ John Bailyn (2012) attempts to provide an intra-syntactic account to the subject inversion; he claims that the preverbal constituent in both (19a) and (19b) (and similar examples) is in the structural subject 14 position, i.e. Spec, TP, and the verb has moved to T.17 Thus, he treats the subject inversion separately from other word order permutations motivated by information structure. If we follow Bailyn‘s analysis of Russian ―generalized inversion‖ constructions, we have to admit that the verb undergo the clause-internal movement out of the VP in order to precede the subject, and the specifier position of the target projection is not allocated to the subject exclusively, that is, Russian is {-EPP} in Roberts‘ (2012) taxonomy. Despite of these properties of Russian, however, we are not inclined to tag it as a ‗partial V2‘ language; and we strongly doubt that other ‗partial‘ / ‗residual‘ V2 languages with similar characteristics exist. 5. Parameters. Table 1. Parameters of the clausal architecture bearing on V2 complementary distribution of Left Periphery restrictions on Diagnostic type of V2 clauses Basic word order Topicalization Ensembles categories Type V>3 CL2 V1 German V2 SOV/SVO root no No no no no Comp/V declaratives Kashmiri V1/V2 SOV all yes No yes yes no — declaratives Old V1/V2 SVO root yes No no no CL2/CL3 — Icelandic declaratives Modern V1/V2 SVO all yes No no yes no — Icelandic declaratives Danish V2 SVO root no No no no no Comp/V declaratives Swedish V2 SVO root no No yes no no Comp/V declaratives kanske/V Norwegian V2 SVO root no no yes yes CL3 — declaratives Old non- SVO/SOV no yes yes yes n/a no — English V2 Old French non- SVO no no yes yes n/a VP- — V2 (rare) internal 17 Bailyn argues that the preverbal constituent in inverted constructions has some properties of subject DPs, namely, those properties that follow from their highest argumental position, i.e. their ability to bind reflexives inside the grammatical (nominative) subject. 15 clitics Old non- SVO no yes yes yes n/a VP- — Sicilian V2 internal clitics Serbian- non- SVO no yes yes yes n/a yes CL2/Comp Croatian- V2 V2/CL Bosnian Ossetic non- SOV no yes yes yes n/a yes Comp/wh- V2 words/Focus Russian non- SVO no yes yes yes n/a no — V2 Old non- SVO/SOV no yes yes yes no yes CL2/Comp Russian V2 In this section we summarize the parameters discussed above as Table 1, which contains classic V2 and V1/V2 languages, as well as non-V2 languages, among them those that recently has been argued to be very close to V2 and to represent a ‗partial‘ or ‗residual‘ V2 subtype. The table shows that despite the parametric variation attested in classic (V1/)V2 languages, the ‗bottleneck‘ condition (V>3) clearly separates them from other languages with verb movement; at the same time, the allegedly ‗partial‘ / ‗residual‘ V2 languages share many parameter settings with the clear instances of non-V2 languages. <insert Table 1 here> 6. Conclusions. In this paper, we have discussed the characteristic properties of V2 languages and the parameters of variation attested among them. This parametric representation of the complex V2 phenomenon can constitute an empirical basis of a theoretical work aiming at modelling the clausal architecture, as well as be implemented in rule-based parsing systems making use of adjustable parameter setting. We have paid special attention to the recent claim that the broader definitions of V2 phenomena are possible which rely on a single parameter of the clause-internal verb movement and abandon other conditions on the clause architecture. We have shown that V2 is a complex phenomenon that cannot be reduced to the verb movement but involves different parameters‘ setting. Hence, non-restrictive word order systems disregarding the ‗bottleneck‘ condition cannot be an extension of restrictive V2 and V1/V2 languages. Abbreviations 1/2/3 – 1st/2nd/3rd person, ACC – accusative, Adv – adverbial, AdvP – adverbial phrase, AUX – auxiliary, CL – clitic, COMIT – comitative, Comp – complementizer, CP – complementizer phrase, DAT – dative, DEF – definite, DP – determiner phrase, ERG – ergative, FinP – finiteness phrase, FUT – future, GEN – genitive, INF – infinitive, IS – information structure, LA – linear-accent transformations, LAT – lative, M/F/N – masculine/feminine/neuter, NEG – negation, NOM – nominative, NP – noun phrase, O – object, PERF – perfective, PL – plural, PP – prepositional phrase, PRS – present, PRT – preterite, PST – past, S 16 – subject, SG – singular, TP – tense phrase, TR – transitive, V – verb, V1 – verb in the 1st position, V2 – verb in the 2nd position, Vfin – finite verb, VP – verbal phrase, wh – interrogative pronoun. References Bailyn, John Frederick. (2012). The Syntax of Russian. Cambridge, Cambridge University Press. Bech, Kristin & Salvesen, Christine Meklenborg (2014). Preverbal word order in Old English and Old French. In: Information Structure and Syntactic Change in Germanic and Romance Languages. Pp. 233–269. Belyaev, Oleg (2013). Superiority effects in Ossetic: structural vs. linear constraints // (Typology of Morphosyntactic Parameters, 15–17 окт. 2013). — Moscow. Belyaev, Oleg (2014). Korrelyativnaya konstrukciya v osetinskom yazyke v tipologicheskom osveshchenii [Ossetian correlative constriuction in the typological perspective]. (In Russian). PhD thesis, Moscow State University. Beninca, Paola & Poletto, Cecilia (2004). ―Topic, Focus, and V2‖. In: L.Rizzi (ed.). The Structure of CP and IP: Oxford Studies in Comparative Syntax. Oxford. Pp. 52–75. Bhatt, Rajesh M. (1999). Verb movement and the syntax of Kashmiri. Dordrecht. Den Besten, Hans (1983). ―On the interaction of root transformations and lexical deletive rules‖. // On the formal nature of the Westgermania /Abraham W. (ed.). John Benjamins, Amsterdam, pp. 47–131. Diderichsen, Poul (1976). Elementary Danish grammar [Elementær Dansk Grammatik]. (In Danish). København, 3rd edition (11946). Erschler, David (2012a). ―Sluicing-like Phenomena and the Location of CP in Ossetic‖ // Proceedings of IATL28 (Tel Aviv University, 16–17 окт. 2012). — [URL: https://www.academia.edu/2494508/Sluicing- like_Phenomena_and_the_Location_of_CP_in_Ossetic ] Erschler, David (2012b). ―From preverbal focus to preverbal ―left periphery‖: The Ossetic clause architecture in areal and diachronic perspective‖ // Lingua, Vol. 122, № 6. Pp. 673–699. Ershler, David & Volk, Vitaly (2009). Puzzles of Digor Ossetic Negation // International conference on iranian languages (ICIL3), September 11–13, 2009, Paris. Faarlund, Jan Terje, Lie, Svein, and Vannebo, Kjell Ivar (1997). The reference grammar of Norwegian [Norsk referansegrammatikk]. (In Norwegian). Universitetsforlaget, Oslo. Frascarelli, Maria & Hinterhölzl, Roland (2007). ―Types of topics in German and Italian‖. // On Information Structure, Meaning and Form / K. Schwabe & S. Winkler (eds.). Amsterdam, John Benjamins. Pp. 87–116. Gareyshina, Anastasiya, Grashchenkov, Pavel, Lyutikova, Ekaterina, and Tatevosov, Sergei (2011). Sentential proforms and complementation in Ossetian // 33.Jahrestagung der Deutschen Gesellschaft für Sprachwissenschaft. Text: Strukturen und Verarbeitung (Georg-August- Universität Göttingen, 23–25 Feb. 2011). [URL: http://www.zas.gwz- berlin.de/fileadmin/veranstaltung_zas/workshops/archiv/proforms/gareyshina_ossetian.pdf ] Holmberg, Anders & Platzack, Christer. (1995). The Role of Inflection in Scandinavian Syntax. Oxford. Holmberg, Anders (2015). ―Verb second‖. // Syntax – Theory and Analysis. An International Handbook. T. Kiss & A. Alexiadou (eds.). Vol.1. Berlin: Mouton de Gruyter, pp. 242–283. Kaiser, Georg & Zimmermann, Michael (2015). ―Exploring verb order differences between (Medieval) Romance and (Modern) Germanic‖. Traces of History Conference. The University of Oslo, March 9–10, 2015. Kaiser, Georg (2002). Verb position and its change in Romance. [Verbstellung und Verbstellungswandel in den romanischen Sprachen]. (In German). Tübingen: Niemeyer. Kovtunova, Irina (1976). Russian syntax: word order and information structure [Russkiy sintaksis: porjadok slov i aktual‘noye chleneniye predlozheniya]. (In Russian). Moskva: ―Prosveshcheniye‖. Lambrecht, Knud (1994). Information Structure and Sentence Form. Cambridge. Legendre, Geraldine (2001). ―Masked second-position effects and the linearization of functional features‖. // Optimality-theoretic syntax. Cambridge. Pp. 241–277. Lyutikova, Ekaterina & Tatevosov, Sergei (2009). The clause internal left edge: Exploring the preverbal position in Ossetian // International conference on iranian languages (ICIL3), September 11–13, 2009, Paris. Mathieu, Eric (2006). ―Stylistic Fronting in Old French‖. Probus, 18. Pp. 219–266. 17 Nunes, Jairo (1995). The copy theory of movement and linearization of chains in the Minimalist Program. Doctoral dissertation, University of Maryland, College Park. Paducheva, Elena (2008). ―Categoricity and ways to overcome it: subject inversion‖ [Kommunikativnaya raschlenennost‘ i puti ee preodoleniya: inversiya podlezhashchego] (In Russian) // Phonetics and Non-phonetics: To Sandro V. Kodzasov on the occasion of his 70th anniversary [Fonetika i nefonetika. K 70-letiyu Sandro V. Kodzasova]. A. Arkhipov et al. (eds.). Moskva, ―Yazyki slavyanskikh kul‘tur‖. — P. 417–426 . Pereltsvaig, Asya (2008). ―Split Phrases in Colloquial Russian‖. Studia Linguistica 62(1), pp. 5–38. Platzack, Christer (2008). The Edge Feature on C. Lund, University of Lund, Ms. Pollock, Jean-Yves (1989). ―Verb Movement, Universal Grammar, and the structure of IP‖. // Linguistic Inquiry 20, pp. 365–424. Rizzi, Luigi (1997). ―The Fine Structure of the Left Periphery‖. // Elements of Grammar: Handbook of Generative Syntax. Dordrecht. Pp. 281–337. Roberts, Ian (2012). ―Phases, Head-movement and Second-Position Effects‖. // Phases developing the framework / Angel J. Gallego (ed.). Berlin, Boston: Mouton de Gruyter, pp. 385–440. Salvesen, Christine Meklenborg (2013). ―Topics and the Left Periphery. A comparison of Old French and Modern Germanic‖. // In search of universal grammar: from Old Norse to Zoque. Pp. 131–172. Stabler, Edward P. (1997). ―Derivational minimalism‖. // Logical Aspects of Computational Linguistics. Christian Retore (ed). Springer, pp. 68–95. Vikner, Sten (1995). Verb Movement and Expletive Subjects in the Germanic Languages. Oxford Studies in Comparative Syntax, New York, Oxford: Oxford University Press. Wackernagel, Jakob (1953). ―Indo-European word order law‖. [Über ein Gesetz der indogermanischer Wortstellung] // Wackernagel J. Kleine Schriften. Bd. I, Basel. Pp. 1–103. Wanner, Dieter (1996). ―Second Position Clitics in Medieval Romance‖. // Approaching Second: Second Posiition Clitics and Related Phenomena. A.L.Halpern & A.M.Zwicky (eds.). Stanford. Pp. 537– 578. Wolfe, Sam (2015). A New Perspective on V2 and the Evolution of Romance Clausal Structure // Traces of History Conference. The University of Oslo, March 9–10, 2015. Zimmerling, Anton (2002). Typological Scandinavian syntax. [Typologicheskiy sintaksis skandinavskikh yazykov] (In Russian). Moscow. Zimmerling, Anton (2013). Slavic word order systems from the viewpoint of formal typology. [Sistemy poryadka slov slavyanskikh yazykov v tipologicheskom aspekte] (In Russian). Moscow. Zimmerling, Anton & Kosta, Peter (2013). ―Slavic clitics: a typology‖. STUF – Language Typology and Universals. Vol. 66. No 2. Pp. 178–214. Zimmerling, Anton & Lyutikova, Ekaterina (2015). Approaching V2: verb second and verb movement // Ко п те а л т ка теллектуал е те оло : о ате ала е е од о Ме ду а од о ко е е «Д ало » Мо к а, 7– а . . п. : т. . : О о а п о а а ко е е . — М.: И д- о , . — . 663–675. 18