Negative Concord in Russian An Overview

In this article I will describe the general properties of Negative Concord in Russian, which is a strict Negative Concord language, where all negative indefinites must co-occur with sentential negation. However, there are several cases where the negation marker can be absent (like in fragment answers) or can appear in a non-standard position (like at the left of an embedded infinitival). I will take into consideration all these specific cases described by the literature on the negation system of Russian and analyse them according to current approaches to Negative Concord.

In this paper I illustrate some peculiarities of the Russian Negative Concord system and discuss them in the light of current analyses of Negative Concord.
The presence of multiple negative items in the same clause is a phenomenon with some consequences for both the syntax and the semantics of negation. Following Zeijlstra (2016) it is possible to individuate two main groups of analyses of Negative Concord systems. Some, like Haegeman and Zanuttini (1996), De Swart and Sag (2002), Watanabe (2004) and others, assume that all n-words, i.e. negative indefinites and adverbs, always introduce a semantic negation (i.e. they are negative quantifiers) and when multiple n-words are present some kind of absorption mechanism produces a unique sentential negation. The main problem of similar approaches is that they do not predict why in languages like Russian the preverbal negation must always be present (cf. Horn's 1989 NegFirst criterion). The other type of explanations (e.g. Ladusaw 1993;Giannakidou 2000;Zeijlstra 2004;Haegeman, Lohndal 2010) is based on the assumption that n-words are negative polarity items (NPIs) or a special type of NPIs: they are similar to English any-terms, which have to be licensed by a negative operator and do not convey a negative interpretation per se. The main problem for this approach is to explain the difference between strict Negative Concord languages like Russian and nonstrict Negative Concord languages, like Italian, where preverbal nwords are not accompanied by preverbal negation:  In what follows I examine the properties of Russian Negative Concord. In section 2 I present fragment answers with n-words. In section 3 I discuss some other cases where n-words appear without the preverbal negative marker. In section 4 I present some cases where n-words seem to be licensed by a lower negation. In section 5 I briefly present the diachronic development of Negative Concord in Russian.

Fragment Answers
An important property of Russian n-words is that they can appear in isolation in fragment answers and convey negative semantics: This is a context where these items encode negation without the preverbal negative marker ne. Pereltsvaig (2004) has proposed that these cases are to be analyzed as elliptical constructions, where the preverbal negation is elided with the verb. If Russian n-words are negative quantifiers, these data do not need a specific account, since it is assumed that they are inherently negative. Recall, however, that in standard non elided clauses the sentential negation is mandatory with all these items. On the other hand, approaches assuming that n-words are NPIs need to postulate that either the deleted negation or an abstract negative marker can license the n-word. Fălăuş and Nicolae (2016) present an interesting property of nwords in strict Negative Concord languages (they mainly discuss Romanian data, but Russian behaves in a similar way). When isolated n-words are used as answers to negative questions, they are ambiguous between a Negative Concord and a double negation reading: This is unexpected since the double negation reading is not possible in the non-elided version of the answer. Fălăuş and Nicolae (2016) accept the hypothesis that n-words are a special type of NPIs, able to license a covert negation (CN) operator, as a last resort strategy, when the vP is not spelled out. The double negation reading corresponds to a structure where the n-word has moved to a higher position like focus, where it is c-commanded by the covert negation operator and the sentential negation (SN) is still present underlyingly: Notice however that this explanation is adequate for Romanian or Greek, where a double negation reading is marginally possible with multiple n-words 1 , while in Russian there is not an overt version of the structure in (5). 2 A possible alternative is to assume that in the double negation case there is a different type of covert negation which has scope only over the n-word and not over the whole clause. My proposal to analyze these cases assumes that the fragment answer, besides the elided part, can have a Theme/Topic projection containing a silent version of the stimulus question. The negation inside the 1 Fălăuş and Nicolae (2016) point out that in some nc languages a Double Negation reading is possible in negative spreading contexts: (i) Nimeni nu a citit nimic. nobody not has read nothing 'Nobody has read anything' (nc) or 'Nobody hasn't read anything' = 'Everybody read something' (dn). The authors argue that, since this is possible only in negative spreading, i.e. when there are multiple n-words, one n-word is licensed by the sentential negation, but the other one requires the covert negation operator. Notice that this type of reading is not available in Russian.
2 Interestingly, the ambiguity disappears if in the stimulus question there is a modal verb and the negation appears before the infinitival (Letuchiy 2017) Letuchiy (2017) discusses this example in relation to the non elided version in (ii), where the subject n-word seems to be licensed by the embedded negation (see also Grenoble 1992 andMinor 2013): (ii) Nikto mog ob etom ne uznat'. nobody could about this not know 'It could be the case that nobody knew it' I will come back to similar cases in section 4.

Jacopo Garzonio Negative Concord in Russian. An Overview
Studi e ricerche 20 179 Studi di linguistica slava, [175][176][177][178][179][180][181][182][183][184][185][186][187][188][189][190] topic has scope over the n-word in focus position, there is no sentential negation inside the elided clause, and this configuration triggers the double negation reading (6b). A necessary ingredient of this analysis is the assumption that the negation in the silent topic is too embedded to enter in a Negative Concord relation with the n-word. In the case of the Negative Concord reading, there is no silent topic or it does not contain a negative operator, and the elided part (here As-pectP) contains the sentential negation (6a): The advantage of a similar solution is that it does not require to postulate that n-words are semantically ambiguous between negative quantifiers and NPIs. Notice also that this account does not require a biclausal structure, which can normally contain a double negation configuration: Freestanding n-words Fitzgibbons (2008) has described two other environments where nwords can appear without the preverbal negative marker in Russian. The first type of construction involves the presence of a small clause with a copula or a verb like sčitat' 'to consider': 8 a. Kto byl ničem, tot stanet vsem who was nothing that-person will-become everything 'Those who were nothing will become everything.' b. Ja sčitaju tvoego brata nikem. I consider your brother nobody 'I consider your brother a nobody.' (Fitzgibbons 2008, 53) The second type involves n-words inside PPs, like in (9): 9 Ty isčez v nikuda. you disappeared into nowhere 'You disappeared into nowhere.' (Fitzgibbons 2008, 53) An interesting observation made by Fitzgibbons is that these constructions can have also a double negation reading if there is another negative item, like the sentential negation. She provides the following example: 10 Vanja ne sčital Iru nikem Vanja not considered Ira nobody dn: 'Vanja did not consider Ira a nobody.' (he considered her a worthy person) nc: 'Vanja did not consider Ira anybody.' (i.e. had no opinion of her) (Fitzgibbons 2008, 55) The analysis proposed by Fitzgibbons (2008) for these cases is similar to the analyses of elided structures I have discussed in the previous section, as she proposes a phonologically null negative head ∅ NEG present in small clauses and PPs, different from the sentential negation in TenseP or AspectP and able to enter in double negation configurations with n-words. The locus of ∅ NEG is a Polarity projection optionally present above the PredicateP of small clauses and PPs. Under this view the ambiguity of (10) corresponds to the following two different structures. In (11a) there is only the higher Polarity projection containing the whole clause, while in (11b) there is also the lower one, which contains only the small clause. Notice that in this approach the licensing of embedded n-words is computed assuming their covert movement to either the projection containing ne or the one containing ∅ NEG . This movement is based on the Agreement theory proposed by Bošković (2007), according to which the uninterpretable feature is carried by the moving element. What is relevant for the general discussion about Negative Concord is that under this view small clauses can be of two types. In the first type there is no internal polarity and n-words are computed together with the matrix negation, while in the second type there is an internal polarity which can license n-words (and produces a double negation reading when combined with a matrix negation). According to Fitzgibbons (2008) the same is true also for PPs, since the following example is ambiguous: (12) Dokladčik ne obraščaetsja ni k komu. speaker not addresses neg to who.dat. nc: 'The speaker does not address anybody' (He just likes to listen to his own voice) dn: 'The speaker does not direct his talk to nobody' (The thing is, he is almost blind and is not sure where the audience is) (Fitzgibbons 2008, 57) A consequence of this analysis is to admit that PPs can have an independent polarity projection and are similar to clauses in this respect. The presence of a separate projection for the encoding of polarity above PPs could be related to the splitting, triggered by the preposition, of the negative morpheme ni-and the wh component of the negative indefinite (Harves 1998

The Boundaries of Negative Concord
In this section I present some constructions where an n-word appears to be licensed by a negation lower than Tense/Aspect, i.e. lower than the main finite verb. Letuchiy (2017) has labeled these constructions non-standard Negative Concord cases. There are three main types: adjectival constructions, constructions with moč 'can' and other subject control verbs, and constructions with object control verbs. The two latter types are called inter-clausal Negative Concord constructions by Kholodilova (2015). Adjectival constructions have a semi-copular verb as matrix predicate (e.g. okazat'sja 'turn out to be') and an adjective in the short form: (14) Nikto iz nas okazalsja ne nužen. nobody from us turned out not necessary 'Nobody of us turned out to be necessary.' According to Letuchiy (2017) these are cases of raising, i.e. the nword appears in the matrix portion but is interpreted in the small clause, where the Negative Concord configuration is licensed. In oth- Since adjectival constructions involve copular predicates, I propose to analyze them according to Fitzgibbons' (2008) idea about polarity in the structure of small clauses, assuming that in similar cases the negative marker ne preceding the adjective overtly realizes the small clause polarity. In (16) I represent the structure of (14) with movement of nikto to the matrix subject position. Letuchiy (2017) also points out that an example like (17) is ambiguous. This means that the elided part of the fragment answer can correspond to a full negated Tense/AspectP or just to the small clause: (17) Kto okazalsja ne gotov? Nikto. who turned.out not ready nobody (i) 'Nobody was not ready.' (from Nikto ne okazalsja ne gotov.) (ii) 'Nobody was ready.' (from Nikto okazalsja ne gotov.) The two other types of non-standard Negative Concord involve a matrix control verb and an embedded non-finite verb form. They can be classified according to the control type of the matrix verb. In (18) I provide two examples with a subject control verb (starat'sja 'to try') and on object control verb (prosit' 'to ask'): These cases are problematic since in general n-words require clausemate negation, and a negation in an embedded finite clause cannot license an n-word in the matrix clause: There are different possible analyses to account for these constructions (even if it should be pointed out that there is variation regarding their acceptability by speakers). The first possibility is to assume n-word raising as in adjectival constructions. However, Neg-raising is not normally blocked with embedded finite verbs, as the contextual equivalence of the following examples shows:  Kholodilova (2015) also points out that the matrix verbs allowing this type of non-standard Negative Concord are different from those found in typical Neg-raising cases (for instance obeščat' 'to promise' displays non-standard Negative Concord but not Neg-raising).
A second possibility is to assume that these constructions are in fact monoclausal and the matrix verb is a type of semi-auxiliary verb. According to this view these cases are instances of syntactic restructuring into a single clause or grammaticalization of the matrix verb. Obviously, this analysis applies only to the subject control type. Some evidence in favor of this approach is provided by the fact that the acceptability of the clause significantly degrades if the matrix verb is associated with an intentional energy consuming activity or if there is another overt argument of the matrix verb. Kholodilova (2015) proposes a third possibility, namely that the matrix verb undergoes pragmatic bleaching (Partee et al. 2011). In other words, the core semantics of the verb does not change, but it forms a latu sensu modal frame, which does not influence the proposition. In this account, Negative Concord is computed only semantically and the morphosyntactic component is not relevant. However, as pointed out by Letuchiy (2017), a purely semantic approach cannot explain why in elliptical contexts only the interpretation corresponding to the standard Negative Concord configuration is possible.  In (21) the answer nikto can be interpreted only as corresponding to nikto ne staralsja etogo ne delat', with the n-word licensed by the negation on the elided matrix verb (or a covert negative operator as discussed in section 2) and triggering a double negation interpretation with the elided negated infinitive.
For this reason I propose here a syntactic account based on a strong version of the theory of restructuring, namely the cartographic account developed by Cinque (2006). Under this approach, the syntactic transparency of some structures involving a matrix verb and an infinitival form (like for instance the climbing of object clitics in Romance) is observed precisely because the whole structure is monoclausal and the matrix verb realizes one of the functional heads of the articulated and hierarchical clause structure. Intuitively, the matrix verb is functionally equivalent to an adverb modifying the embedded lexical verb. For instance the verb starat'sja 'to try' in cases like (18a) encodes Conative aspect (Cinque 2006, 47n4): In (22) I assume for simplicity that the constituent [ob etom ne vspominat'] corresponds to the vP. A similar analysis requires to postulate that ne can surface in the lexical layer of the clause structure (an assumption in line with the idea that also small clauses have an internal polarity). This is not strange since a sentence with a modal verb can have a 'constituent negation' ne under a sentential ne, with a double negation interpretation: A potential problem is the fact that the inter-clausal Negative Concord is possible also with some object control verbs (like prosit' 'to ask'). However, Cinque (2006, 24-5) discusses some cases of apparent object control verbs allowing clitic climbing in Romance. The solution he proposes is based on the idea (first discussed by Kayne 1989)  In general, all the constructions where the licensing negation surfaces with an embedded adjective or infinitive can be considered as monoclausal and the generalization that Negative Concord is clausebound is not contradicted. Interestingly, what all these phenomena show is that the item licensing Negative Concord words does not have to be the sentential negation marker on the finite verb in Ten-seP/AspectP. All the phenomena discussed so far can be captured by the two following generalizations: N-words are licensed by a negated verb form or small clause

On the History of Negative Concord in Russian
In this section I provide a brief description of the diachronic development of Negative Concord in Russian. The negative cycle in Russian has beed described and analyzed by Tsurska (2009) and Willis (2013). All Slavic languages display a negation derived from the inherited preverbal negative marker *ne. This item usually precedes the inflected verb, but in some languages (e.g. Slovak) it appears before the participle and not before the auxiliary in complex past forms. Interestingly, Old East Slavonic had the same pattern:  Willis 2013, 346) This suggests that ne is a verb prefix, merged in the low portion of the clause structure (AspP or even vP).
Strict Negative Concord is clearly an innovation in Slavonic. Tsurska (2009) describes the following stages for Russian: a) In Early Russian of the 11th and 12th centuries n-words appearing in the preverbal space could encode negation without the preverbal marker ne, as in (27a) Willis (2013, 370) points out that this optionality is a feature of Old Slavonic in general, and in Old Church Slavonic it is possible to find the two opposing patterns in different gospel translations: b) In the 15th-17th centuries the presence of ne with preverbal nwords becomes the preferred variant even if it is still possible to find cases of non-strict Negative Concord like the following one: (29) Nikto že bez truda venčan budet. nobody prt without work wed will-be 'Nobody will be wed without labor' (Domostroj, 16th c., from Tsurska 2009, 81) c) Finally, from the 17 th century, the Negative Concord is strict, like in the present day system: (30) I ja is požaru ničevo ne pospel vynest' and I from fire nothing not managed take 'And I didn't have time to take anything out of the fire' (An Appeal of V. Krečatnikov, 17th c., from Tsurska 2009, 81) Only some Northern Russian dialects still display the possibility to have preverbal n-words without ne before the verb.
The diachronic development of Negative Concord in Russian strongly supports the idea that the change has involved the internal semantics of n-words. As I have mentioned in section 3, n-words