Connective negation and negative concord in Balto-Slavic.

With negative indefinite pronouns the Balto-Slavic languages all exhibit strict negative concord. In this study we investigate how negative concord functions in a context in which a connective negator (‘neither ... nor’) combines either phrases or clauses. We show that there are various types of non-concordant patterns.


Introduction
The citation from Alexander Sergeyevich Pushkin could apply to Axel, sitting at his desk and working on Baltic linguistics, possibly even clause-linkage (cf. Holvoet & Judžentis eds. 2003), but the relation of clause-linkage to negation probably hasn't landed on that desk yet. This connection is a complex matter and in this paper we focus on one issue, viz. the relation between connective negation, as in English neither … nor, and negative concord, as in substandard English you ain't seen nothing yet. We study this issue in Balto-Slavic, and with a few examples from other languages we sketch a wider typological horizonvery much so in the spirit of Axel's work.
In sections 2 and 3 we sketch our understanding of connective negation and negative concord. In section 4 we focus on the relation of negative concord and connective negation in Balto-Slavic languages. Section 5 is the conclusion.
The study is restricted in more than one sense. First, though the paper touches on diachrony, the focus is on synchrony. Second, though we deal with a fair number of the Balto-Slavic languages, viz. Latvian, Lithuanian, Russian, Belarusian/Ukrainian, Pannonian Rusyn, Kashubian/Polish, Czech/Slovak, Upper Sorbian, Lower Sorbian, Slovenian, Croatian/Serbian, Bulgarian, and Macedonian, we only deal with the standard languages. 1 Third, our observations are based on grammatical descriptions, which usually pay little attention to connective negation, and on the intuitions of native speaker colleagues, but often just one for every language. This paper thus constitutes a plea for more work, particularly, corpus work. Because of the wider cross-linguistic orientation we introduce connective negation with some brief observations about non-Balto-Slavic, viz. Spanish and French. The latter two languages, as well as Croatian, are treated in more detail in van der Auwera (Forthc.). Most examples come from native speaker colleagues.
(1) Spanish No somos ni de izquierda ni de derecha. neg be.prs.1pl coneg of left coneg of right 'We are neither from the left nor from the right.'

47
We will call ni a 'connective negator', henceforth 'CoNeg'. That ni, neither and nor are negative is obvious. The choice for the term 'connective' is less obvious. We prefer 'connective' to 'conjunctive', in a wish to avoid two debates: (i) whether neither and nor are fundamentally more like the conjunction and than like too and also, which are adverbs, and (ii) whether neither and nor are more like the conjunction and than like the disjunction or (cf. Nau & Ostrowski 2010 for another plea for a broad concept of 'connective').
CoNegs vary within and across languages along many parameters. One that we will not study is what happens when there are more than two negatively connected constituents, as when (1) is continued with ni del centro 'nor from the centre'. A parameter that we will focus on is the nature of the constituents that are connected: are they phrases or clauses? The difference is illustrated with French (2) and (3).
(2 There is more to be said about the distinction between phrasal and clausal uses. What happens, for example, when a clause functions within another clause and whose external syntax is thus that of a phrase? Also, not all phrases pattern alike. In this paper we pay attention to phrasal negators connecting finite verbs -we call it the 'finite' type. This is particularly interesting for languages that express standard negation with a preverbal particle, like French and Span-ish, but no less also the Balto-Slavic languages. If the language expresses finite connective negation with a preverbal particle as well, the preverbal slot will have two contenders, a standard negator and a CoNeg. Will a language settle for just one negator or will it host two? Spanish allows only one negator there, viz. the CoNeg. (4) Spanish Ni puedo ni debo exponerla a coneg can.prs.1.sg coneg must.prs.1.sg expose.inf-obl.3.sg.f to ciertos riesgos. certain risks 'I neither can nor should expose her to certain risks.' The French cognate ni, however, does combine with the standard negative particle ne, at least, for the second finite verb, giving ni ne.

(5) French
Je ne veux, ni ne peux répondre. 1.sg neg want.pres.1.sg coneg neg can.prs.1.sg answer.inf 'I don't want to answer and I can't either.' The phrasal, finite and clausal CoNegs may be different or they may be the same, but have different syntactic properties. French (2) and (3) illustrate that the markers may be different. The contrast between Spanish (1) and (4) illustrates that ni may show up in both the phrasal and the finite use, but that its properties differ. In (1) ni co-occurs with the standard negator no, but not in (4). In Balto-Slavic, the CoNegs tend to be the same in the three uses, cf. Russian (6). (6) also illustrates that markers may have different syntactic properties in the different uses. In both the phrasal and the clausal uses Russian ni combines with a negative verb, but not in the finite use.
In Balto-Slavic the typical CoNeg is ni, as in (6), or, more commonly, a complex form containing ni, as in (7)b. The ni forms derive from *nei, which may itself derive from *ne + i 'not and' (lemma ni in [ESSJ]). 3 In the complex forms ni is preceded by an originally connective element a-or followed by a strengthening particle, like -ti (Kovačević 2016, 262) or it has undergone both processes, like older Czech/Slovak aniž, which survives with the meanings 'without' and 'although not'. An interesting double complex form is Lower Sorbian daniž, with the expected strengthener but with da 'so' in front. Bare ni is older than the complex forms (lemmas in [ESSJ] and [SP]). The preposed forms are found in East and West Slavic, the postposed forms in Baltic and South Slavic -and the circumposed ones in West Slavic. 4 As already illustrated in Croatian/Serbian, some languages have or had both simple and complex forms. The new form may oust the old one (Czech/Slovak ani, ni) or leave it only in relic phrases (Polish ani, ni) or in an archaic register (Slovenian ni -lemma ni in [SSKJ]), the two may be alternatives with roughly identical uses (Bulgarian ni, nito) or different ones (Croatian/Serbian ni, niti), or the new form may disappear again (Lithuanian nei, neigi). A CoNeg may also appear in a slot that either had no marker at all or only had a standard negator. For instance, in Old Lithuanian (Ostrowski 2014, 125-131) phrasal nei ... nei was very rare, the established pattern having nei only for the second finite verb and the standard negator ne in front of the first one. The picture is further complicated by the fact that there are a few CoNegs that do not seem to relate to ni -Latvian and Slovenian ne, which have the same form as standard negators, and Lithuanian nė.
The above forms have connective negation as a typical function, but this need not be their only function. Most of them have a scalar use ('not even'), illustrated in Polish (8).

(8) Polish
Nie słyszy ani słowa. neg hear.prs.impf.3.sg scneg word.gen.sg 'He/she doesn't even hear a word.' In all of Balto-Slavic languages ni combines with interrogative-indefinite pronouns to yield negative indefinite pronouns, like in Romance (Gianollo 2018) -and in some languages (e.g. Russian) prepositions can still split the two components. There are several other uses, not found throughout the family. Thus in Lithuanian (but also in old Russian and West Slavic) the ni word serves as a comparative marker and Russian ni has an expletive use.

(9) Lithuanian
Jaučiuosi geriau nei jaunystėje. feel.prs.1.sg.refl better than youth.loc.sg 'I feel better than in my youth.' Most of the CoNeg markers are thus not dedicated. This also applies e.g. to Slovenian ne, if the conclusion is warranted that ne in (11) is really only a CoNeg use of the standard negator.

(11) Slovenian
Ne jaz ne ti ne bova odgovarjala. neg 1.sg.nom neg 2.sg.nom neg be.pfv.fut.2.du answer.ptcp.du.m 'Neither I nor you will answer.' But what is dedicated in (11) is the entire pattern, the construction with two ne's or, in the earlier examples, two ni forms, usually called 'correlative'.
Next to the correlative pattern, there are other dedicated patterns. Thus all the languages have strategies with either the normal conjunction or disjunction and/or a positive additive marker. This point is illustrated for the phrasal domain with Lithuanian (12). Next to the dedicated correlative nei ... nei construction in (12)a there is a construction with the ordinary disjunction ('not ... or ...') - Russian (13)a illustrates 'not and also not' as an alternative to (6)c, and (13)b 'not and not' as an alternative to (6) The strategies with 'not or', 'not and also not' or 'not and not' may be preferred to the 'more' dedicated ones. The CoNeg pattern in Russian (6) b, for instance, though provided with an international grammaticality seal (Wade 1992, 490), cannot count on a high acceptance rate with native speakers (Sannikov 2018, 301). It may have been more common in Pushkin's time (Bulaxovskij 1953, 367), but in present-day Russian it is extremely rare. In the Russian National Corpus with over 300 million words (https://ruscorpora.ru/ old/en/index.html), Vladimir Plungian (p.c.) has found no more than a dozen natural examples, compared to 20,000 examples of the 'not and not' pattern of (13)b. Also, in the clausal use, a mixed strategy may be common, with a standard negator in the first clause and a CoNeg in the second one. This is illustrated with Croatian/Serbian (14). It is an alternative to example (7)c, with a CoNeg only in the second clause. So much for an overview of the properties of connective negation.

Negative concord
The goal of this paper is to study how connective negation interacts with negative concord. So we need to explain what we mean by 'negative concord' ('NC'). The idea behind the notion of 'negative concord' is simple: a single clausal negation is expressed both with a standard negator and one or more constituents. Thus (16)a has two exponents of negation, ain't and nothing, but they do not cancel each other. Semantically, there is only one negation, which is reflected in the Standard English versions in (16)b and (16)c, with either just nothing or n't.
b. You have seen nothing yet. c. You haven't seen anything yet.
Despite the simplicity of this observation, there are many controversies. The most important one concerns the question whether words like nothing in (16)a are truly negative, like nothing in (16)b, or only negative polarity items like anything in (16)c. In this paper we take a constituent like nothing in (16)a to be negative, but anyone who is convinced of the alternative view can rephrase the present study in terms of negative polarity. There is also quite some variation, both within one language and cross-linguistically, as can be appreciated from skimming through Déprez & Espinal (eds.) (2020) or van der Auwera & Van Alsenoy (2016,2018). Most of the studies give pride of place to the NC type illustrated in (16)a, i.e., the collocation of a standard negator like n't and an indefinite pronoun like nothing. Other types have drawn much less scholarly attention. A type that has the same negator but something other than a negative pronoun is shown in (17).
The constituent that combines with the clausal negator is not a negative indefinite pronoun, but a negative noun phrase, and negative noun phrases and pronouns do not behave in the same way (van der Auwera & De Lisser 2019). The remark about the lack of scholarly attention also holds for the type investigated in this paper, i.e., the collocation of a standard negator and a CoNeg construction. This is also a matter of non-pronominal NC and we will see that this type does not behave like a negative pronoun either.
For the collocation of one negative indefinite pronoun and a standard negator the parameter that has received most attention concerns word order. NC is said to be 'strict' if the negative indefinite pronoun doubles up with the standard negator, independently of the position of the negative indefinite relative to the finite verb. This is the case in Lithuanian. NC is said to be 'non-strict', if the double exponence depends on the position of the negative pronoun relative to the finite verb. There are subtypes, and the case that has been studied most has NC with a negative indefinite pronoun following the finite verb and no NC with a negative indefinite pronoun preceding the finite verb. We see this in Italian (19) In Balto-Slavic, simple clauses with one negative pronoun overwhelmingly 5 show strict NC, the type illustrated with Lithuanian (18), but we will come close to non-strict NC showing up with CoNegs.

Negative concord and connective negation
In this section we look at the relation between negative concord and connective negation. As regards connective negation, we only look at patterns containing at least one ni form, either a simple or a complex one. We discuss phrasal, clausal and finite uses separately.

Phrasal connective negation and negative concord
In the phrasal domain the Balto-Slavic languages show a great deal of uniformity: most of them allow a CoNeg construction that has two CoNegs and that exhibits the strict NC that negative pronouns show. Thus Latvian nedz ... nedz collocates with a negative verb, independently of word order, just like the Lithuanian negative pronoun in (18) One exception is Bulgarian -but not, interestingly enough, Macedonian. Ivanova & Gradinarova (2015, 80) claim that when the phrasal CoNeg precedes the finite verb, the latter cannot be negative, and when the phrasal CoNeg follows the finite verb, this verb has to be negative.  If this is correct, we are dealing with an almost classically non-strict NC of the type found in Italian (19): it is not fully classical, because it only applies to phrasal CoNegs -and in a language that has strict NC for negative indefinites.
But the claim is not fully correct. We understand (Zlatka Guentchéva, Olga Mladenova, p.c.) that in some preverbal contexts, like the one in (21)a, the verb can be either positive or negative. Thus the near-classical non-strict NC is not the Italian type, but the Catalan type. i.e., the type in which the preverbal negative indefinite can be followed by either a positive or a negative verb (van der Auwera & Van Alsenoy 2016).
And there is more non-classical variation. CoNeg constructions may exhibit a kind of non-strict NC, called 'quirky' in van der Auwera (Forthc.) -'quirky', because it differs from the 'classical' type illustrated with Italian (19) -and also the near-classical type of Bulgarian (21). In the classical and near-classical types the position of the negative phrase relative to the finite verb determines whether or not the clause needs a standard negator. In the quirky type, the position of the negative phrase relative to the finite verb determines the absence of the first of two CoNegs. Thus in Spanish (1) the first ni is optional, but only when ni ...ni follows the standard negator. We have this in Lithuanian, Bulgarian and Croatian/Serbian too.

Clausal connective negation and negative concord
In the phrasal use the Balto-Slavic verb is nearly always negative. In some languages we find the same pattern in the clausal uses. Thus the second clauses in Slovenian (24) we get the CoNeg niti and the verb is negative. This almost invites the claim that Slovenian clausal niti shows the same kind of NC as phrasal CoNegs, but not quite. Phrasal CoNegs can appear both before and after the verb. Clausal CoNegs, however, can only appear before the verbs of their clauses. So we are not dealing with classical strict NC, but, once again, with a 'quirky' type of NC. Also, if, in these languages, the clause contains an indefinite pronoun, then it is negative, as shown in Russian (25) This must mean that these CoNegs express the clausal negation: after all, they scope over entire clauses and there is no need for a second clausal negator, i.e., there is no need for the standard negator. 6 Interestingly, if the clausal CoNeg takes the place of the standard negator, one would expect them to trigger NC in indefinites. But this is not the case. This is illustrated with Croatian/Serbian (27), with an 'anybody' or a 'somebody' pronoun, and with Bulgarian (28), with a 'somebody' pronoun. What nearly all languages rule out is the pattern with a negative verb and a positive indefinite, which makes sense, given the strict NC profile for negative indefinites. But this would also make one expect both the verb and the indefinite to be negative, but this is not the case in Baltic as well as in 'non-Slovenian' South Slavic. In these languages, the indefinites are negative only if the verbs are negative; the presence of a clausal CoNeg is not sufficient. In clausal CoNeg patterns the verbs are not negative, because the CoNeg takes care of the clausal negation. The CoNeg thus succeeds in expressing clausal negation instead of the standard negator, but it does not succeed in licensing the negative pronoun.

Finite connective negation and negative concord
We have suggested with Spanish (4) and French (5) that finite connective negation may be special. This is true for some Balto-Slavic languages too. We have already shown this with the CoNeg pattern in Russian (6)b, as compared to (6)a and (6)c: only in the finite CoNeg pattern of (6)b are the verbs not made negative by the ne particle. Polish is special too, in allowing both the pattern with and without the standard negator. And there is another difference: whereas in Russian the pattern without a clausal negator is extremely rare to the point of sounding artificial, the Polish counterpart is colloquial, to the point of being substandard. Belarusian/Ukrainian is special too: both versions are allowed.
(30) Ukrainian Vin (a)ni (ne) tancjuje (a)ni (ne) spivaje. 3.sg coneg neg dance.prs.ipfv.3.sg coneg neg sing.prs.ipfv.3.sg 'He is neither dancing nor singing.' But there is a semantic difference: the pattern without the clausal negators conveys that when the subject is neither dancing nor singing, he is doing something in between (Ljudmila Popović and Maria Koniushkevich, p.c.). Such subtleties may, of course, be under our radar for the other languages, but it seems that, overall, the negativity of the verbs with finite CoNegs is the same as with clausal CoNegs.
We find a similar situation with respect to the negativity of the indefinite pronouns: they are negative or non-negative in finite CoNegs for the same languages as in clausal CoNegs and, again, Russian, Polish and Belarusian/ Ukrainian are special. When the latter allow a positive pronoun, one would expect, judging from Latvian, Lithuanian, Croatian/Serbian, Bulgarian and Macedonian, that the pronoun will be positive or negatively polar. However, indefinite pronouns are simply unacceptable, not just negative ones, but also positive and negatively polar ones. 7 We mark this in Table 3    The Balto-Slavic languages are textbook cases of strict NC. This paper has shown that connective negation disturbs this simple picture: there are various types of non-concordant patterns. When there is a non-concordant pattern in a NC language, we are used to the pattern in which the verb is not negative, although the accompanying indefinite still is. When there is furthermore an alternation between NC with a postverbal indefinite and no NC with preverbal indefinite we speak about 'non-strict NC'. We have something close to this with Bulgarian phrasal CoNegs. What we also find with phrasal CoNegs is a non-concordant pattern in which the lack of concord does not concern the negativity of the verb, but that of the first of two CoNeg phrases. Interestingly, what determines the variation may again be word order and more particularly, again, whether the relevant constituent is preverbal or postverbal. This is the case in Croatian/Serbian, for example, but in e.g. Czech/Slovak we see that a first phrasal CoNeg may be absent independently of word order.
Clausal CoNegs show a different non-concordant pattern. In e.g. Slovenian the clausal CoNeg combines with a negative verb, as one would expect, but in e.g. Lithuanian the verb is positive. This looks like what we find in classical non-strict NC. There a preverbal negative indefinite goes with a positive verb and clausal CoNegs are necessarily preverbal. However, there is a difference, in classical non-strict NC the negative indefinite can be postverbal and they then show NC. Clausal CoNegs, however, cannot be postverbal. So their non-strict NC is not really the classical type. And there is more, when in a clausal CoNeg construction the verb is non-negative and there is a focal indefinite pronoun, it is not negative either: depending on the language, it may be negatively polar or positive.
The most quirky type is the finite type. It resembles both the clausal and the phrasal type, but it is different from both. Like a clausal CoNeg a finite CoNeg may combine with a positive rather than a negative verb, but there are differences. Most importantly, whereas with clausal CoNegs no language allows a choice between a positive and a negative verb, with a finite CoNeg this is possible. As to the similarity of finite CoNegs to phrasal CoNegs, we see again that of two CoNegs the first one may absent. In the case of phrasal CoNegs this alternation may depend on whether the CoNegs are preverbal or postverbal.
In the case of finite CoNegs, however, there is no word order dependence, because the finite CoNegs, like the clausal ones, are always preverbal.
As a sobering note, this study barely scratches the surface. Issues not dealt with include the following: 1 / We have studied how a first CoNeg alternates with Ø in the case of phrasal and finite CoNegs. But there is more to be said about the first CoNeg slot, especially in the case of clausal and finite CoNegs, which can contain standard negators instead of CoNegs. 2 / To explain the various (ir)regularities the paper referred to the 'Neg Early' constraint and to a competition of two kinds of negators for the slot immediately in front of the finite verb. More work is needed here. 3 / We have seen with Russian (6)b that a construction can be extremely rare: we need solid corpus work to document the frequency of the various patterns and of the extent that they are typical of regions or registers. 4 / We need corpus work to document the diachronies, including that of contact interference. and 5 / We need work on both the synchrony and the diachrony of the non-CoNeg uses of the markers studied here.