When ni- and -nibud’ are logically equivalent: Evidence from Russian nominalizations

This paper deals with the two sets of polarity sensitive items in Russian: niand -nibud’ pronouns. Non-specific indefinite -nibud’ pronouns (NSIs) are possible only in propositions that do not ensure truth, i.e. non-veridical contexts. Although clause-mate negation creates such a context, NSIs are incompatible with it and are substituted by negative nipronouns that are licensed only by negative concord. The incompatibility of NSIs with negation can be resolved in subjunctive sentences and embedded purpose čtoby-clauses, however, the licensing conditions in these cases are not defined. In this paper I introduce another context which licenses both types of pronouns, namely, negated process nominalizations. I determine the licensing conditions for the two types of pronouns in nominalization, and test previous approaches against the new data. In particular, I argue that -nibud’ pronouns are licensed in the scope of the nonveridical operator that is introduced in the main clause.


ni-and -nibud' pronouns in Russian
Polarity sensitive items (PSIs) are elements that are restricted to a set of contexts that have certain truth-conditional properties. In Russian there are traditionally distinguished four main classes of PSIs: negative nipronouns, non-specific indefinite -nibud' pronouns, negatively polarized -libo pronouns and free-choice NPIs lyuboi and ugodno [Paducheva, 1985;Haspelmath, 1997].
Two series of polarity sensitive items in Russian are in complementary distribution, namely, ni-and -nibud' pronouns. Negative ni-pronouns belong to strict negative polarity items [Giannakidou, 2011], or n-words [Laka, 1990], as they are licensed only under negative concord, can provide a negative fragment answer and block double-negation readings. According to [Paducheva, 2014], ni-pronouns are licensed in the context of clausemate sentential negation and banned from the scope of constituent negation. The licensing of ni-pronouns in the scope of superordinate negation is subject to structural restrictions (contra [Pereltsvaig, 2004]): as shown by [Gerasimova, 2015], negative pronouns can be licensed in infinitival clauses not bigger than TP.
Non-specific indefinite -nibud' pronouns (NSIs) are licensed only in nonveridical context, which is introduced by operators that do not ensure truth [Paducheva, 1985[Paducheva, , 2014Giannakidou, 2011]. Importantly, clause-mate negation creates an anti-morphic context that belongs to non-veridical contexts. Therefore, one would expect that NSIs are licensed by negation. However, -nibud' pronouns are incompatible with negative concord and are obligatory substituted by negative ni-pronouns (cf. (1)- (2)). This property of Russian NSIs is referred to as the Bagel Problem [Pereltsvaig, 2004]: the anti-morphic context figuratively speaking creates "a bagel hole" with respect to NSIs as they are not licensed in it (see Fig. 1 Paducheva supposes that the substitutability of the two pronouns is not absolute: the two NPIs create logically equivalent, but not synonymous sentences. This lack of synonymity can be seen in (5): while in (5a) a nipronoun is used entailing that no one from the known set of people was hurt, in (5b) NSI refers to a particular person, though randomly chosen. To model the licensing conditions of NSIs in negative concord, Paducheva introduces the notion of non-standard negation. This is negation which appears in the scope of non-veridical operator introduced by conjunction čtoby and by subjunctive mood. NSIs can appear under the scope of nonstandard negation: that is, they are not banned as in case of (standard) negative concord. However, Paducheva states that NSIs are still licensed by the nonveridical operator.
The idea of non-standard negation is quite problematic as this kind of negation is postulated ad hoc only for two contexts and it is not clear how exactly negation receives its specific properties. In this paper I introduce another context which licenses both types of pronouns, namely, negated process nominalizations. The research question is the following: What licensing conditions help to resolve the Bagel problem? To answer this question I determine the licensing conditions for the two types of pronouns in nominalization, and test Paducheva's approach against the new data. In particular, I argue that -nibud' pronouns are licensed in the scope of the nonveridical operator that is introduced in the main clause. The rest of the paper is organized as follows. In section 2 I present a brief overview of negated process nominalizations in Russian. In Section 3 I analyze the licensing conditions for ni-and -nibud' pronouns taking into account how the presence of negation, non-veridicality and specificity of a nominalization can influence acceptability of NSIs. In Section 4 I present the consequences that my analysis has with respect to one more context in which NSIs are licensed. Section 5 concludes.
(6) Yavlyaetsya li nepodderzhanie is whether neg-supporting blagotvoritel'noi initsiativy grekhom charity initiative sin 'Whether not supporting a charity initiative is a sin' The corpus study revealed that negated process nominals possess the same structural properties as affirmative event nominalizations: in particular, they obligatorily take internal arguments (6), and may take aspectual modifiers (7). neg-going silent 'the everyday not going silent' According to Pazel'skaya, negation in nominalizations creates the same context as clausal negation. Therefore, we would expect to observe the Bagel Problem: the licensing of ni-pronouns and unacceptability of NSIs in negated nominalizations. However, the GICR study shows that both types of pronouns are available within negated event nominalizations (8) I argue that negated process nominalizations constitute the same context as in cases observed by Paducheva. The structural position of negation was examined in [Gerasimova, 2019]. The diagnostics show that negation appears high in the syntactic structure, at least above all arguments and possibly even above the nominalizer. In particular, a pilot acceptability study has shown that n-words in nominalization can be licensed distantly from the matrix clause. Consequently, the nominalizer does not serve as a barrier for the strict NPI licensing. This means that n-words in negated nominalizations can be licensed by negation that is located above the nominalizer. Another proposal made by [Gerasimova, 2019] is that all arguments of a nominalization are generated before the [neg]-feature is introduced and fall under the negative scope. That is, in nominalizations we observe instances of clausal negation which has properties similar to those of negation in subjunctive sentences and embedded purpose čtoby-clauses.

Licensing conditions for ni-and -nibud'
So far we have seen that negated process nominalizations provide context which licenses both negative ni-pronouns and non-specific indefinite -nibud' pronouns. While for ni-pronouns the hypothesis is that negation is the licensor, it is less clear what operator could provide the nonveridical scope in nominalization that would license -nibud' pronouns. One possible solution could be that the nominalizer itself contains the covert non-veridical operator, as according to [Weinreich, 1963], nominalization is an assertion suspending device. However, in this case the nominalization would always provide the non-specific interpretation for its arguments, which is not the case.
I propose that nominalizer keeps the stem neutral with respect to quantificational operators. In order to establish the licensing conditions, I suggest systematically examining in which contexts which type of pronouns is licensed taking into account the three parameters. The first parameter is the presence/absence of negation in nominalization, which determines the complementary distribution of the two types of pronouns. The second parameter is the presence/absence of the non-veridical operator. On the one hand, nonveridicality can be introduced in the main clause with sentential aspectual operators such as habitual, generic and iterative: e.g. vsegda 'always' in (10a). In this case aspectual operators allow for the nonspecific interpretation of the nominalization and therefore NSIs may be licensed. On the other hand, nonveridicality can be introduced within The final parameter is the specificity of the nominalization. This parameter is the crucial one as specificity of a noun phrase restricts the usage on nonspecific indefinite pronouns (11a). I suppose that specificity is introduced at DP which serves as the referential semantics domain. If the non-veridical operator is within a specific DP, it can license -nibud' pronouns (11b). However, when the noun phrase is specific, the clausal non-veridical operator cannot license NSIs (11c). Therefore, specificity serves as a restrictor for NSI licensing by a clausemate non-veridical scope and can be used as a diagnostic of whether there is non-veridical scope created in nominalization. The different combinations of the mentioned parameters allow us to establish the licensing conditions for ni-and -nibud' pronouns and define whether there are any interactions between different scopes. Below we provide observations that are based on judgments from 15 native speakers (ages 22-55).
First, we shall examine the factor of specificity. When the nominalization is specific, non-veridical operators from the main clause cannot license NSI in nominalization. The licensing of NSIs in specific nominalizations does not depend on whether nominalization is negated or not and whether the context is affirmative or non-veridical.

c. Specific nominalization in the non-veridical main clause
Ego podrazhanie *nikakomu / *kakomu-nibud' masteru his copying after no(n-word) / any(nsi) master vsegda menya udivlyalo always me impressed 'His copying after some master always impressed me' d. Specific nominalization in the veridical main clause Ego podrazhanie *nikakomu / *kakomu-nibud' masteru His copying after no(n-word) / any(nsi) master udivilo nas impressed us 'His copying after some master impressed us' On the contrary, when the nominalization is non-specific, NSIs can be licensed by non-veridical operators from the main clause (13). The licensing conditions of NSI do not depend on whether nominalization is negated or not. However, when the clause is affirmative -nibud' pronouns are no longer licensed: even though there is no restriction on licensing in the form of specificity, there is no non-veridical operator in the main clause either (14)-(15). Лингвистика ISSN 2500Лингвистика ISSN -2953 (13) a. Non-specific negated nominalization in the non-veridical main clause Eto motiviruet menya na ne-napisanie this motivates me to neg-writing ok nikakoi / ok kakoi-nibud' eresi no(n-word) / any(nsi) nonsense 'This motives me for not writing any nonsense' b. Non-specific nominalization in the non-veridical main clause: Podrazhanie *nikakomu / ok kakomu-nibud' masteru copying after no(n-word) / any(nsi) master vsegda ubivaet individual'nost' always kills individuality 'Copying after some master always kills individuality' (14) Non-specific negated nominalization in the veridical main clause: Direktor odobril ne-vmeshatel'stvo principal approved neg -intervening ok ni v kakie / * v kakie-nibud' dela in no(n-word) / in any(nsi) business 'The principal approved not intervening in any business' (15) Non-specific nominalization in the veridical main clause: Direktor odobril podrazhanie principal approved copying after *nikakomu / *kakomu-nibud' masteru no(n-word) / any(nsi) master 'The principal approved copying after some master' In case there is a non-veridical operator in the nominalization, the observation is, as expected, that this operator would be the licensor of NSIs (16).
(16) Specific (negated) nominalization in the non-veridical main clause: Ego postoyannoe (ne)podrazhanie ok kakomu-nibud' masteru his constant (neg)copying after any(nsi) master sdelalo ego izvestnym made him famous 'His copying after some master made him famous' To sum up, we observe the following distribution of ni-and -nibud' pronouns. NSIs are not licensed in specific nominalization and in nonspecific nominalization in affirmative clause. However, -nibud' pronouns can be licensed in non-specific nominalization in non-veridical clause and by non-veridical operator within specific nominalization. These observations are summarized in Table 1. The provided examples also show that nipronouns are licensed only in negated nominalizations. The data above allows us to conclude on what is the relative order of the operators at LF. The crucial observation is that the licensing of -nibud' pronouns does not depend on the presence/absence of negation. Importantly, the specificity of nominalization influences the availability of non-veridical operator from the main clause, which can license -nibud' pronouns.
I argue that there is no need in postulating the non-standard negation, as negation is not necessary for NSI-licensing. Although linearly -nibud' pronoun appears under the scope of two operators, it is not necessarily in the negative scope at LF. In other words, NSI undergoes LF movement, gets out of the scope of negative operator but is still in the scope of non-veridical operator. Herewith, nonveridical operator is located above the negative one. The same course of reasoning can be applied to the cases of subjunctive sentences and embedded purpose čtoby clauses.

Consequences of eliminating the non-standard negation
The elimination of the non-standard negation can help in solving another puzzle connected to -nibud' licensing conditions. In particular [Paducheva, 2018] reports some marginal examples that contradict the generalization that negation cannot license -nibud' pronouns. In particular, in (17a) -nibud' pronoun is licensed under the negative scope of nepravda 'lie', while it is prohibited in an affirmative variant of the same sentence (17b). Thus, it can be argued that derivational negation in nepravda 'lie' is the operator that is licensing the NSI.

Conclusion
To summarize, in this paper I have examined the licensing conditions for the two polarity sensitive items in Russian, ni-and -nibud' pronouns. While these types of pronouns are usually in complementary distribution, two contexts were found in which both of them are available. In order to model this exceptional licensing of NSIs under the negative scope, [Paducheva, Rhema. Рема. 2020. № 1 2018] introduced the notion of non-standard negation, which appears under the scope of non-veridical operator. The limitation of such an approach is that this is an ad hoc solution which does not provide any explanation regarding the mechanisms that underlie licensing.
To investigate the exceptional NSI licensing conditions I introduced another context which licenses both types of pronouns, viz. negated process nominalizations. I tested Paducheva's approach against the new data. In particular, by assessing acceptability of sentences with all possible combinations of presence/absence of negative, non-veridical operators and specificity, I have shown that negation does not affect NSI licensing in any way. The crucial observation is that NSIs are always licensed by non-veridical operators in case they are not in a specific DP which restricts the scope of the clausal operator. Remarkably, the obtained results correspond to the crosslinguistic generalization from [Giannakidou, 2006]: n-words obey syntactic locality restrictions and are licensed by a clause-mate antiveridical expression, while non-veridical operators exhibit long distance licensing.