To Be or Not To Be?

Though the occurrence in Russian of genitive of negation (GenNeg) with the verb byt’ ‘to be’ and with intransitive verbs that allow genitive arguments under negation, termed genitive verbs (GVs) (Padučeva 1997), is assumed to be one phenomenon, the two contexts are not consistent with one another. The account we propose explains these inconsistencies, which existing analyses leave as puzzling counterexamples. First, GVs tend to display optionality in assigning GenNeg to inanimate arguments (1); in the case of byt’, however, GenNeg assignment is obligatory (2).

Second, GVs generally disallow GenNeg marking of proper names, allowing only nominative case (3) instead, while with byt’ either nominative or genitive case is possible (4).

Third, GenNeg with GVs is linked to an interpretive effect. In particular, a GenNeg-marked argument of a GV is interpreted as non-referential (see (1a)), while a nominative argument is referential (see (1b)) (Bailyn 1997, Babyonyshev 2003). However, with the verb byt’, GenNeg can occur with either a referential argument, such as the proper name Svety ‘Sveta.GEN’ in (4b), or a nonreferential argument, such as the NP kefira ‘kefir.GEN’ in (2a).

The availability of GenNeg with referential arguments, such as the proper name Sveta in (4b), inhibits an appealing generalization about the direct relationship between GenNeg and existential interpretation of the sentence (Borschev and Partee 2002).¹ The “Existence Is Relative” Principle in Borschev and Partee 2002:188 resolves the problem only partially, as it only accounts for GenNeg on necessarily referential arguments, such as proper names. However, a GenNeg- marked NP that is not lexically specified for such a construal can be interpreted as referential in byt’ constructions. In fact, it must be interpreted as referential in sentences like (5a), where the NP is preverbal.² The nonreferential interpretation is only available for a postverbal NP, as in (5b). By equating the assignment of GenNeg with existential interpretation, Borschev and Partee are unable to account for interpretive differences that arise solely through variations in word order in GenNeg sentences with byt’.

The analysis of byt’ that we believe comes closest to accounting for the above properties is Padučeva’s (1997, 2004) lexical-semantic account. Padučeva posits a division of byt’ occurring with the specification of a location, as in (2), (4), and (5), into two separate verbs with distinct lexical semantics, which she calls locative byt’₁ and locative byt’₂. Byt’₁, illustrated in (2a), (4b), and (5), belongs to the so-called perceptual group of GVs: its semantics includes a perceptual component related to the interpretation of an implicit observer, made salient by the assignment of genitive case to the argument under negation. According to Padučeva (1997), assignment of GenNeg to the argument of a GV is optional and results in the interpretation of the sentence as asserting the absence of the subject from the field of vision of this implicit observer. Conversely, byt’₂ does not belong to the group of GVs, as it lacks the relevant semantic component; accordingly, GenNeg cannot be assigned to its argument (see (4a)).³ It is a separate predicate, whose semantics requires an agentive argument; it is therefore not compatible with the inanimate subject kefir in (2b).

The advantage of Padučeva’s analysis is that it distinguishes different meanings of byt’ that demonstrate different behaviors. Its disadvantage is that it treats byt’ with a genitive-marked argument and a specified location—Padučeva’s locative byt’₁—as an intransitive verb that behaves in the same way as other perceptual GVs, such as prijti ‘to come, arrive’ in (1). In particular, assignment of GenNeg to the arguments of perceptual GVs is optional and interpretively conditioned (see (1)). Yet byt’₁ does not allow this optionality: the sentence in (2b) cannot be interpreted as byt’₁ with a nominative subject (Chvany 1975). Furthermore, byt’₁ does not conform to Padučeva’s assertion that GVs imply an observer as part of their semantics, as evidenced by (6), which is grammatical even though no implicit observer is possible at the specified location.

This constitutes an interpretive basis for proposing that byt’₁ is a separate type of verb whose syntactic and semantic behavior differs from Padučeva’s characterization of GVs.

Following Padučeva (2004), we propose an analysis that distinguishes two independent lexical entries for byt’ with a specified location, one of which (Padučeva’s byt’₂) is predicative/argument-taking—that is, it is a predicate with a meaning similar to ‘to visit’ or ‘to have been to (at some point in time)’. Unlike Padučeva, we analyze her byt’₁ not as a predicate but as the copula in a copular construction with a locational prepositional predicate, with the meaning ‘to be (located) in’.⁴ Both predicates—byt’₂ and the prepositional predicate used with copular byt’₁—are transitive, involving two obligatory arguments: one functioning as the grammatical subject and another denoting a location.⁵

In the byt’₂ construction, as in (7), the verb byt’ functions as the main predicate of the sentence. It takes two arguments: an internal Location argument—usually denoted by a PP—and an external Agent argument. In both affirmative and negated sentences, the subject must bear nominative case, which is the default case assigned to overt subjects in Russian; GenNeg is not possible with this verb. The assignment of nominative case to the subject in the context of sentential negation in (7b) indicates that this is predicative byt’, and the sentence is interpreted as asserting that the Agent has not visited the Location denoted by the PP.

In (8) and (9), the verb byt’ functions as a copula, establishing a predication relation per Williams (1980) between the grammatical subject and a predicate, which in these sentences is the preposition v ‘in’. This prepositional predicate takes two arguments: an internal nominal Location argument and an internal nominal Theme argument, the latter functioning as the grammatical subject.⁶ Affirmative sentences with this prepositional predicate, as in (8a) and (9a), have a meaning similar to ‘to be located at x (at a specific point in time)’, where x is the place denoted by the internal Location argument of the preposition. In the context of sentential negation, as in (8b) and (9b), the assignment of genitive case to the subject of this copular construction is obligatory (Chvany 1975, Bailyn 1997). We argue that this is because morphological case is used to distinguish these two lexical items: GenNeg identifies the copular construction, while nominative case is reserved for the subject of predicative byt’ (see also Matushansky 2012 on the connection between lexical semantics and morphological case).

The affirmative nonexistential copular construction in (8a) is identical to the affirmative predicative construction in (7a). However, the obligatory assignment of genitive case to the grammatical subject in the context of sentential negation in (8b) disambiguates copular byt’.

In the present tense, copular byt’ is disambiguated in a negated sentence by a distinct form of negation (see (10b) and (11b)). As argued in section 3, the form net results from a syntactic operation that merges the copula with the negation, creating a complex Neg head.

In negated sentences with predicative byt’, this form of negation is unavailable (12). By hypothesis, this is because only a copula can undergo the aforementioned syntactic process.⁷

The obligatory nature of GenNeg with copular byt’ under negation indicates that, in this context, the default case assigned to the grammatical subject is in fact the genitive case (Chvany 1975, Bailyn 1997). As the unmarked variant, genitive case is assigned indiscriminately and is not linked to some specific interpretive effect or other, such as referential or existential interpretation. Instead, we propose that word order determines the interpretation of the NP: in the postverbal (VP-internal) position, the GenNeg-marked NP is a second-order predicate that expresses properties of sets—that is, a “generalized quantifier” (Barwise and Cooper 1981, Keenan 1987, Francez 2009). Conversely, movement to Spec,IP removes the quantificational construal of the NP and turns it into a referential expression (cf. McCloskey 2014, Jasinskaja and Šimík to appear), where referentiality is understood in the sense of Geist’s (2008) epistemic specificity: when using a referential expression, the speaker has a particular referent in mind that they believe to exist in the real world (see also von Heusinger 2011, Farkas 2002). The existential reading is only available with the former.⁸

The copular constructions in (8) and (9) have the same numeration but demonstrate two different neutral word orders (contra Brown 1999), that is, word orders that do not require any contextual license and that can be uttered out of the blue and with neutral intonation. In (9), the postverbal position of the grammatical subject results in its interpretation as a generalized quantifier: the speaker does not have a specific referent—a specific doctor—in mind; rather, any member of the set of doctors could make the sentence true. The two sentences are thus interpreted existentially. In (9b), the negation takes wide scope over this generalized quantifier subject, and the sentence is interpreted as asserting that it is not the case that the city contained any member of the set of doctors (at a specific point in time). Compositional semantics preclude the narrow scope reading of the negation with respect to the generalized quantifier. Meanwhile, in (8), NP-movement of the grammatical subject to the preverbal position encodes its referential reading: here, the speaker has a specific doctor in mind. The negated sentence in (8b) is interpreted nonexistentially, asserting that the referent of the subject cannot be found at the location x at a specific point in time; in other words, it asserts the referent’s absence from a certain place at a certain time.

This analysis easily accounts for the data with byt’ presented in section 1. The predicative verb byt’—like most Russian verbs—disallows genitive case assignment to its agentive argument (for reasons to be explained in section 3) and requires the assignment of nominative case in all instances (see (4a) and (7)). Conversely, the genitive-marked subjects of the negated sentences in (2a), (4b), (8b), and (9b) indicate that in these sentences byt’ functions as a copula, with which GenNeg case assignment to the grammatical subject is obligatory, even if this subject is referential, as in (4b) and (8b).⁹ The ungrammaticality of (2b) thus results from the assignment of nominative case—which disallows the copular reading and forces the predicative reading of byt’—to an inanimate subject, which cannot be interpreted as an Agent and is therefore incompatible with the lexical semantics of this predicate.¹⁰

Plausibly, the two distinct lexical interpretations of byt’ have two distinct syntactic realizations. Thus, predicative byt’, shown in (13), functions as a standard agentive transitive verb, with an external Agent argument and an internal Location argument. Following Kratzer (1995), we assume that VP-internal positions are reserved for internal arguments: the verb’s external Agent argument does not originate predicate-internally in Spec,VP. It may be assumed that this argument is merged in Spec,vP or directly in Spec,IP (Neeleman and Van de Koot 2002, 2010), as in (13). To express its internal Location argument, the verb byt’ selects for a PP, v gorode ‘in the city’, which is merged VP-internally in the verbal complement position;¹¹ and it selects for an NP, doktor ‘doctor.NOM.MASC’, which is merged directly in Spec,IP and assigned the Agent thematic interpretation. This NP receives structural nominative case via agreement with the finite Infl.

In the case of the copular construction, illustrated in (14), the preposition functions as the main predicate of the sentence,¹² selecting for two internal arguments: a Theme NP, which acts as the grammatical subject of the sentence, and a Location NP. The prepositional predicate merges first with the Location argument—an order of merger that seems to be semantically forced—then with the Theme argument doktora ‘doctor.GEN.MASC’ in Spec,PP. Both arguments are θ-marked in PP-internal positions, but the Theme NP undergoes case-driven movement to Spec,VP headed by the copula in order to receive genitive case from the complex Neg head under exceptional case-marking (cf. Bailyn 1997).¹³ As sentential negation is quantificational, it plausibly patterns with other Russian quantifiers in being able to assign structural genitive (cf. Bošković’s (2006) genitive of quantification); furthermore, in the absence of agreement of the NP with Infl, this appears to be the only option. Even though Neg, as a proclitic to V, is not typically in the right syntactic configuration to assign case, movement and head adjunction of the copula to Neg, and the resulting formation of a complex Neg head where ne acts as a prefix, plausibly makes case assignment by Neg possible. One of the consequences of this incorporation is that Neg and the copula form one prosodic word, where word stress in the past tense form obligatorily falls on the prefix;¹⁴ and the present tense form, as in (10b) and (11b), is plausibly derived by the PF rule Neg⁰ /ne + BE_Null/ ⇒ /net/. The Theme NP then undergoes further A-movement to Spec,IP not for reasons related to case, but to encode a referential (and hence nonexistential) interpretation.

In the existential construction, as in (9b), the Theme remains in Spec,VP in order to be interpreted as quantificational. We propose that the surface word order is derived via topicalization of the remnant PP to an IP-adjoined position (see (15)), which is motivated by a phonological constraint that bans sentences from beginning with sentential negation.¹⁵

We argue that GenNeg occurring with the verb byt’ must be analyzed independently of GenNeg occurring with subjects of intransitive verbs or objects of transitive verbs. Crucially, we distinguish two distinct interpretations of the verb byt’ with differing syntactic and semantic characteristics.

Predicative byt’ is a transitive predicate with an external Agent argument and an internal Location argument, whose subject always bears nominative case; GenNeg assignment is not possible with this predicate. Copular byt’ is a copular verb that occurs with a PP, whose head functions as the main predicate of the sentence. This predicate is also transitive, with two internal arguments: a Theme and a Location. In the context of sentential negation, the assignment of genitive case to the grammatical subject of the copular construction (the Theme NP) is obligatory and—unlike in GV constructions—is not used to encode its nonreferential status. Instead, referentiality is encoded by means of word order. With a referential subject, the sentence is interpreted nonexistentially as asserting the absence of the subject’s referent from the specified location. With a nonreferential quantificational subject, the sentence is interpreted existentially as asserting the nonexistence of an individual bearing the property denoted by the subject in the location. Hence, the verb byt’ does not display the optionality observed in the assignment of GenNeg to GVs. Rather, GenNeg is disallowed in the construction with predicative byt’ and obligatory in the construction with copular byt’.

Syntactically, we propose that predicative byt’ acts as a standard transitive verb. In the copular byt’ construction, a referential NP A-moves to Spec,IP; a nonreferential NP remains in Spec,VP and the surface word order is derived through topicalization of the PP constituent.

Our analysis accounts for the apparent inconsistencies of GenNeg assignment with byt’ and captures the observation that copular byt’ is signaled by a different morphological form of negation in present tense negated sentences in which GenNeg is assigned.

We would like to thank Ruslan Sabitov for grammaticality judgments and the two anonymous LI referees for very detailed and useful comments. Material from the squib was presented at the Workshop on (Morpho)syntax at UCL. We would like to thank the audiences for useful comments.