The paper claims that the independent partitive case in Finnic languages and the independent partitive genitive case in Baltic and East Slavic (henceforth: ip(g)) show considerable correlations that cannot be accounted for but by language contact. Given that both the ip(g) in Baltic and East Slavic as well as the ip(g) in Finnic are inherited from the respective proto-languages, the paper also offers a methodological discussion of how inherited categories may also be shown to be subject to language contact. A typologically not infrequent category must be individualized on the basis of a list of properties. Thus, 13 semantic and 5 morphosyntactic properties have been discussed. While the study reveals that in general the ip(g) is or was subject to intensive language contact, there is no common hotbed for all properties analysed and different properties have different hotbeds and are distinct with respect to their geographical distribution and entrenchment. North Russian and Finnic show the greatest degree of correspondence as, e.g., the aspectuality related functions of the ip(g) or the morphological distinction between the possession (sensu lato) and the partitive-related functions are concerned. Here, Finnic is the donor language. However, other properties such as the semantic and syntactic merger of the acc and ip(g) marking must have spread from Russian to Finnic and, to some extent, Baltic. Similarly, the genitive/partitive-under-negation probably developed first in Baltic and Slavic and spread then into Finnic, since preconditions for this rule are already found in the ancient Indo-European languages. Finnic, however, preserves this rule best.
The Circum-Baltic area is an established linguistic area with some subareas, extensively discussed in the literature (cf., inter alia, Matthiassen, 1985; Stolz, 1991; Klaas, 1996; Nau, 1996; Dahl and Koptjevskaja-Tamm, 2001; Wälchli and Koptjevskaja-Tamm, 2001; Wiemer et al., 2014 on the Balto-Slavic Contact Zone, see also Sarhimaa for Karelian-North Russian sprachbund). The languages in focus here are primarily Lithuanian and Latgalian (Baltic), Russian and North Russian (East Slavic) as well as Finnish, Estonian, Karelian, Veps, Ingran (Finnic). I will also sporadically mention Polish (West Slavic), which partly replicates features found in these three branches. Other languages of the area such as Low German, Yiddish, Romani or Karaim will not be considered here for various reasons, one of them being the fact that these languages entered the area at a much later time.
In this paper semantic and morphosyntactic properties of two cases will be compared: the partitive case in Finnic languages and the genitive case in Russian and Baltic. While Finnic languages distinguish between the genitive case that encodes possession and the partitive case that encodes partitivity-related functions to be discussed below, Russian and Baltic (like other conservative Indo-European languages) do not morphologically discriminate between the possessor (sensu lato) genitive and the partitivity-related use of the genitive. To this extent, the languages under investigation are different among themselves, and this difference is due to inheritance.
Though even here, there is a tendency to encode possession and partitivity-related functions differently. Thus, generally in Slavic, possessive adjectives (derived from nouns incl. proper names and pronouns) may be used to mark possession and related functions. In Baltic, the singular of the first and second person pronouns have dedicated possessive forms that are distinct from those used for partitivity-related functions. Lithuanian has dedicated possessive genitive forms man-o ‘of me’, tav-o ‘of you’ as opposed to the genitive man-ęs ‘of me’, tav-ęs ‘of you’ that is used for partitivity-related functions (e.g. under negation or lexicalized partitive use of the genitive with certain verbs). Latvian employs adjectival pronominal forms (man-s ‘mine-nom.sg.m‘, tav-s ‘your-nom.sg.m’) for possessor-related functions and the genitive man-is ‘of me’, tev-is ‘of you’ for the partitivity-related functions (such as negation). Moreover, there is a recent tendency in North Russian dialects (to some extent rendered by colloquial standard Russian) to morphologically distinguish between the possessor genitive and the partitive (genitive) (see subsection 3.2 below). For the sake of simplicity, I refer to all non-possessive uses of the Baltic and Russian genitive that have parallels in the Finnic partitive case as partitive genitive. Note that partitive here is just a label encompassing a whole bundle of semantically and/or diachronically related functions, true partitive uses constituting just a small subset thereof.1
Apart from the partial “possession-partitivity” syncretism of the Baltic and Russian genitive, but not of the Finnic partitive, there are striking correlations between the use of the genitive and the partitive case in these three main language branches of the Eastern Circum-Baltic are. I will explore convergence effects of the ip(g) in this area, tying in to previous research (Larsson, 2001: 244–6; Koptjevskaja-Tamm and Wälchli, 2001: 649–60; Koptjevskaja-Tamm, 2001; Bjarnadóttir & De Smit, 2013: 35–50).
The dependent partitive case in Finnic and the dependent partitive genitive case of Baltic and East Slavic is immediately syntactically dependent on an overt head which may either represent a measure phrase, a quantifier or a verb that governs the partitive or genitive case respectively as in the following example with a measure phrase butelis ‘bottle’ governing the partitive genitive alaus ‘of beer’:
As is obvious from this example, there is no possession relation between the head and the genitive phrase; instead, the genitive phrase encodes the kind of liquid that is being measured by the head np.
In turn, the syntactically independent or bare partitive (genitive) is, so to speak, any other syntactic constellation: the independent partitive (genitive), henceforth the ip(g) (for both the independent partitive case of Finnic and the independent partitive genitive of Baltic and Russian/North Russian), is a case that overrides the default/structural case assignment required by the lexical verb or by a particular construction in order to encode some added meaning (to be discussed below). Thus, the existential predicate byt’ ‘to be’ in Russian/North Russian subcategorizes for nominative subject. However, if the definite reading of the subject np should be excluded the independent partitive genitive may be used as in (2):
Note, however, that the restriction as to structural cases only is not a “universal” condition of the ip(g). Thus, the ip(g) could override also datives, instrumentals and other non-structural cases in ancient Indo-European languages and most probably in Proto-Indo-European (see subsection 3.1.1 below). Crucial to the definition of the ip(g) is that it is not governed by the verb, some other np or a quantifier in terms of a subcategorization frame. While the dependent and the independent partitive (genitive) are obviously related categories, they have developed quite far from each other. Thus, it is only the ip(g) that has functions pertaining to (in)definiteness, aspectuality, negation, etc. (see below).
There are several contexts in which the former ip(g) has been lexicalized and became a part of the case frame of the respective verb (typically intensional verbs allowing for the opaque reading of the object such as to look for, to wait for, to want or verbs that require their object to be quantified, e.g. to lack).2 I will leave these instances outside of the scope of this paper. Similarly, in the context of a predicate negation the ip(g) has been generalized as the only possible object and, with some intransitives, subject marking in some languages of the area (see subsection 3.13 below). To this extent, one may claim that negation licenses the ip(g) at some level which is then no longer syntactically independent. The same is true with the process of generalization of the ip(g) as the direct-object marking for some np types whose referents are high on the animacy scale. This generalization yields various degrees of the acc/ip(g) syncretism in a number of languages of the area (see subsection 3.5 below), making the ip(g) into an allomorph of the accusative case and, hence, syntactically dependent.
While these uses of the partitive genitive and the partitive case are, strictly speaking, no longer syntactically independent, historically they were ip(g)’s to begin with. As such they therefore provide further evidence in favour of our main claim, namely, that there are a number of shared innovations with the ip(g) that have shaped the category of the ip(g) in the languages of the Eastern Circum-Baltic area. The fact that these instances represent exceptions to the definition given above only strengthens the claim that language contact has heavily constrained the ip(g) in the area.
The approach I have selected is semasiological. Thus, the aim is to look into various functional and formal facets of this morphosyntactically and semantically defined category and explore correlations both inherited and acquired across the languages of the area.
Semantically, the ip(g) represents a cluster or multi-faceted category with bearings on at least three domains: quantification, definiteness and discourse prominence. Taken together, these functions may roughly be subsumed under the (non-technical) notion of decreased referentiality (cf. Partee, 2008 on Russian). This is, however, just an approximate semantic definition which does not entirely account for the aspectuality-related function and for the ip(g) under negation in those languages where predicate negation obligatorily requires the ip(g) in terms of a syntactic rule. Here the ip(g)-marked nps may have also definite reference. Moreover, quantification functions may be further divided into those with np-internal functions and those with vp-related functions (see subsections 3.7–3.12); in turn, np-internal quantification functions may be further distinguished into pseudo-partitive functions (Selkirk, 1977) and true partitive functions (see subsection 3.3). It is nearly impossible to provide details of all denotational facets that the ip(g) exhibits in the languages under investigation in the introduction section; the reader is therefore referred to the respective sections.
More generally, the alternation of the structural cases with the ip(g) discussed here falls under the wide notion of Differential Object Marking (dom) and Differential Subject Marking (dsm). Following arguments in Iemmolo (2013) I concede that this type of dom/dsm deviates from the classical dom systems such as the one in Spanish with the preposition a. In the latter there is an alternation between a case-marking and a morphological zero (cf., inter alia, von Heusinger and Kaiser, 2007, 2011). The dom/dsm found with the ip(g) is different in that it involves an alternation between two (at least syntactically) distinct cases, namely, the ip(g) and the accusative case in the object position. However, the differences between the Spanish dom type and, e.g., the Finnic one reveal themselves as not that dramatic. In Finnic languages, the partitive case alternates with a morphologically marked accusative (traditionally referred to as genitive) in the object position and the zero (nominative) in the subject position. Crucially, in Finnic and Russian but not in Baltic, accusative is quite defective and often lacks a dedicated morphological exponent: the accusative forms are morphologically unmarked and indistinguishable from the nominative ones in the plural of Finnish and Russian as well as in the accusative singular of a number of declensions in Russian while Finnic languages employ a syncretic genitive-accusative marker in the singular.3 Table 1 provides a simplified overview over the case syncretism found in the representative languages of the three language branches under investigation.
Note that – confusingly – the possessor-genitive case is syncretic with and typical of acc in Finnic but part in Baltic and Russian: it is the canonical direct object marking in the singular in Finnic, and it is the only way to express partitive-like functions in Baltic and Russian.
The dependent partitive case and the dependent partitive genitive have been extensively discussed in the literature with regard to the Circum-Baltic sprachbund. It constitutes a firmly established feature of the Eastern part of the Circum-Baltic language area (Koptjevskaja-Tamm, 2001). The partitive genitive in East Slavic and Baltic exhibits functional correlations with the partitive case in the Finnic languages that “…are typologically too infrequent to be explained by a coincident parallel development” (Koptjevskaja-Tamm, 2001: 541). Also the idea that the ip(g) shows considerable convergence effects in the area is not new (inter alia, Larsson, 1985, 2001: 244–6; Koptjevskaja-Tamm and Wälchli, 2001: 649–60; Bjarnadóttir & De Smit, 2013). Thus, the excellent overview in Koptjevskaja-Tamm and Wälchli (2001: 649–60) provides important insights that have been integrated into this paper. However, a number of details remained either understudied (such as the interaction of the ip(g) with aspectuality, subsection 3.10) or just not mentioned at all (e.g., the merger of nom-acc and nom-ip(g) alignments in the languages of the area, subsection 3.5).
2 Accounting for Language Contact: Framework
My aim is to provide a thorough analysis of various processes leading to convergence effects of this latter category. This is an especially challenging undertaking as this category is arguably inherited in both Indo-European (Russian and Baltic) and in Proto-Finnic. Moreover, as will be argued below, various properties of this category have been developed in different hotbeds, covering different parts of the area and to different extents. In effect of this layered language contact with no dominance of a particular language in the donor function the emerged category shows considerable variation across the area. A further difficulty is that, on the superficial level, an emancipated/independent partitive expression is also not infrequent cross-linguistically. Thus, the typological triviality is another factor impeding claims about language contact.
Correlations across languages need not be due to language contact per se, but may rather represent typologically unmarked or dominant features motivated by language processing and less by a diffusion process. The extent to which a feature is cross-linguistically common has to be taken into account in order to exclude correlations driven solely by typological unmarkedness or dominance principles (Koptjevskaja-Tamm and Wälchli, 2001). Similarly, etymologically cognate correlating features represent rather a weak indication of language contact and require additional justification (cf. Koptjevskaja-Tamm and Wälchli, 2001; Heine, 2009; Wiemer et al., 2014). I adhere thereby to the Triangulation approach while accounting for the convergence effects put forward in Wiemer et al. (2014: 25). Here, triangulation is meant as a cover term of methods aiming at an equilibration of three factors potentially responsible for the emergence of convergent features, namely: (i) typologically frequent patterns of diachronic change, (ii) contact, (iii) properties inherited from common ancestors. Even though typologically unmarked features as well as features inherited from a common ancestor are not per se an indication of a language contact, the latter nevertheless should not be excluded solely on the basis of null hypothesis. It seems natural that exactly etymologically related features and/or features that are typologically frequent may be prone to language contact. Crucially, even with these features, I believe, it is possible to establish individual parameterization that would be typologically uncommon and, hence, proof for language contact if recurrent in the area.
There is strong evidence that both the ip(g) in Baltic and Slavic and the ip(g) in Finnic is an inherited category, which means that it must have been present in the three branches (i.e. Baltic, Slavic and Finnic) before these languages came into contact. Thus, the ip(g) in Finnic arose from a separative or ablative case which had already entered the domain of the direct object before the Proto-Finnic period. The evidence for this is provided by non-Finnic languages of the Uralic branch such as Sami and Mordvin (cf., inter alia, Itkonen, 1972: 185; Laanest, 1982; Campbell, 1990: 66ff; Kiparsky, 1998; Harris and Campbell, 1995: 362–3). Analogically, the Baltic and Slavic partitive genitive is the etymological and functional continuation of the Proto-Indo-European partitive genitive widely attested across the ancient ie languages (cf., inter alia, Bauer, 2007; Dahl, 2009; Luraghi, 2003: 60ff; Nachmanson, 1942; Napoli, 2010; Schwyzer and Debrunner, 1950; Seržant, 2012a, 2012b).5 The ip(g) is also an established category in all old and some modern Slavic languages such as Old Church Slavonic, Russian, Ukrainian, Czech, Polish (Miklosich, 1926: 427).
Given the autochthonous origins of the category in each of these language branches, a question emerges whether language contact still may sufficiently be argued for. To prove this, I will zoom in into particular properties of the category at issue, looking for a specific subset of properties that are: (i) innovative in at least one of the language branches (i.e. not present or different at a proto-stage) and (ii) recurrent in at least two branches. Such a correlating set of (at least, partly) innovative properties pertaining to different grammatical layers (such as morphology, syntax and semantics) establishes typologically individual parameterization of an areal category which makes this category stand out against the typological background. Moreover, complexity of correlations is yet another factor disambiguating language contact from typologically trivial patterns as the reason for convergence (cf. Principle of Complex Correlations in Seržant, 2010: 195). Establishing such an at least partly common set of properties will not only provide for individual parameterization of the category in question but will also answer an essential question: To what extent has there been language contact with regard to the category of the ip(g) in the Circum-Baltic area?
For this purpose, I carry out a comparison of a range of particular properties (P1-P18) among the languages under investigation subdivided into denotational properties (P1-P13, Section 3) and formal properties (P14-P18, Section 4), supplied with historical data whenever available. Section 5 summarizes the results.
3 Properties Check: Functional Properties (P)
This section treats a set of functional properties. Properties are understood more generally here in the sense that there is some rule or a semantic context that triggers the ip(g) marking. Notably, none of the properties to be discussed below is typical for the functional domain of case (as per Blake, 1994: 1–2).
The following discussion of the denotational properties is divided into np-internal properties (P1-P6) and clause-level properties (P7-P13).
3.1 Decreased referentiality (P1)
Generally, a bare np can readily have definite interpretation in these languages, as there are no grammaticalized means to mark the definiteness of an np. However, the ip(g) marking blocks this interpretation, and the respective np can only be interpreted as low referential (e.g. indefinite):6
In most of the cases the ip(g) induces even non-specific indefinite reading in terms of scopal specificity (Farkas, 1995), signalling narrow scope with regard to other quantifiers on the np-internal function of the ip(g), i.e. excluding negation and aspectual functions. Cf. de Hoop (2003) for Finnish, Kuznecova (1964: 7–10), Markova (2008: 155) and Seržant (2014b) for North Russian, Padučeva (1998), Partee (2008) or Timberlake (2004: 324) on standard Russian, Seržant (2014a: 267–8) on Lithuanian.
It is quite difficult to provide an invariant meaning here. However, it is obvious that the ip(g) considerably decreases the referentiality of the np (cf. Partee, 2008 for Russian). Typical for the decreased referentiality is a frequent combination with such attributive modifiers requiring indefinite non-specific nps as the Latg. vysoks ‘any kind of’, Lith. visoks ‘idem’, Russ. vsjakij ‘idem’ and its lexicalization (becoming part of the verb’s case frame) with intensional verbs such as to look for, to search, to wait for, to want, etc. which are conducive of decreased referentiality. This functional property of the ip(g) is well attested already in the ancient ie languages (cf., inter alia,Luraghi, 2003: 60ff; Napoli, 2010; Seržant, 2012a) and, hence, must have already been present in the ancestor language, namely, Proto-Indo-European (henceforth: pie). This property is thus inherited in Baltic and Slavic. It is also inherited in Finnic: the ablative case – the precursor of the Finnic ip(g) – is also found with the same function in Mordvin:
Recall that this language represents a more archaic state of affairs with respect to the ip(g) in comparison to Finnic (Itkonen, 1972; Laanest, 1982; Campbell, 1990: 66ff; Kiparsky, 1998; Harris and Campbell, 1995: 362–3).
These facts above suggest that the decreased referentiality function of the ip(g) is not itself due to language contact in either of the languages but has to be explained as concomitant to independent partitive and partitive-like expressions.
From the typological perspective, two points reveal this property of the ip(g) as striking. In addition to the fact that the ip(g) case-marking has properties that are not typical for a morphological case (Blake, 1994), it is also striking for the following reason. The dom phenomena typically employ marking of the more salient participant while leaving the less referential participant unmarked. In this case, however, the ip(g) yields dom that is based on marking of the less referential participant while leaving the more salient participant in many places unmarked. Moreover, the very fact that the same category has been employed in these languages for both dsm and dom phenomena seems revealing.
3.2 Discursive Backgroundedness (P2)
Decreased referentiality is implicationally related to discourse backgroundedness. It is in fact often difficult to discriminate between these two. The ip(g) marked argument may form with the verb a unified information-structure unit. More often than not, the whole vp is in the focus and not its ip(g) marked nominal. This has to do with the fact that the ip(g) usually do not introduce discourse topics but rather provide for background information, cf. Lithuanian:
The following example from North Russian illustrates the same point. While elsewhere existential clauses are typically used to introduce new discourse participants, the ip(g)-marked subject kulakov ‘kulaks’ is only locally relevant. The speaker tells of her life before. She says that her family was poor, while there were some kulaks in the village (some of which were later dispossessed). Her point is not about kulaks themselves, which only provide contrastive background information:
Helasvuo (1996) states that the ip(g) in Finnish is also conducive to discourse backgroundedness in the existential sentences; Tveite (2004: 150) comes to the same conclusion for Livonian. The discursive backgroundedness is also found in North Russian (Seržant 2014b: 307–9). Symptomatic for this function is the respective word order. Thus, the ip(g) marked subjects tend generally to occur in postverbal position in Finnish (Karlson, 1987: 77; Sands and Campbell, 2001: 257), North Russian and Lithuanian, which is an unusual position for subjects in these languages. Moreover, the ip(g) marked objects are almost never fronted.
Analogically to the previous property, this property is already found in the ancient Indo-European languages (Seržant, 2012a), and represents rather a common inheritance in Baltic and Russian. At the same time, the ablative-case-marked direct object nps of Mordvin (as in ex. (7) above) indicate that this property is inherited also in Finnic.
3.3 Partitive vs. Pseudo-Partitive Functions (P3)
Consider example (10) from English:
True partitivity is found only if a subset (Some) of a particular, discursively accessible delimited group, namely, the superset (our students) is affected by the event while the remainder (the other students) is not (cf. Heusinger, 2002; Heusinger and Kornfilt, 2005; Kornfilt and Heusinger, 2009). Pseudo-partitivity, in turn, is found when the superset does not encode a particular, discursively retrievable, delimited set but is rather extended to kind-referring nps.
Thus, in example (11), the subset ‘some water’ does not have a proper superset in the discourse model, the superset being extended by default to the whole kind water. This, in turn, is a violation of the Partitivity Constraint first formulated in Jackendorf (1977) with different refinements by, inter alia, Barwise and Cooper (1981), Ladusaw (1982), Dowty and Brodie (1984), Reed (1989), de Hoop (1997, 2003), Ionin et al. (2006). This constraint requires the superset to be a definite and discursively accessible set. The violation of the Partitivity Constraint makes the whole expression devoid of its original “set-from-set nature” (that is part-of-relation) and turns it into an instantiation expression like the English a cup of water. The latter is basically parallel to some water, cf. German eine Tasse Tee ‘a cup of tea’, and does not presuppose any other water parts not affected by the predicate in the given discourse model. The ip(g)-marked np only informs about the kind of the referents to be selected by the head and does not provide their superset; hence, there is no part-of relation in any reasonable sense (Selkirk, 1977; Koptjevskaja-Tamm, 2001: 523). The development from the partitive into a pseudo-partitive is thus immediately related to what kinds of selectional restrictions are imposed on the np marked by the ip(g): if this np is (non-generic) definite, then we have a partitive expression, otherwise it is a pseudo-partitive one.
The ip(g) in Baltic and Slavic does not impose selectional restrictions on its np such as required by the Partitive Constraint. In addition to the definite nps it freely allows also for indefinite and non-specific, kind-referring nps to occur in:
The same loosening of selectional restrictions is found in Finnic.10 Even more, there is a tendency in this branch towards assuming only kind-referring nps with the ip(g), while true partitivity (with definite supersets) tends to be encoded by means of the new partitivity marker, namely, the elative case (Alho, 1992; Itkonen, 1972: 181 for the same development in West Saami).
The loosening of the selectional restrictions on the np is a common innovation of Baltic and Slavic, because the ancient ie languages typically prefer definite nps with the ip(g) (Seržant, 2012a: 122), cf. from Ancient Greek:
Thus, Baltic, East Slavic and Finnic show the same tendency. However, one cannot claim that language contact is the main motivation behind this convergence. This is so because the development from the partitive into the pseudo-partitive is widely attested cross-linguistically and represents quite a general developmental cline. For example, the partitive meaning has been lost with the so-called partitive article of French des (as well as with mass nouns du/de l’/de la), the latter having rather the function of a pseudo-partitive, cf. also faded partitives in Dutch with the originally partitive preposition van ‘from, of’ (de Hoop, 2003: 193–99).
3.4 Selectional Restrictions and Productivity (P4)
3.4.1 Selectional Restrictions on the np
Modern Standard and even Middle Russian (Krys’ko, 2006: 225–6), as well as Lithuanian (Seržant, 2014a) and, to a much greater extent, Latvian (Berg-Olsen 1999) impose considerable restrictions on the lexical input of the ip(g), especially when the latter is used on its np-internal readings. Here, Lithuanian allows only for mass nouns, abstract nouns and plurals to be marked with the ip(g). The ip(g) is almost extinct in Standard Latvian, continuing to be productive only in some varieties of Latgalian (Nau 2014).
Moreover, intensional verbs show coherence across all the languages under investigation except for Latvian in that they all require or allow for the ip(g) which, in turn, is not restricted in any way. I assume that this coherence is motivated by inheritance and typologically frequent developments for the following reasons. The use of the ip(g) with intensional verbs and its use in np-internal quantification are acquired earliest diachronically, which is why the languages do not considerably differ in this respect (subsection 3.4.2. immediately below).
3.4.2 Selectional Restrictions on Transitive Verbs
Finnic languages stand out by having much fewer selectional restrictions on the verb than any other language in the area. Finnic languages allow both perfective and imperfective viewpoint aspects with the ip(g) and disallow it only with inherently telic and atelic verbs (see details in subsection 3.10 below).
Standard Russian allows the ip(g) with only few transitive verb classes. These are mainly transfer verbs like to buy, to take, to give, consumption verbs as to eat or to drink or intentional verbs such as to want, to wait for, to look for. Moreover, the np-internal quantificational use of the ip(g), i.e. with the first two verb classes, is only possible with perfective verbs in Standard Russian and the perfective context in Lithuanian. The situation in the modern Russian and Lithuanian dialects is less restrictive, e.g., in Belarusian (Karskij, 1956: 319, 403) or North Russian, Northwestern Russian (the so-called Pskov Group) and in some neighboring Central Russian subdialects (Filin, 1972: 514–5; Kuz’mina, 1993: 36–7). In general, the ip(g) reduced its productivity from the 19th c. to 20th c., also becoming less frequent in North Russian and Belarusian due to the growing influence of Standard Russian. Nevertheless, there are fewer lexical and positional restrictions in these East Slavic branches than in Standard Russian. Furthermore, clear-cut dialectal clines are observable: the ip(g) is most unrestricted in the North and Northwest Russian subdialects, while it becomes more constrained gradually towards the East and the South (Kuz’mina, 1993: 29, 36–7). Thus, in contrast to Standard Russian and from other Russian dialects, the ip(g) may freely occur with imperfective verbs, including the incremental-theme verbs in North and Northwest (Kuz’mina, 1993: 30). North Russian is thus even less constrained than Lithuanian. The latter namely typically blocks the ip(g) with incremental-theme verbs in the contexts of imperfective viewpoint (Seržant, 2014a: 283–9).
Other indications for gradual loosening of selectional restrictions towards the geographic area of Finnic can be adduced. Such accomplishment verbs as ‘to open’ do not allow for ip(g) in Lithuanian, Latgalian or Standard Russian, while this verb readily allows for the ip(g) in North Russian, cf. (61) from North Russian and (62) from Finnish below. This concerns not only the aspectually relevant uses of the ip(g). There are also much fewer restrictions with quantificational, np-internal uses of the ip(g) in North Russian and neighboring West Russian subdialects than in Standard Russian with both imperfective and perfective verbs (Kuz’mina, 1993: 30–1). Moreover, the data from North Russian suggest that not only has North Russian extended its use of the ip(g) onto such verbs as to open or to wet with an aspectual function (Seržant 2014b: 287–8) but it also has preserved a number of verbs that allowed for the ip(g) in the ancient Indo-European languages and in Old Russian (Krys’ko, 2006) but have generalized the accusative in Standard Russian and even in the conservative Lithuanian. These are, for example, the verbs of perception that generally encode the stimulus np with the partitive case in Finnic (which is thus syntactically no longer independent here), cf. North Russian: smotret’ ‘to watch’, videt’ ‘to see’, slušat’ ‘to hear’ but also other psych verbs zabyt’ ‘to forget’ (Kuz’mina, 1993: 32–3). While the ip(g)-marked stimulus is most probably an inheritance in North Russian, it is nevertheless striking that precisely North Russian has best preserved this marking. I take this as another piece of evidence for the impact of Finnic languages on North Russian which, yet, has a conservational and not borrowing effect.
Larjavaara (1991) reasonably argues that the semantic evolution of the partitive case in Finnic went through a quantificational meaning such as ‘some Swedes’ in (5) above to the aspectual function (to be discussed in detail in subsections 3.10–3.12 below). I also adopt this scenario for Baltic and Slavic. This relative chronology must also be extended by other uses of the ip(g) such as predicate negation or the intensional function. It is striking that the verb classes used with the ip(g) in the ancient Indo-European languages and in Mordvin match to a considerable degree, although any kind of language contact can be safely excluded here. Thus, Mordvin, which is our source for the Proto-Finnic situation of the ip(g), attests intensional predicates such as to need or to want (Larsson, 1984: 97), experiential verbs to listen or to see, transfer verbs to bring, to take, to steal, to give and some others used with the ablative case on the direct object (Itkonen, 1972: 170; Larsson, 1983: 125ff.; Kiparsky, 1998). Pretty much the same verbs are found with the ip(g) in the ancient Indo-European languages (inter alia, Schwyzer and Debrunner, 1950; Kuryłowicz, 1964: 184). The use of the ip(g) with intensional verbs is already attested in the ancient Indo-European languages and must also be of Proto-Finnic origin. This suggests that the use of the ip(g) with intensional verbs is also diachronically prior to the aspectual functions. Furthermore, the ip(g) with negation has been generalized only within Finnic and Lithuanian, Polish, Old Russian. The relative chronology can be represented as follows:
Since different functions impose different input restrictions (on both np and the verb), the relative chronology in Table 2 must correlate with the relative chronology of the loosening of the respective input restrictions. Obviously, aspectual functions enlarge the range of verbs that may take the ip(g).
3.4.3 Selectional Restrictions on Intransitive Verbs
The selectional restrictions found with intransitive verbs are conditioned by two sets of factors: (i) the lexical properties of the verb and (ii) the syntactic constraints on non-nominative subjects in the given language, since the sole argument of intransitive verbs is mostly the subject.
While Lithuanian, Latgalian and North Russian allow for the ip(g) in the subject position of many inactive intransitive verbs (such as to be, to rise, to arrive, to appear, cf. Nau, 2014; Seržant, 2014a, 2014b), Modern Standard Russian, southern and eastern Russian subdialects only sporadically attest the ip(g) in the subject position (Kuz’mina, 1993: 114). In Latvian, the ip(g) is attested only in the earlier language layers such as folklore texts or in some more archaic subdialects but not in the present day language. The same holds for Modern Standard Russian which did attest a number of existential predicates allowing for the ip(g) subject marking in Early Modern Russian (18th–19th cc.).
Dialectal Polish spoken in Lithuania (polszczyna kresowa, not mentioned in Table 4 below) – differently from Standard Polish – patterns here with Lithuanian, e.g., by allowing the ip(g) in the subject position (Adamovičiūtė and Čekman, 1984: 9–11). Data from archaic Latvian texts suggest that Latvian was originally quite similar to Latgalian but then for some reason lost the ip(g) except for the negation function where it is also quite marginal (Berg-Olsen, 1999). Otherwise, languages that are more distant from Finnic, Baltic and the western dialects of East Slavic show even stronger selectional restrictions as Koptjevskaja-Tamm and Wälchli (2001: 659–60) note citing Standard Polish as being more conservative and Czech with almost no instances of ip(g) subjects.
Again, Finnic languages exhibit much fewer restrictions here. The ip(g) is allowed not only in the subject position of the unaccusative and unergative intransitives in many Finnic languages but, in colloquial Finnish, it is also found in subjects of transitive verbs in sentences with a full direct object (Hakulinen and Karlsson, 1979: 16711); the same holds for Veps (Lytkin et al., 1975: 108; Koptjevskaja-Tamm and Wälchli, 2001: 658):
‘Expecting mothers, servicemen, and students used their right to vote in Salo on Monday.’ (Hakulinen and Karlsson, 1979: 167)
This is especially striking, since this type of dsm works against the discrimination between subjects and objects (A/S and O in typological terms) – something quite atypical of dsm (see, e.g., the discussion of dom/dsm in Malchukov, 2008: 211ff; Moravcsik, 1978: 250). The following table provides the relative chronology of the ip(g) in A, S and O marking (Moravcsik, 1978; Sands and Campbell, 2001; 256; Seržant, 2013: 336–7):
As the data from Mordvin show, the ip(g) was assumedly not possible in the subject position in Proto-Finnic. At the same time, ancient Indo-European languages do attest the ip(g) in the subject position of a number of inactive intransitive verbs (inter alia, Conti & Luraghi 2010; Seržant, 2012b). One could thus argue that the extension of the ip(g) in Finnic onto the subject position might have been the result of Indo-European, more specifically, Baltic and Russian/North Russian impact. At the same time, as mentioned above, the extension of the ip(g) onto subjects has proceed in Finnic much further affecting not only active intransitive but also some transitive subjects in colloquial Finnish. One is thus forced to assume that there was first some Indo-European influence providing for subject ip(g)s in Finnic, which then became even more productive than in the source language.
3.4.4 Selectional Restrictions, Summary
This geographic distribution of different productivity degrees and selectional restrictions speak in favour of areal impact. It is, however, quite difficult to establish a common epicentre here. The selectional restrictions are of course immediately related to other properties. Thus, the aspectual functions of the ip(g) are most developed in Finnic; as a consequence, these languages allow more verbs to have ip(g) object marking (cf. subsections 3.7–3.10 below).
As can be observed from Table 4 below, Finnic scores highest by having the weakest input restrictions in all functions of the ip(g). North Russian is second – it allows for some non-incremental-theme verbs such as otvorit’ ‘to open’ to take the ip(g) to measure the event, cf. ex. (61) below. Thirdly, Lithuanian, allows aspect-related functions of the ip(g) only with incremental-theme verbs and transfer achievements in its East while scoring similar to Finnic and North Russian with respect to other functions, Latgalian ensuing. I summarize this in the following table:
The negation function that does not impose any restriction on the argument type whatsoever is clearly a new development in both Finnic and Baltic/Slavic, however, with some prerequisites in Proto-Indo-European (see subsection 3.13 below).
To conclude, in addition to the inherited contexts, new contexts allowing for the ip(g) have been created, indicating the mutual influence of the languages at issue.
3.5 Conflation of Two Transitive Alignment Patterns: nom-acc and nom-ip(g) (P5)
Typical for many languages of the area is the tendency – albeit to different extents – to semantically and syntactically merge both competing transitive alignment patterns nom-acc and nom-ip(g) into one.
3.5.1 Russian and North Russian
Russian is the most progressive language in this respect. In all varieties of Russian the nom-acc and nom-ip(g) patterns merged into one pattern, namely nom-ip(g), reanalysed as nom-acc2 with animate nps of the following types: all pronouns, (masculine) animate nouns of the o-declension and plurals of all nouns. Since this process has affected only those nps that refer to animate referents but not those that refer to inanimate referents, it yielded the well-known animacy-driven Differential Object Marking of Russian. Here, inanimates and singular animates of the a-declension still allow the meaningful alternation between the nom-acc and the nom-ip(g) pattern while the aforementioned animate np types take only the nom-acc2 (< ip(g)) pattern which represents in itself the result of the merger. This is not the place to lay out how exactly the semantic neutralization of the nom-accvs.nom-ip(g) patterns happened in the history of Slavic and, specifically, of Russian/Old Russian. The reader is referred to Klenin (1983) who speaks about „accusative-genitive transitivity” (see also Kry’sko, 1994; 1997). It suffices to say that inherently definite expressions such as proper names, demonstrative and personal pronouns and, to some extent, any animate nps were not subject to scope ambiguities which were, in turn, typical for inanimate nps with regard to intensional and negation operators – contexts in which the ip(g) was generalized in Old Russian. Lack of scope ambiguities with the aforementioned np types (in contrast to inanimate np types) made the low-prominence interpretations typical of the ip(g) unavailable in the aforementioned contexts with animate nps and led to the semantic conflation of the nom-acc and nom-ip(g) with animates (Seržant, to appear-b).
Moreover, it seems that North Russian shows a slightly higher degree of progress than Standard Russian by treating some collective nouns designating animals such as skot ‘cattle’ as animates in the direct object position and, hence, subject to the semantic conflation of the nom-acc and nom-ip(g) patterns (Seržant 2014b: 317–9).
3.5.2 Baltic: Latvian and Latgalian
Furthermore, the conflation of acc with the genitive (and, hence, originally the ip(g)) is also found with personal pronouns of Latgalian (Baltic, Latvian). Thus, Leikuma (2010: 63) reports that central dialects tend to generalize the genitive forms maņa/mane ‘me’ and teva ‘you’ of the first and second person singular pronouns for both accusative (= Finnic total/acc) and genitive (= Finnic partitive). They thus fully adhere to the generalization of the (partitive) genitive form in Russian. The south-western Latgalian subdialects, in turn, tend to generalize the accusative forms mani ‘me’, tevi ‘you’ respectively (Leikuma, 2010: 63), adhering to the common Latvian way of abandoning the acc/ip(g) alternation in favour of the acc. This is probably due to the closeness of the latter subdialects to Standard Latvian, which has lost the ip(g) in favour of the accusative altogether.
Nau (2014: 214–17) provides detailed figures for the competitive use of the acc and gen (< ip(g)) forms with personal pronouns including the third person in the subdialect of Viļāni which is rather central than south-western. Nevertheless, in this subdialect, it is the accusative form that most frequently replaces the genitive form with rare instances of the opposite constellation of the gen for the acc. Moreover, Nau (2014: 237–8) finds attestations of the proper names marked by the ip(g) in a variety of Latgalian (while other varieties she has checked lack this) which cannot be accounted for but assuming the begin of the syntactic/functional syncretism of the ip(g) and acc with proper names in the Slavic manner.
Crucially, in both cases the encoding of the (partitive) genitive and accusative objects has been levelled out in favour of a unified case-marking with personal pronouns. Revealingly, in none of the Latgalian subdialects has the merger of accusative and (partitive) genitive with personal pronouns affected the encoding of the possession-related functions. Here, there is another dedicated morphological means: internal possession is typically encoded by the agreeing adjective-like forms of the personal pronouns. The same is true for Russian/North Russian.
3.5.3 Finnic: South Finnic and Saami
This merger has not yet affected Finnish, whereas the situation is different in the closely related Estonian. Here, the conflation of the nom-acc and nom-ip(g) patterns is underway with personal pronouns. Estonian strongly prefers partitive case-marking with 1st, 2nd person and reflexive pronouns in the singular and requires the partitive marking in the plural in those contexts in which all other np types would have been marked with the canonical direct-object (total) marking, i.e. by the syncretic “accusative” (gen in the singular and nom in the plural, see Table 1 above), in the context of totality (non-negated, non-intensional, definite, etc.) (L. Lindström, p.c.; Lees, 2003: 1). Contrast the accusative with a noun in (16) with personal pronouns in (17) and (18) in the same sentence:
In the same context personal and reflexive pronouns may also take the partitive case with no differentiation in the meaning. While singular indistinguishably allows for both options, plurals take only the partitive case (Liina Lindström, p.c.):
Moreover, as Lees (2003: 2) shows, the frequency of the ip(g) in the same parallel text in Estonian and Finnish is quite different, with Estonian having many more partitives than Finnish.12
The Differential Object Marking accvs.ip(g) is abandoned also in other South Finnic languages to various degrees from a meaningless alternation such as in (17) to the solid ip(g) in (18) exclusively. It is found in Livonian (Kont, 1963: 103–6; Tveite, 2004: 38–9), Votic (only rarely accusative/genitive plural forms may be found) (Markus and Rozhanskiy, 2011: 230) but also in the North Finnic Ingrian (Rozhanskiy, p.c.). Notably, the non-Finnic Saami has generalized the former partitive plural as the only accusative plural marker with no differential semantics whatsoever, e.g. in the eastern Saami branch in Russia (Itkonen, 1972: 178). Finally, there is an expansion of the ip(g) on the lexical level. Thus, several verbs in Estonian have generalized the ip(g) object marking (Tamm, 2006).
In conclusion, there is a common tendency to South Finnic (plus Ingrian), some Latgalian dialects, North Russian and Russian to merge acc and the ip(g) into one case, thereby abandoning the acc/ip(g) dom to a different extent. This loss proceeds along the Extended Animacy Hierarchy (Croft, 2003: 130) as in Table 5 below:
In addition to the Extended Animacy Hierarchy the conflation of both alignment patterns is also constrained by verb classes and construction types (cf. Nau, 2014: 240 for Latgalian).
The abandonment of the ip(g) vs.acc alternation is unambiguously a common innovation in all three branches, since it is not inherited from any of the respective proto-languages. While Russian has progressed much further on the Extended Animacy Hierarchy, Latgalian and Southern Finnic languages still attest the dom with the third person pronoun and with other np types further down the hierarchy.
Since this merger is already attested in Old Church Slavic (within the singular only) and is, hence, a common Slavic development, I assume that Russian represents here the hotbed for this process. This is also suggested by the internal East Slavic geographic distribution of this process. Thus, Ukrainian, south-western Belarusian and western Russian dialects do retain old accusative forms with animate non-human plurals here, not replacing them with the genitive as in Standard and North Russian (Kuz’mina and Nemčenko, 1964: 166–8).
3.6 ‘One’ as a Possible Interpretation (P6)
The ip(g) induces an implicit quantifier. Being implicit, it lacks a particular value and therefore has “to look” elsewhere for its interpretation (Neidle, 1988; Franks, 1995: 182). Thus, it may be specified by another quantifier in the clause (see subsection 3.8, 3.9), or, alternatively, by default, it receives an indeterminate or arbitrary value. This is the state of affairs found in the ancient Indo-European languages, where the interpretation is indeed fully arbitrary and not restricted. Being indeterminate, it consequently allows for the interpretation as ‘one’ (Seržant, 2012a: 123; 2012b: 190–2), cf.:
The ip(g) thygatrôn [lit.] ‘of daughters’ refers to one particular participant, namely, just one daughter (scil. Deipyle) that he (scil. the father of the speaker, namely, Tydeus) has married. This interpretation is no longer available in Baltic or Russian and, crucially, is equally impossible in the Finnic languages. Thus, examples like (19) would result in ungrammaticality in all three branches.
A specific property of a category can be lost without any contact influence, simply because any loss is a typologically trivial process not needing a particular external trigger. Nevertheless, it is striking that Baltic and Slavic have lost exactly that interpretation of the ip(g) that is not attested in the Finnic languages. In terms of cumulative evidence, I consider this to be another indication for the process by which these three branches gradually accommodate their inherited categories into a common pattern.
3.7 Development of the Clause-Internal Quantifier Readings (P7)
Originally the implicit indeterminate quantifier of the ip(g) applied only np internally in both Indo-European13 and Proto-Finnic (as evidenced by Mordvin). One of the original functions has been the meaning of partial affectedness of the np’s referent (quantificational function in terms of Larjavaara, 1991). This function has also been preserved in the modern languages, cf. Jakobson (1936: 38), Babby (1978: 15–18), Crockett (1976: 314) on Russian. The partial-affectedness reading can be found, if the np embedded under the ip(g) is definite or has a definite interpretation:
The likeliest interpretation of (20) is that a particular subset of the water delimited by the jar has been consumed, which means that the water in the jar has been partially affected. In turn, partial affectedness with count-noun singulars is already scarcely attested in Old Russian, cf. the following example from 16th c. (Nikon’s Chronicle, xiI 155) from Malyševa (2008b: 237):14
A striking property of the ip(g) is that the domain of application of the implicit quantifier has been extended from purely np-internal quantification to the predicate quantification, and in some cases to the quantification of the whole event. At some point in history there must have been a development from ip(g) being a D(eterminter)-quantifier (np-internally only) into being an A(dverb)-quantifier (np-externally, clause level), cf. Löbner (1985) or Partee (1995) on these notions. Ambiguity contexts necessary for the functional extension have been provided by utterances as in (21), which is in fact ambiguous between (a) ‘…the gravestone of metropolitan Filipp has been destroyed a little bit.’ and (b) ‘a little bit of the gravestone of metropolitan Filipp has been destroyed.’ These two readings basically boil down to the same state of affairs, namely, of the gravestone being somewhat destroyed, the differences being hardly pragmatically relevant. Moreover, incremental theme verbs must have played an important role here, because they provide a natural “bridge” between the np-internal and vp-quantifier.
The acquisition of the A-quantifier functions (in addition to the original D-quantifier functions) is found in all three branches, but, again, to a different extent: Baltic, East Slavic and Finnic. There are three steps in this development: (i) the implicit quantifier of the ip(g) interacts with other quantifiers present somewhere in the clause, (ii) the implicit quantifier of the ip(g) is the only quantifier in the clause but applies to both levels: clause and np-internally, and (iii) the implicit quantifier is the only quantifier in the clause quantifying the whole event but not the referent of its host np. Thus, in Baltic, Russian/North Russian and Finnic, the ip(g) may now be triggered by an overt A-quantifier in any position in the clause. Most obviously: verbs with prefixal quantifiers (subsections 3.8) or adverb quantifiers representing constituents on their own (subsections 3.9), i.e. step (i). Furthermore (step ii), the np-external quantification also manifests itself in the interaction between the implicit quantifier and the aspectual interpretation of the verb (subsection 3.10). The final step (iii) in this development is found when the implicit quantifier no longer quantifies both – its host np and the predication simultaneously – but starts quantifying only the predication, while the respective np is affected holistically throughout the event, as in (61) or (62), cf., e.g., temporal transfer in subsection 3.11 below.
While developments from a D-quantifier into an A-quantifier – and not the other way around – are frequently attested cross-linguistically (Keenan and Paperno, 2012: 948), such cases are reported only with respect to overt, non-flectional quantifiers that can be moved across the clause. This is crucially different with the implicit quantifier coded by the ip(g) which is morphologically tied to its host np. Moreover, in most of the developments from a D- into an A-quantifier, one finds some additional marking with the new A-quantifier that is not present with the former D-quantifier (Gil, 1993; Keenan and Paperno, 2012: 941), cf. English some (np-internal) vs. somewhat (np-external). The ip(g) implicit quantifier is marked in the same way in its A- and D-functions. Furthermore, the morphotactic (paradigmatic) nature of the morphological marking of the implicit quantifier is the one of a case marking being thereby heavily distinct from other quantifiers that have undergone this development. I conclude that the np-external quantification functions of a quantifier with these morphological and syntactic properties make the ip(g) a typological quirk.
There is strong (negative) evidence from ancient Indo-European languages that the ip(g) has ever had functions outside its host np (cf. Fn. 13 above). Nor are such cases reported for Mordvin either. Thus, I consider this emancipation of the implicit quantifier from its host np onto the clause level as a common Eastern Circum-Baltic innovation with no prerequisite in the respective ancestor languages. I discuss different aspects of this process in the following subsections.
3.8 Sensitivity to Adverbs Quantifying the Whole Situation (P8)
The change in the domain from an np-internal into a clause quantifier explains its sensitivity to overt adverbial quantifiers that quantify over the event, such as, e.g., North Russian malen’ko ‘a little bit’ or edva ‘scarcely’ (cf. Malyševa, 2008: 237):
In this example the adverbial quantifier malen’ko (accompanied by the delimitative prefix po-) does not quantify over the referent of the npgolovy marked with the ip(g) (in terms of the affected parts), but rather over the whole predication (the act of raising), cf. analogous to (23) not intending to mean *so much of her face and (24) equally not intending to mean *so much of my eye:
This semantic dependency of the ip(g) from predicate quantifiers is not typical for Baltic or Standard Russian, but is productive in the North Russian subdialects (cf., inter alia, Malyševa, 2008) and the Finnic languages:
As can be concluded from the grammaticality judgements of examples (26a) and (26b) (Tuomas Huumo, p.c.), it is the predicate quantifier paljon ‘much’ that blocks the accusative and requires the ip(g) marking on the object. Note that parallel to the North Russian examples above, the object is affected holistically (not intending *much of the room).
To conclude, property P8 groups Finnic and North Russian together, while Standard Russian patterns with Baltic in this respect. On the bases of this distribution, I conclude that this property has been copied from Finnic into North Russian.
3.9. Sensitivity to Prefixal Quantifiers (P9)
While there are no comparable examples from Finnic, in Baltic and Russian/North Russian the predicate quantifier may also be expressed by means of derivational morphology, i.e. by prefixes, such as Russian po-na- or Lithuanian pri-, at-, už- (cf. Ambrazas et al., 1976: 24–25), Latvian at-, pie-, Latgalian pī-, pa-sa- (Nau, 2014: 246–7) all inducing the meaning ‘so much’:
Note that the verbs in (27) and (30) govern regular nominative subjects and in (28)-(29) the regular accusative object without these quantifiers. It is only the presence of the quantifying prefixes that makes the use of the ip(g) grammatical.
Crucial to the language contact account is the fact that verbal quantifiers such as these probably did not exist in Proto-Slavic or in Proto-Baltic, let alone in Proto-Indo-European – no parallels are found in the ancient Indo-European languages. One may assume that the creation of this type of quantifiers incorporated by the predicate must have resulted from language contact between Baltic and Russian/North Russian, the latter being the source language. From this it follows that also the use of the ip(g) with these predicates is a relatively recent innovation, because it is crucially dependent on the existence of these quantifiers. I assume that two processes come together here: (i) the extension of the domain of application of the implicit quantifier from an np-internal into a clause level quantifier (P7) as well as (ii) the creation of the prefixal quantifiers in Baltic, assumedly under Slavic influence. This property is thus implicationally related to P7 not representing evidence on its own.
3.10 Interaction with Aspectual Features (P10)
While there seem to be no constraints on the interaction of the ip(g) with actionality classes (Aktionsarten) or grammatical aspect (as instantiated by the distinction between the aorist and imperfect, etc.) in the ancient Indo-European languages (Bauer, 2007: 134; Seržant, 2012a; differently Dahl, 2009: 37–41 and Napoli, 2010, see, however, Fn. 13 above), Baltic and, to a much greater extent, Russian/North Russian introduce restrictions on the occurrence of the ip(g) sensitive to the boundedness value of the predicate.
At first glance, the interaction of the ip(g) and aspect found in Finnic seems to be in contrast to what is found in Russian and Baltic (as per Koptjevskaja-Tamm and Wälchli, 2001: 653): while the ip(g) in Finnic is supposed to induce the imperfective aspect (cf., inter alia, Larsson, 1984: 105; Krifka, 1989; or even progressive aspect as per Filip, 1989),15 in the general discussion on verbal partitivity it has gone unnoticed outside Slavic linguistics that the ip(g) in Baltic and Standard Russian is – with a few exceptions – only compatible with the perfective aspect (see, inter alia,Jakobson, 1936; Padučeva, 1998; Mehlig, 2006; Seržant, 2014a, 2014b). However, as I will argue below, the situation is more complicated than this, and Finnic, Baltic and Russian do yield similar results with respect to aspectuality if analysed at a more fine-grained level. In order to show this however, I first have to give a brief introduction into the bidimensional approach to aspectuality adhered to in this paper (subsection 184.108.40.206.) and the way to account for the interaction between partitives and aspectuality (subsection 220.127.116.11.).
18.104.22.168 Bidimensional Approach to Aspectuality
I adopt the bidimensional approach to aspectuality (inter alia, Bertinetto, 1997; Smith, 1997; Sasse, 2002). There are two cross-cutting dimensions: the viewpoint aspect or aspect1 in Sasse (2002) and actionality (also termed, e.g., as Aktionsart or aspect2 in Sasse, 2002). Crucially, both dimensions operate with boundaries of an event: while actionality is about inherent (Depraetere, 1995) or intrinsic (Sasse, 2002) boundaries, viewpoint aspect is about temporal or established boundaries (Sasse, 2002: 205–6). While actionality refers to the inherent organization of an event such as, e.g., referred to by Vendler’s classes (activity, achievement, accomplishment or state, Vendler, 1957) and their different subsequent modifications as well as by such compositional properties as telicity (telic vs. atelic) or dynamicity, the viewpoint aspect (aspect1 in Sasse, 2002), in turn, refers to the properties that the speaker establishes in a particular utterance and that pertain to such domains as discourse organization or pragmatics and not to the very semantics of the event. Traditionally, one distinguishes between the perfective and imperfective aspect here. The perfective viewpoint aspect is found when a particular event is represented as included in the reference point of the narration, while the imperfective viewpoint aspect entails that it is the reference point that is included into the duration of the event, or, more precisely, the imperfective is often simply non-committal as to its temporal delimitation. In other words, the event is either construed as having boundaries dictated by the discourse organization and not by the internal organization of that event, or no commitments are made as to whether this event is delimited in that particular situation or not. To illustrate this, consider the following examples from English:
The very event to write a letter presupposes a natural or inherent endpoint, namely, that moment when the letter is written and the event cannot continue the same way. The actionality of this event can be defined as telic or culminating (as in Tatevosov and Ivanov, 2009). Thus, all three examples (31)-(33) are culminating as to their actionality. At the same time, only (31) is imperfective as regards the viewpoint aspect, while both (32) and (33) are perfective, because in both cases the event has achieved a boundary and another event began. Note that the imperfective aspect in (31) represents the situation as unbounded, but it does not alter the inherent semantics of the event ‘to write a letter’ which continues to contain a natural or inherent endpoint. The difference between (32) and (33) is that the boundary imposed by the perfective viewpoint aspect (temporal boundary in Depraetere, 1995) coincides with the inherent boundary of the event only in (33) (matching the P-property in Dahl, 1981), while it is prior to the latter in (32). The difference between culminating vs. non-culminating events and perfective vs. imperfective viewpoint aspect will be crucial in understanding the correlations and differences between East Slavic and Baltic, on the one hand, and the Finnic use of the ip(g) in aspectually relevant functions on the other hand.
It has been frequently stated in the literature that there is some correlation between the viewpoint aspect and actionality. For example, telicity often correlates with the perfective, while statives often correlate with the imperfective viewpoint. Indeed, in many cases, the inherent endpoint (the domain of actionality) coincides with the temporal boundary (the domain of viewpoint aspect) – a constellation referred to in Dahl (1981) as P-property, e.g., in he ran a marathon. The typical interpretation of this sentence will be that the temporary boundary resulting from the perfective viewpoint (the situation is seen as a whole) coincides with the inherent endpoint (the entire marathon distance has been accomplished).
22.214.171.124 Aspect, Actionality and the Interaction with the Object
In general, it is a well-known fact that certain verbs allow their objects to determine their actional class (aspect2 property), being themselves ambiguous between accomplishments or activities (Verkuyl’s generalization in 1972), in our terms: between culminating and non-culminating dynamic processes, cf. English:
The lack of a boundary in the object npapples in (34) has to a consequence that the whole vp‘to eat apples’ is an activity such as ‘to work’, e.g., with respect to the commonly used tests as in an hour/for an hour. This is different in (35). Here, the delimited or bounded npthe apples provides for an inherent endpoint, turning the whole vp‘to eat the apples’ into an accomplishment. That is to say, the delimitation of the object np is projected on the verb, or, alternatively, the delimitation of the object np disambiguates the accomplishment reading of the verb to eat, while lack of such a delimitation features the activity reading of this verb (Tenny, 1992: 5). We observe that the quantificational value of the object np is mirrored by the verb and, hence, becoming the value of the whole vp. This phenomenon is well-known under the heading of incremental-theme verbs (cf. Krifka, 1989, term coined in Dowty, 1991). Incremental-theme verbs are typically accomplishment verbs that establish the homomorphic relationship with their object nps which entails that every subpart of the event the verb denotes is unique, and it is coupled with a particular unique subpart of the object np’s referent (Krifka, 1992: 39). Thus, in a sentence like John ate the roll every specific subpart of the roll corresponds to a specific subpart of the event of eating. The object is said to “measure out” the event (Tenny, 1994). From this it naturally follows that, if the incremental theme is bounded, then the event in itself is also bounded.
Yet the case with the ip(g) is not principally different, at least diachronically: the quantificational value of the ip(g)-marked np interacts with the verb by disambiguating its actional and aspectual value in Baltic and Finnic and by adjusting to the actional and aspectual value in Russian/North Russian. Recall that the difference between the former two and the latter is that in the former two language groups the verb is aspectually ambiguous or unmarked, while in the latter it is marked and not ambiguous. The differences between the English examples above and the ip(g) in the languages under investigation are rather in terms of the degree of expansion. While in English this pattern is restricted to incremental-theme verbs only, it has been extended to other verbs in Russian/North Russian, Lithuanian and to a much higher degree in Finnic. I believe that this relation holding between the nominal and the predicate with the incremental theme verbs has been transferred to other predicates and their subjects or objects that are not typical incremental-theme verbs to begin with. To give an example, consider the following example from Lithuanian:
In this example the relationship between the verb and the object is read into a homomorphic one – the subparts/subsets of the subject žmonių ‘people’ can be mapped onto the subevents of the arrivals. The verb requires its ip(g) subject np to have distributive reading while the collective reading is blocked. That is to say, different sets of people correspond to particular arrivals: e.g., first came John and Mary, then came a neighbor and some other people, etc., finally, too many people arrived and the inherent endpoint massive arrival has been achieved. The very event of arrival is not typically homomorphic, neither with singular subjects and collective plurals nor with distributive plurals, since the regular verb to arrive is accomplished with every subpart of such a subject, and there is no inherent endpoint of a higher level comprising different arrivals like in the Lithuanian example above, cf. Several well-known linguists have arrived. Example (36) thus represents a metaphorical extension of the incremental-theme pattern.
Yet how exactly do actionality and viewpoint aspect interact with the ip(g)? The parallelism between the quantification of the event structure and of the nominal has been treated in several works (cf., inter alia, Verkuyl, 1972; Krifka, 1989, 1992; Filip, 1989; Kiparsky, 1998; Borer, 2005 and most recently Champollion, 2010). One of the most influential approaches integrating quantification at the event-structure level and the np quantification is Kiparsky (1998). Kiparsky’s primary concern is to account for the assignment of the partitive case – as opposed to the accusative case – in Finnish. His main claim is that the unboundedness of the vp is the discriminating factor that requires the assignment of the partitive case to the object np, whereas boundedness motivates accusative. According to Kiparsky, a vp is unbounded if the predicate and/or the object is unbounded. There are verbs that are inherently unbounded (such as psych verbs) and, hence, inherently take the partitive case, while there are accomplishment verbs that alternate between partitive and accusative allowing for both bounded and unbounded interpretations. A predicate or np is unbounded “if it is cumulative, divisive and not diverse” (I shall skip the formulaic definitions) (Kiparsky, 1998):
Indeed, this approach has a number of advantages: it coherently accounts for most of the instances of the partitive case assignments in Finnish and, crucially, it provides a unified model for the verbal and nominal quantification as well as for the interaction between the two. Furthermore, it makes an important distinction between (un)boundedness and (a)telicity, which are notoriously used synonymously in the literature (e.g., in Doetjes, 1997).
There are, however, some problems with this account that are relevant here. The predictions that this account makes seem to be counterintuitive with respect to what one might call “restrictedly unbounded” or “weakly bounded” quantities. These are indefinite portions of something: either (i) an np, e.g., headed by a weak quantifier/determiner such as some, or (ii) verbs that refer to only some portion of an action (by means of their morphology, lexically or contextually), e.g., Lithuanian pa-valgyti ‘to eat a little bit/somewhat’. A lack of clear-cut, definite boundaries makes these quantities pattern with unrestricted quantities with respect to cumulativity and divisiveness. At the same time, the presence of some – even if vague – cut-off point unequivocally suggests a bounded interpretation. I will illustrate this in more detail in what follows.
Borer (2005) shows that for Kiparsky’s approach such nps as apples and some apples yield contradictory results with regard to the boundedness value of their vps, contrast (38) vs. (39):
On Kiparsky’s approach in (37) both nps are cumulative and not diverse, while their divisiveness value is dependent upon whether or not the singular atoms are included in the set.
Furthermore, a boundary at the upper edge is explicitly claimed with some apples, because one cannot infinitely add some apples to some apples and get a set that would still fit the description of some apples. Even though, I concede, this boundary is vague and its value is subject to an individual interpretation, it nevertheless exists and is explicitly signaled by the quantifier some. Thus, if one has two apples for some apples and add another some apples, e.g., also two apples then four apples might still fit the description of some apples. Once, however, one has reached, say, 1000 apples the description of some apples will no longer do, at least, not for many speakers. Note that this is not the case with the description apples. There is, thus, a difference between apples and some apples in terms of boundedness, conventional or inherent (Seržant, 2014a, 2014b). Having no distinction between apples and some apples would predict that both vps in (38) and (39) would pattern alike. Crucially, as Borer (2005) correctly points out, only some apples yield a telic interpretation of the event. Any account must therefore discriminate between apples and some apples, cf. Borer (2005).
Moreover, it seems that the same problem also exists with those predicates that morphologically or lexically denote a particular portion/part of an action (Seržant, 2014a, 2014b). In the case of accomplishment verbs, Kiparsky’s model predicts correct results only when this portion exhausts the action completely including the culmination (P-property in Dahl, 1981). However, if just a particular portion/part of an accomplishment event or of an activity event is referred to by the predicate (fully parallel to some apples above), then intuitively this portion must be considered as bounded, but Kiparsky’s model treats it as unbounded. Consider the following example from Kiparsky (1998):
Both readings (i and ii) are parallel in that they both encode an indeterminately quantified action. Crucially, while (i) does not make any commitments as to the temporal boundaries of the event featuring the progressive reading and is indeed unbounded, (ii) does imply such a boundary, even though the latter is an arbitrarily established boundary and not inherent to the very event. Reading (ii) is sometimes neglected in the literature, although it provides the key for understanding correlations between Finnic and Baltic with Slavic. This reading is delimitative (= cessative in the Finnish tradition, cf. Huumo, 2010: 90). Delimitatives signal that the action has taken place for a while and then has stopped for whatever reason without reaching its natural boundary (culmination/telos/inherent endpoint), if such a boundary is implied by the actionality type of the respective verb at all; in the latter case, one is dealing with a certain portion of an activity.17 Now, while reading (i) is perfectly consistent with Kiparsky’s model, reading (ii) suffers from similar problems as the ones mentioned above in relation to the boundedness value of some apples in the nominal domain. The second reading (ii) is thus not unbounded in any non-theory-dependent sense, because it essentially implies certain boundaries. Similar to the case of some apples, these boundaries are not definite.18 The situation in (ii) is not cumulative, because not every partly opening the window ends up with partly opening the window. At some point, the window will be opened completely and the description partly opening the window will simply no longer fit. I suggest below that the discriminating semantic factor in Finnish is not boundedness in Kiparsky’s terms but rather totality (= P-property as defined in Dahl, 1981) as has been claimed in the literature, see most recently Huumo (2010). Those vps that imply totality mark their objects with the accusative. In contrast, in non-total events the object is marked with the partitive. There is some correlation between totality and boundedness1 in that totality (conditionally) implies the positive value of the feature boundedness1 but not necessarily vice versa, (as in the case of delimitatives), since boundedness1 does not automatically imply totality (boundedness2 as per Sasse 2002).
The partitivity account put forward in Filip (1989) or Krifka (1998) faces similar problems. This account crucially relies on the notion of partitivity that is assumed to characterize both verbal aspect such as progressive and mass nouns/bare plurals, given that the incremental-theme relation between the object and the verb holds. In Krifka (1998), the imperfective aspect is obtained by extracting some parts from the whole denotional base of a telic verb. Hence, the parallelism with the nominal partitivity: the imperfective aspect is interpreted here as referring to a part of a telic event in the same way as nominal partitives refer to a part of the np they embed (Krifka, 1998). This account is essentially correct but it does not predict why we get reading (i) and reading (ii) in (40). In other words, a part of an event – in Filip’s sense – may be either bounded/perfective (i.e. be temporally delimited) or unbounded (imperfective). As we will see, this distinction is essential for Lithuanian (and Russian). Thus – as will be argued in subsection 3.2 below – only if the part of the event encoded by an incremental-theme verb is bounded1 the ip(g) may be used, while if it is unbounded1, then only accusative can be used in Lithuanian. To conclude, the accounts in Filip (1989) and Krifka (1989) are not fine-grained enough. I claim that the bidimensional nature of aspectuality must be taken seriously here in order to fix the problem.
This is done by Tatevosov and Ivanov (2009) in a different context. They set two functions (operators in their terms) apart: (a) non-culmination (actionality/aspect2) and (b) perfectivity/imperfectivity (viewpoint aspect/ aspect1). This decompositional approach to verbal partitivity allows a coherent account for the use of the ip(g) in all three branches discussed here. At this juncture, there is a way to capture the similarities and differences between the partitive case in Finnish and the partitive genitive in Lithuanian/Latgalian and Russian/North Russian. The (a) function always creates non-culminating accomplishments in all three branches: Finnic, Baltic and Russian/North Russian, but, crucially, not necessarily imperfective aspect as assumed in Krifka (1998) or progressive as claimed in Filip (1989). Creation of non-culminating accomplishments (that is, basically, derived or secondary activities) is a derivational process pertaining exclusively to the domain of actionality and not to the viewpoint aspect, as Tatevosov and Ivanov (2009: 93–5) correctly maintain. This is also intuitively more likely, since the lack of culmination here is inherent to such vps exactly as the presence of the latter is inherent for the accomplishment vps, independently of the particular position in a discourse. The authors argue that the output of the non-culmination function is the input to the viewpoint aspect (Tatevosov and Ivanov, 2009: 94). Thus, the perfective interpretation is indeed found to be morphologically marked in Russian as well as contextually disambiguated in Bagwalal (Nakh-Daghestanian, Andic) (Tatevosov and Ivanov, 2009: 93–4). In our context, the perfective-viewpoint interpretation is found in the delimitative readings in Finnish such as (ii) in (40) above. On this account, the progressive reading (i) in Finnish (40) – the most prominent reading in the literature – is equally well accommodated: it is non-culminating by the first of Tatevosov and Ivanov’s operators (actionality) and imperfective by the second (viewpoint aspect).
In what follows, I will make the following claims. The data from the languages under investigation presented in subsections 3.10.2–3.10.5 below allow for the following conclusions relevant to language contact: non-culmination marked by the ip(g) is a common feature of all languages under investigation. The differences concern the viewpoint aspect: while the ip(g) in Finnic does not have any impact on the viewpoint aspect interpretation of the vp, the ip(g) of Baltic (Lithuanian and Latgalian) and Russian/North Russian is more restricted in that it allows only for perfective non-culminating accomplishments.
Koptjevskaja-Tamm and Wälchli (2001: 652) argue that aspectual considerations are not relevant for the ip(g) in Lithuanian. Indeed, in contrast to Russian it allows the ip(g) to occur with both telic (41) and atelic predicates (42) and is, hence, independent from the choice of actionality type:
I have argued in detail elsewhere (Seržant, 2014a) that there is a group of verbs in Lithuanian (primarily, incremental-theme verbs) that show strong interaction between the ip(g) case-marked object and their aspectual and actional properties. In what follows I will just recap the main points.
Crucially, with delimitatives (that are overtly marked by means of the prefix pa-) the ip(g) case-marking of the object is the default and accusative is only restrictedly possible. Recall that delimitatives entail that the process had been running for a while and was stopped for whatsoever reason without reaching some natural end, if such an end (telos) is presupposed by the lexical semantics of the verb at all. The typical implication here is that the action could have lasted longer and was not fully exhausted:
Given the homomorphic relation between the verb and the object here, the quantity of the action and the quantity of the np must correlate, i.e. be compatible with each other. The Lithuanian accusative would imply totality on the part of the incremental theme which is blocked by the delimitative marker on the verb. The accusative is only possible if one wishes to emphasize the kind of the np’s referent (cf. Paykin 2014 for a similar phenomenon in Russian). Delimitatives imply that the action stops arbitrarily (typically without reaching its natural endpoint), i.e. some sort of partitivity of events as in Filip (1989).
The ip(g) may also be used with verbs that are explicitly marked as telic (by means of a lexically empty prefix) such as in (44), in which the verb iš-gėrė ‘drank up’ is marked as telic by means of the prefix iš-:
As I have argued in Seržant (2014a), the interpretation of the whole vp in (44) is synonymous to (43) with the verb marked as delimitative – in both cases, at minimum, there is no commitment as to whether there is a culmination. The reason for this is that vandens ‘of water’ is an arbitrary quantity, its boundary may but need not coincide with some natural boundary in the given situation. Thus, if (44) is uttered in the context where there is a glass of water on the table, the entailment of (44) would be that not the whole glass has been emptied, i.e. the event has not been fully exhausted. In turn, if (44) is uttered in a situation in which there is no conventional quantity available in the discourse world, e.g., if that person drank some water from the tap, then naturally there is no inherent boundary whatsoever. In both cases (43) and (44), the interpretation of the whole vp is delimitative and not culminating (= non-committal to a culmination). In order to commit oneself to exhaustiveness, accusative has to be used:
Thus, both utterances (43) and (44) are non-culminating (as opposed to (45)) and both are perfective as to their viewpoint interpretation. Notably, exactly the same is valid for Russian/North Russian perfective telics with the ip(g) (see 3.10.4 below).
In turn, the imperfective reading ‘was eating’ of the verb valgė ‘ate/was eating’ disambiguated by the conjunction kol ‘while’ blocks the ip(g) case marking. Contrast grammatical accusative with the ungrammatical ip(g) in the following example:
Differently from Finnish where the ip(g) can also feature the progressive reading (subsection 3.10.5 below), the ip(g) in Lithuanian is not compatible with this reading at all (except for pure np-internal functions with verbs other than incremental-theme verbs). Seržant (2014a) claims that this is because the value of the ip(g) is indeterminate but bounded, which is why it is compatible with the perfective interpretation of the verb and is incompatible with an imperfective interpretation thereof. Not only is the ip(g) typically ungrammatical with the progressive reading, it is also dispreferred with generic contexts or individual-level interpretations (as per Krifka et al., 1995), cf. (47):
Here, the ip(g) case-marking may be found, but is judged as less acceptable by the native speakers, whereas the accusative case-marking is fully felicitous. I conclude, the ip(g) is generally dispreferred in Lithuanian with the imperfective viewpoint interpretation.
With regard to the Latgalian data I rely mainly on Nau (2014), who extensively discusses the ip(g) of this language. The distribution adheres to the one suggested for Lithuanian above. Recall from example (44) from Lithuanian that telic predicates, when used with the ip(g), become, at minimum, non-committal as to the culmination (i.e. some sort of secondary activities) but always have perfective reading, as evidenced by the event of laying down immediately after the event of drinking the coffee:
The imperfective viewpoint reading of the verb such as progressive induces accusative object marking:
Like the delimitatives in Lithuanian and Russian, the delimitatives in Latgalian (likewise formed by means of the prefix pa-) also select for the ip(g). Thus, the verb pa-sa-dzert ‘to drink somewhat’ selects for the ip(g) only (Bukšs and Plačinskis, 1973: 296; Nau, 2014: 246):
Analogically to the ip(g) in Baltic, the Russian ip(g) when interacting with the aspectual and actional properties of the verb entails a particular, i.e., an indeterminate but bounded (delimited) quantity only (cf. Padučeva, 1998: 80; “unspecified but delimited” in Timberlake, 2004: 319; Seržant 2014b). Differently from Baltic, the Russian verbs are fixed with respect to their viewpoint aspect. Here, it is not only the context as in Baltic, but also the very verb form that is indicative of the viewpoint aspect. Otherwise fully parallel to Baltic, Russian/North Russian generally allow only for the combination of the ip(g) with the perfective aspect, most prominently with the incremental-theme verbs (cf. Jakobson, 1936; Padučeva, 1998), cf. grammatical (51) versus ungrammatical (52):
Again, parallel to Lithuanian, Russian delimitatives are naturally used with ip(g) case-marked objects:
Just as in Lithuanian, Russian does allow for the accusative case assignment with delimitatives:
The accusative case-marking implies either a definite amount of water or, alternatively, it emphasizes the kind of the np’s referent (Paykin 2014). In this case there is no longer a homomorphic relation between the quantity of the verb and that of the object: the verb implies that the action has been carried out for some period of time and then stopped without reaching its end, while the accusative object implies discreteness of the referent.
While the restriction of the ip(g) to exclusively perfective verbs in Standard Russian – although a solid rule – is not without exceptions (cf. Padučeva, 1998), the compatibility of the ip(g) and imperfectives is much higher in the North Russian dialect. Here, a number of imperfective verbs can be used with the ip(g) as well, cf. gret’ ‘to warm’, delat’ ‘to do’, vozit’ ‘to carry’, kosit’ ‘to mow’, and several others (see Kuz’mina, 1993: 30). In this respect North Russian comes closer to Finnic, where the ip(g) encodes just non-culmination while not having any impact on the viewpoint aspect interpretation of the verb.19
Differently from, e.g., Russian where every verb is marked for aspect and actionality, but also from Lithuanian or Latgalian that prefer to mark actionality as well, Finnish is not so explicit here. Most of the verbs are either telic or atelic by virtue of their inherent lexical semantics (cf. Kangasmaa-Minn, 1984; Kiparsky, 1998). There is, however, a group of verbs that allow for different readings. With these verbs, it is only the case-marking of the object that provides for disambiguation, cf. the non-culminating readings in (55)-(57), disambiguated only by the case-marking on the object (from Huumo, 2010: 93 and Kiparsky, 1998):
As can be observed from (57) both perfective non-culminating (ii, iii) and imperfective non-culminating (i) readings are available in Finnish. This holds for other Finnic languages as well. Note that Finnic exhibits the greatest progress, using the clause-related functions of the ip(g) to disambiguate the aspectuality of the verb: while incremental-theme accomplishment verbs such as to read represent rather an incipient state of affairs for this function of the ip(g) due to their natural homomorphic relationship between the quantity of the object and the quantity of the action, non-incremental-theme accomplishments such as to open attest further development. Here, the quantity coded by the ip(g) does not quantify the object’s referent but the verbal action (it is the same quantity of the window at any time of the process of opening). Another frequently cited example is the achievement verb to shoot, such as Fin. ampua, which is construed as an accomplishment verb to shoot to death with a preparational phase of shooting (cf. German er-schießen ‘to shoot to death’):
The ip(g) signals a non-culminating event. Note that under the meaning ‘to shoot to death’ the culmination (inherent endpoint) is not the shot but rather the death of the object’s referent.
Crucially, the development of the aspectual function must be attributed to the Balto-Finnic period (after the split from Finno-Ugrian), as argued in Itkonen (1972: 188). This suggests that the aspectual or rather actionality function of the Finnic ip(g) is a later, inner-Finnic innovation.
3.10.6 Summary: Aspect and the ip(g) in Finnic, Russian/North Russian and Baltic
In addition to the differences between the aspectual/actional make-up of the verbal system, other differences between the languages concerned are due to the semantics of the accusative case with the relevant verbs in every language. It is the default case in Russian, default but not indeterminate bounded in Lithuanian and only determinate in Finnish. Crucially, these differences influence the patterns found in the languages concerned, but they are not indicative of differences in the functional load of the very category of the ip(g). The following table summarizes the results from 3.10.1–3.10.5:
As it can be observed from the table above, in Lithuanian the aspectually relevant ip(g) is sensitive to both boundary types – be it viewpoint aspect boundedness (boundedness1 in Sasse, 2001) or actionality boundedness (boundedness2 in Sasse, 2001), while the Finnish partitive is only sensitive to actionality boundedness (boundedness2 in Sasse, 2001), i.e. totality (as per Huumo, 2010). The common feature is thus that both the Finnic partitive case and the Baltic and Slavic partitive genitive encode non-culmination on the actionality level, i.e. lack of commitments as to boundedness2 in terms of Sasse (2001), while differences mainly pertain to the compatibility with the imperfective viewpoint aspect. I summarize:
If one puts aside the differences pertinent to the semantics of the accusative case in each language, it can be observed that the ip(g) interacts with aspectuality in quite similar ways in the three languages, the only difference being the complete loss of the imperfective (non-culminating) option in Standard Russian and Baltic but not in North Russian (Seržant, 2014b), cf. Table 7.
The interaction between the ip(g) and the domain of aspectuality is constrained by the organization of the latter in the respective languages: while viewpoint aspect plays an important role in Russian and is to some extent relevant in Baltic (although not marked on the verb), it is grammatically irrelevant in Finnish. At the same time, crucial for all three language branches is the domain of actionality. Here, culminating vs. non-culminating events tend to be overtly distinguished in Russian and Baltic, as well as in Estonian with its detached bounders (Metslang, 2001; Koptjevskaja-Tamm and Wälchli, 2001). In Finnish, on the other hand, a covert distinction must be assumed, which is then highlighted by the respective object marking.
Furthermore, in all three languages, there is a development to extend the interaction of the ip(g) with the aspectual and actional properties of the verb from incremental-theme verbs only to other accomplishment and achievement verbs, whereby the homomorphic relation between the event and the object referent gradually gets lost. This development has, of course, progressed most in the Finnic languages where nearly every accomplishment and some achievement verbs participate in the accvs. ip(g) alternation relevant to the aspectual/actional interpretation of the verb. North Russian takes the second position here followed by Lithuanian, cf. the respective examples in subsection 3.11 and 3.12. It is thus natural to assume that the Finnic languages constitute the hotbed of this particular property. This conclusion is all the more likely given the geographical distribution of this development in which the North Russian dialect represents the second most developed situation with the high number of non-incremental-theme accomplishments and achievements allowing the interaction between the ip(g)-marked objects and the aspectual interpretation of the vp (see the next subsection). The view expressed, for example, in Larsson (2001) that the functional profile of the partitive case in the Finnic languages is the result of Baltic influence, is highly implausible for this domain.
3.11 Event Measure (P11)
The implicit quantifier induced by the ip(g) case-marking on the object is interpreted – if no other quantifiers are present in the clause, i.e. by default, – as having an indeterminate value. The implicit quantifier (with an indeterminate but bounded value, see Seržant, 2014a, 2014b) measures the extent to which the event encoded by the predicate has/will have proceeded towards its inherent endpoint. At the same time, the object itself remains holistically affected throughout the event, and there is no (np-internal) partitivity with this reading, cf. the examples:
The extension of the semantic scope of the implicit quantifier of the ip(g) from its original host (np) must be considered as a purely Eastern Circum-Baltic development not inherited from the respective ancestor languages. This is the third and final step in the development (cf. subsection 3.7 above). There are no analogical examples from the ancient Indo-European languages. They do not seem to be attested in Old Russian either. While sentences like (61) and (62) do not have an unequivocal correlate in Baltic, the languages of this branch do attest the event measuring function emancipated from np-internal functions with the verbs of transfer, see P12 below.
3.12 Temporal Transfer (P12)
The temporal transfer (temporally restricted usage in Koptjevskaja-Tamm and Wälchli, 2001: 654) function of the ip(g) is yet another property common to the Eastern part of the Circum-Baltic area including Polish (Larsson, 1983; Holvoet, 1991: 110; Koptjevskaja-Tamm and Wälchli, 2001: 654). It is used with the verbs of transfer replacing canonical object marking and implies that the result of the transfer will last only for a short period of time:
The regular, canonical accusative case-marking of the objects in (65)–(67) would not induce the implication ‘for a moment/for a while’. The verbs of transfer are typically achievement verbs. The indeterminate quantifier cannot measure the event encoded by an achievement verb, because these verbs do not presuppose a (preparational) phase that could be measured. In turn, the resultant state, the after-effects is the only phase with these verbs that does have the potential to be measured. This is why the implicit quantifier applies to the stage after the transfer event has already taken place, providing for indefinite boundaries at this stage (Seržant 2014a: 286, 2014b: 288–9).
This function is not attested in the ancient Indo-European languages and cannot be shown to be an inheritance in Finnic either. At the same time, it is attested throughout the area of concern: Late Middle Russian (the first attestation is from 17 c., Krys’ko, 1997: 200), Early Modern Russian,22 North Russian (inter alia, Kuz’mina, 1993: 31–2; Markova, 1988: 98), Belarusian (dabm Kamentary, 756; Lopatina, 1998: 234, 236), Finnish (Larsson, 1983; Kiparsky, 1998), Eastern Lithuanian (Ambrazas et al., 1976: 24–5) and dialectal Polish of the area (polszczyzna kresowa, Adamovičiūtė and Čekman, 1984: 10). The temporal-transfer function has been present originally in a broader area of Russian including subdialects not immediately belonging to the Circum-Baltic area. Thus it is still found in some South Russian subdialects. However, it is most frequent on the token frequency in the area of North Russian (Kuz’mina, 1993: 32). It seems possible that this is also the original area of its use from whence it spread to other dialects of East Slavic and then disappeared in these dialects.
Like the previous properties, this property is a recent and common innovation. Notably, the ip(g) represents here a typologically rare case of a quantifier (Tatevosov, 2002: 56) that is formally realized as internal to the respective np constituent, but semantically applies to the quantification of the whole event; it is thus formally a D(eterminer)- but functionally an A(dverb)-quantifier. Such a quantifier seems to be extremely rare typologically, cf. the overview in Corbett (1994: 202, 2000: 251) where such a quantifier is said to be unattested (similarly Tatevosov, 2002: 56).
3.13 Negation and the ip(g) (P13)
The predicate negation does not trigger ip(g) in the ancient Indo-European languages in terms of a grammaticalized formal dependence; the occurrence of the ip(g) in the context of negation is rare and encodes the part-of-relation:
Thus, (68) does not mean ‘they did not die’ but rather ‘none of them died’ (lit. ‘any of them did not die’), which has somewhat different semantics than the former translation in terms of a different emphasis and implications.
This changes crucially in Baltic and (East) Slavic, where the ip(g) becomes more and more subject to the grammatical rule that requires ip(g) case-marking on the verb’s internal argument under negation in purely syntactic terms. Kuryłowicz (1971) was the first to historically relate the ip(g) and the genitive of negation. There are two developmental stages here. At the first stage, the genitive with negation had partitive function and was an ip(g). The motivation for the ip(g) in the context of negation was emphatic as in (68) above. The emphatic use was conventionalized at the second stage, gradually turning into a grammatical rule (cf. the double negation in French pas, jamais, etc.). Once the use of the ip(g) became regular under the negation, other nps (not only mass nouns and plurals) could occur here. It is fully grammaticalized in Lithuanian, Polish and Old Russian, fully productive in Standard Russian and in Latgalian, extinct though with some traces in Latvian. Thus, Brown (1999) argues for Standard Russian that the genitive must be syntactically licensed by sentential negation, even though it is not always semantically interpreted as under the scope of a real semantic negation (cf. also Brown and Franks, 1997; Partee, 1998: 298). This rule can thus be considered as representing grammaticalization of an originally semantically- and pragmatically-driven phenomenon in Baltic and Slavic, the obligatoriness as well as semantic bleaching of the ip(g) under negation being symptomatic (cf., inter alia, Heine et al., 1991: 2; Traugott, 2003: 645). Interestingly, diachronically this is presumably the first case of the interaction of the quantificational function of the ip(g) with the predicate’s quantification (namely the negated universal quantifier), i.e. its np-external usage. Note also that this is the only case in which ip(g)-marked subjects show up with the np-external reading.
Analogically, Finnish and Estonian require the ip(g) case-marking on the object wherever there is a negation. Livonian is less strict here and also allows for accusative case-marking (Tveite, 2004: 52) which is most probably secondary.23
The ip(g)-under-negation is not an inherited feature in the Finnic languages exactly as in Baltic and Slavic, which is evinced by the comparative evidence with other languages such as Mordvin (Larsson, 1983: 97). This even leads Larsson (1983) to the assumption that the ip(g)-under-negation rule is a purely Baltic feature copied later onto the Finnic languages. While this cannot be sufficiently argued for, in view of the lack of grammaticalization in the proto-ie ancestor language, only a common development of the ip(g)-under-negation rule and its grammaticalization is a reasonable alternative. Moreover, it seems that Mordvin does attest instances of occasional ablative object marking with verbs under negation (Lytkin et al., 1978: 109), which means that an independent outset of this phenomenon may be assumed for Finnic as well. While the ip(g)-under-negation rule has been argued to be subject to language contact in Koptjevskaja-Tamm and Wälchli (2001: 652–3), I add that this rule is a common Eastern Circum-Baltic innovation, not inherited from any of the ancestor languages.
Common to many languages of the area is a weakening of the this rule. This can be observed by comparing Old Russian with Standard Russian, which has introduced the acc marking here competing with the ip(g). Among Baltic languages, it is Lithuanian that is most conservative. While Standard Latvian has almost lost the genitive-under-negation rule, retaining it as an option with the subjects of existential predicates only, Latgalian takes an intermediate position between Latvian, on the one hand, and Lithuanian, on the other. It attests – although optional – a nevertheless frequent use of the genitive in the context of negation (Bukšs and Placinskis, 1973: 296), “roughly as frequent as accusative” (Nau, 2014: 224). With the negative indefinite pronouns such as nikas ‘nothing, nobody’ the use of the genitive is even mandatory here. There is a general tendency towards canonical object (acc) and subject (nom) marking alongside some more conservative varieties, which is assumedly due to the influence of the neighbouring languages: (a) Latvian with only some few traces of this rule (with the subjects of existential predicates), (b) Russian and (c) Belarusian dialects with differential marking (Nau, 2014: 224) constrained by scope specificity and other parameters (inter alia, Babby, 1978; Paducheva, 1998; Partee, 2008).
Restrictedly subjects may also turn into ip(g) if their predicate is negated. This is most frequently found with the subject of existential clauses in the languages under investigation (Koptjevskaja-Tamm and Wälchli, 2001: 657). However, as M. Vilkuna (apudKoptjevskaja-Tamm and Wälchli, 2001: 657) notes, subjects of several unergative verbs may also turn into ip(g) if the overall reading allows for an existential interpretation:
The same marginally holds for Russian:
The question about the source language for this property is somewhat speculative. Since there is some evidence that the ip(g) has been used with negation in the ancient Indo-European languages in order to signal some emphasis, cf. non-emphatic ‘they didn’t die’ and the emphatic ‘none of them died’ attested in (68). I assume that these emphatic uses of the ip(g) in Proto-Indo-European might have provided the source for the ip(g)-under-negation rule in conservative Baltic and Old Russian which, subsequently, has been copied into Finnic. Later processes related to the merger of the nom-acc and nom-ip(g) patterns in Russian, Latvian and Livonian have obscured this, Finnish and Lithuanian being most conservative in this respect.
4 Formal Properties
4.1 Subject Properties in the Subject Position
The formal properties of the ip(g) of Baltic and East Slavic correspond very much to those of the ip(g) in Finnic.25 At the same time, they do differ considerably from the ip(g) in the ancestor language, namely, Proto-Indo-European. It is thus likely to assume that the syntactic properties of the ip(g) have been accommodated to those of the Finnic ip(g), or, alternatively, both represent developments leading to the creation of a common pattern.
4.1.1 From no Restrictions on the Syntactic Position to Overriding Structural Cases Only (P14)
As stated in, inter alia, Schwyzer and Debrunner (1950: 101), Kuryłowicz (1964: 184), Luraghi (2003: 60), Bauer (2007: 133–4) or Seržant (2012b), there has been no restriction for the ip(g) as to which syntactic position in the surface structure it may occupy. Thus, the ip(g) does not only override structural case in the ancient Indo-European languages, but also datives governed by a preposition (Conti and Luraghi, 2010), indirect object datives, adverbial datives (Kuryłowicz, 1964: 184) and non-argumental accusatives (accusativus graecus/relationis). Furthermore, it also overrode the accusative case of controlled subjects in the accusativus cum infinitivo construction in the ancient ie languages (Seržant, 2012b). Although prototypical agents and patients marked by the ip(g) were an extremely rare option, generally there has been no restriction as to the semantic role the ip(g) may mark. This has changed into modern Baltic and Slavic. Here – exactly as in Finnic – the ip(g) may override structural cases only. The only exception might be the dative subject of the embedded dativus-absolutus-like construction of Lithuanian, which may be overridden by the ip(g); notably, the dative case is a structural case here (Arkadiev, 2011b).
4.1.2 No Semantic (ad sensum) Verbal Agreement (P15)
In what follows I distinguish between the syntactic and semantic agreement following Corbett (2006: 155). It has been argued in Seržant (2012b; to appear-c) that the ip(g), while being in the subject position, triggers semantically-based verbal agreement in the ancient Indo-European languages and, hence, in the proto-language. This means that the controller and the target are morphologically not covariant as to their number (and person) values, while it is the logical number of the controller’s referent that is morphologically rendered by the verb. Here, the verb copies the logical singular (72) vs. plural (71) number of the logically implicit subset, not of the formal, explicit number of the np which is plural in both cases:
None of the Baltic, Slavic or Finnic languages, in turn, attests semantic agreement with the ip(g) marked subject. Typically, there is only the default, singular (neuter) agreement with no impact from the semantics of the ip(g) referent.
Moreover, as regards Finnic, the agreement between the ip(g) subject and the verb cannot have existed at an earlier stage, because partitives were not originally allowed in the subject position, as clearly evidenced by Mordvin, which represents the more original state of affairs with regard to the partitive (Larjavaara, 1991: 378; Kiparsky, 1998). In this language, the use of the partitive (ablative) is restricted to some incremental-theme and intensional verbs only.
This property on its own does not provide strong evidence in favour of language contact, since any kind of loss or simplification may take place independently and does not require an external motivation. I find it nevertheless important to bring this property into the context of the discussion pursued here. I believe that the cumulative evidence of a number of parallel – although trivial – changes in some neighbouring languages, resulting in the same pattern across these languages, proves that the overall convergence is hardly accidental.
4.1.3 Rise of the Formal (ad formam) Agreement (P16)
V.I. Trubinskij (in Meščerskij, 1972: 211) was the first to draw attention to the ip(g) that behaves like a “subject-like object” (Rus. “v roli subjektnogo dopolnenija”) in triggering number agreement on the verb, cf. Meščerskij (1972: 211):
The rise of the formal number agreement is a very recent innovation and is assumedly motivated by a gradual acquisition of canonical subjecthood here (Seržant, 2013: 346). The formal number agreement is also found in the Finnic language Veps (Lytkin et al., 1975: 108):
Differently from the semantic agreement found in the ancient ie languages (4.1.2), in this recent development the syntactic agreement, i.e. the agreement ad formam is found: the morphologically coded number value of the controller np’s is copied on the verb. The agreement is nevertheless non-canonical in Corbett’s terms (2006) given the “wrong” syntactic condition – only nominative subject nps control agreement elsewhere in these languages. This is a local common development, but it is difficult to determine the donor language here.
4.1.4 Almost no Coordination with Otherwise Case-Marked nps, except for some Structurally Marked nps (P17)
In the ancient ie languages the ip(g) could be coordinated with any case-marked np (including non-structural nps) with no regard to their syntactic position (Seržant, 2012b). This has been partly preserved in Old Russian, cf. (77) where the regular instrumental object of the verb vladyčestvovati ‘govern’ is coordinated with the ip(g)-marked np:
This crucially changes in Baltic and Slavic, in which languages the ip(g) may be coordinated only with other genitives and occasionally with structural nominatives and accusatives. Thus, sentences such as (77) are ungrammatical in Modern Russian or in Baltic.
There is another property of the ip(g) that makes it pattern with a structural case, namely, when the ip(g) marks the subject. The ip(g) in the subject position is partly endowed with the same syntactic subjecthood properties in the languages of concern, as the respective (structural) nominative subjects. Thus, the deletion of a coreferential pro is possible in Finnish (Sands and Campbell, 2001: 267–8) and in Lithuanian:
The verb pavogti ‘to steal’ is a transitive verb and assigns the nominative case to its subject. The subject of (78) is, however, left out on identity with the ip(g) subject of the preceding clause ‘several people’. Unfortunately, Lithuanian is a pro-drop language and examples such as (78) do not provide a strong argument in favour of the subjecthood. In fact, a better test for subjecthood with the ip(g) in Lithuanian is its ability to be the subject of the adverbial “dativus absolutus” subclause. Thus, the first existential clause in (79) can be transformed into the absolutive construction in the adverbial clause of (80):
The grammaticality of (80) shows that the ip(g) of Lithuanian does have some subject-like behavior, because only true subjects can form the dativus absolutus construction in Lithuanian.
4.2. Morphological Distinction between (Pseudo-)Partitivity and other Functions (e.g., such as possessiveness) (P18)
Finnic languages formally distinguish between the possessive and (pseudo-)partitive relations by means of two different sets of markers: the genitive case encodes possessiveness (sensu lato) while the partitive case encodes various functions (diachronically or synchronically) related to (pseudo-)partitivity. In East Slavic, where both functions are expressed by the genitive case, there has been a tendency to morphologically discriminate between these two meanings along the Finnic pattern. Thus, there arose a new, dedicated partitive ending for the singular of the o-stems, namely, -u as opposed to the default genitive ending -a for the same declension. Both the endings -u and -a are equally possible and interchangeable in a (pseudo-)partitivity-related context, cf. (82), but, in a possessive-like context, only the regular genitive ending -a is allowed, cf. (81) (inter alia, Breu, 1994):
I gloss the dedicated partitive ending here as part and not as gen. The dedicated partitive ending is found in the singular of masculine o-stems only, representing a rather less entrenched and highly lexically restricted pattern. It has been claimed in the literature that this new case is not a full-fledged partitive, since it may also be governed by a preposition within a context that is atypical for a (pseudo-)partitive-related function (Daniel 2014), cf.:
While this is essentially correct, this does not change the fact that this new case does pattern with the partitive case in Finnic. In these languages, there are several pre- and postpositions that govern the partitive case (Karlsson, 1987: 85). This fact only reinforces the correlation between this new case-marker with the Finnic partitive case. Moreover, crucial for the language contact account, Breu (1994) points out that it is only Russian/North Russian that has “recycled” the old u-genitive of the distinct u-declension in terms of a new case, while other Slavic languages (including closely related Ukrainian and Belarusian) use the old genitive ending for other purposes. Furthermore, I have argued elsewhere that there are fewer lexical restrictions as to which nouns may take the partitive ending in North Russian than in the colloquial Standard Russian (Seržant 2014b) which, in turn, provides additional evidence for a replicated pattern.
While convergence effects in the use of the partitive genitive in Baltic and Russian and the partitive case in Finnic are beyond any doubt (cf., inter alia,Larsson, 1983, 2001; Koptjevskaja-Tamm, 2001; Koptjevskaja-Tamm and Wälchli, 2001), the exact development and adaptation of the ip(g) in the area was yet not clear. Moreover, as the preceding discussion shows, the question about the donor and recipient language cannot be answered in a straightforward way cumulatively for all properties. Other difficulties are inheritance and typological frequency: this category is an inherited one in all three language branches and the incipient functions of this category are not infrequent cross-linguistically.
In order to provide solid evidence in favour of language contact, I have tried to individualize the category against the typological and genetic background. While such individualization is already given with typologically quirky features which are therefore generally easier to argue for language contact as the ultimate source for convergence, typologically less infrequent categories must be treated at a more fine-grained level, i.e. must be individualized, in order to meet the criterion of typological idiosyncracy or quirkiness necessary for excluding potentially accidental correlations. For this purpose I have selected 18 properties that individualize the ip(g) on the typological background (if compared with other independent partitive expressions, e.g. such as the partitive de in French).
I did not go into the discussion about how exactly the process of the contact-induced changes in grammatical properties should be modeled (cf., inter alia,Heine and Kuteva, 2005; Johanson, 2008: 64ff). I assume that the process of copying properties – even if not being a diachronic process per se (Johanson, 2008) – might involve a series of copying and adjustments across languages of concern, thus producing an impression of a common diachronic process, e.g. the grammaticalization of the negation property (subsection 3.13, P13).
I analyzed a number of semantic (section 3) and formal (section 4) properties of the ip(g) in Finnic, East Slavic and Baltic languages. Wherever possible, historical-comparative evidence from genetically related ancient and conservative languages outside the area has been adduced in order to establish the values of the properties in the respective proto-languages. The Proto-Indo-European ip(g) of which the Slavic and Baltic ip(g) are the immediate descendants has been quite exhaustively discussed in the literature (inter alia,Bauer, 2007; Dahl, 2009; Nachmanson, 1942; Napoli, 2010; Seržant, 2012a, 2012b, to appear-c; Schwyzer and Debrunner, 1950; Luraghi, 2003: 60ff). For the ip(g) of Proto-Finnic, the data from Mordvin – a non-Finnic Finno-Ugric language – provide a more archaic state of affairs with respect to the ip(g), and, hence, can and have been used for diachronic considerations (Kiparsky, 1998).
In order to assume contact-induced development with regard to a particular property two criteria have to be met: (i) the present makeup of the property or the property itself must be absent from the ancestor language, and (ii) the new makeup must not represent the result from a frequently recurrent development or such a trivial development as simple loss (cf., inter alia, Koptjevskaja-Tamm and Wälchli, 2001: 628). While (i) is uncontroversial, (ii) can be modified insofar as saying that one particular change – if representing a frequently recurrent development – cannot be used as an argument in favor of language contact, but the cumulative evidence of a number of such changes affecting one and the same category seems to represent a stronger argument. To give an example, there are several losses: (P5) loss of the ability to alternate between accusative and partitive (genitive) with personal pronouns, (P6) loss of the value ‘one’, (P15) loss of semantic verbal agreement. Each of these losses on its own is a weak argument for assuming language contact. In turn, cumulatively, they indicate a development towards a common pattern: only those properties have been lost in Baltic and Russian that have been lost or never existed in Finnic, while, in turn, other properties that do exist in Finnic have not been lost. The cumulative evidence is quite strong, in my eyes, to corroborate the claim of language contact with regard to these properties as well. To put it differently, it would be counterintuitive to assume that the language simply accidentally undergoes only those typologically trivial processes (such as loss) with this category which transform this category towards its functional correlate in the other languages of the area.
Furthermore, the very category of the ip(g) is typologically marked for several reasons. Thus, the typical function of case is to encode “the type of the relationships the dependent nouns bear to their heads” (Blake, 1994: 1–2). Crucially, none of the denotational properties discussed so far would fit this typological generalization, revealing the category of the ip(g) as typologically non-trivial. Other properties, such as the discrepancy between the syntactic position of the implicit quantifier and its domain of application, are also typologically striking: the ip(g) represents here a typologically rare case of a quantifier that is formally realized as internal to the respective np constituent, but applies to the quantification of the whole event; it is thus formally a determiner-type, but functionally an adverb-type quantifier. Other properties, such as the formal agreement acquired by some North Russian dialects and Veps, the ability to coordinate with otherwise case-marked nps, etc. contribute to the claim that the ip(g) stands out on the typological background.
In the following Tables (Table 8 and Table 9) I summarize the results from the discussion of each particular property:
As can be observed, one finds a number of correspondences across the languages of concern. Even though the ip(g) is an inherited category in Baltic and Slavic, most of the properties discussed above are not attested in the ancient ie languages and are thus likely to be a recent innovation of Baltic and Slavic. This, however, should not imply that Finnic must necessarily be the source language. It is equally possible that recent innovations may represent historical developments in the area with no clear-cut donor language, but nevertheless lead to a new common pattern as a sort of adjustment of the inherited categories to each other. Thus, the necessary precondition for identifying the donor language is that the change in question must be anterior to the parallel change in the other (target) language. Here, only the data for the property P17 allow it to be considered to be copied (to some extent) from Finnic (source) into Russian (target).
Indeed, most of the properties not inherited from Proto-ie in Baltic and Slavic – as far as I can judge from the data available – are not inherited in Finnic either. This means that these properties were created relatively recently, quite long after the split of the Finnic branch from the Finno-Ugric family. The majority of these properties were created when Baltic, Finnic and Slavic had tied contacts, even if not simultaneously and with different hotbeds. In any event, it does not seem to be the case that the major role in developing the category of the ip(g) to its modern make-up is mainly due to a Baltic influence on Finnic, as is assumed in Larsson (2001). With the interaction of the ip(g) with aspectuality this is arguably exactly the other way around: all Finnic languages show the widest range of verbs allowing for this interaction, while Baltic is the least developed here. According to the hypothesis suggested in this paper, namely, that this interaction starts out from incremental-theme verbs and gradually spreads to non-incremental-theme verbs, the Finnic languages exhibit the highest degree of development. They must constitute the hotbed here for the whole area. This is also suggested by the geographical distribution: the North Russian dialect has progressed much further here than did Standard Russian (i.e. central Russian), the former being geographically closer to Finnic and having Finnic as a substrate. It also preserves a number of lexicalized usages of the ip(g) that are no longer found in Standard Russian.
While the creation of the major part of the properties must be explained as the result of intensive language contact between Baltic, Finnic and Slavic, there are some correspondences that may equally reveal parallel, but independent developments along frequent and recurrent paths. Or, to put it differently, there is no additional evidence at my disposal that would allow a decision in favour of language contact vs. an independent parallel development. Consider P3. The rise of the pseudo-partitive function (P3) is a frequent development attested in many languages, cf. the so-called partitive article in French or Italian, faded partitives in Dutch (inter alia, de Hoop, 2003), pseudo-partitives in Turkish (Selkirk, 1977), etc. This development, as mentioned in Subsection 3.3., has to do with the loosening of the selectional restrictions on the np embedded under the partitive expression from originally definite nps only to the inclusion of indefinite nps as well. This abandonment of lexical-aspectual restrictions represents a development from more specific or restricted contexts to more general ones found in different kinds of grammaticalization processes (Bybee, 2003: 605). The very beginning of this process can even be found in the ancient ie languages such as Ancient Greek (Kuryłowicz 1971; Seržant, 2012a). It seems, thus, that the property P3 need not necessarily be accounted for in terms of contact-induced development.
While there are properties that are independent from each other implicationally, other properties are not. For example, the following properties are mutually interrelated: P1+P2, P1+P3, P3+P4, P6+P15, P7–P12 or P14+P17. Although each implicationally dependent property contributes weak evidence on its own, taken together as cumulative evidence they represent a strong argument in favour of contact-induced correlations; cf. Seržant (2010) on complex correlations as an argument in favour of language contact. I summarize the judgements on every particular property in Table 9 below:
Other properties can be added. Thus, it is common for the partitive case in Finnic and the partitive genitive case in Lithuanian and Russian to mark objects in intensional contexts, i.e. contexts that do not refer to real-world situations (with “world-creating” predicates). One of these contexts is possibly the so-called genitive of purpose, which is widespread in North Russian, Lithuanian (Seržant, 2014a: 291, 2014b: 298–301), Karelian (Fedotova, 1990: 74–5) or Livonian (Tveite, 2004).
I have argued that North Russian groups with Finnic to a larger extent than do the other languages. Many properties found in North Russian can also be found in West Russian (the Pskov Group) although the dialectological data are notoriously scarce and scattered across the literature, which makes it difficult to assess the picture. Thus, there are properties that are found only in North Russian outside Finnic, for example, the verbal agreement ad formam. I have mentioned that some North Russian subdialects show verbal agreement with the ip(g) marked subjects – a phenomenon that is also found in Veps, but is not attested outside this small subarea (Tuomas Huumo, p.c., Fedor Rozhanskiy, p.c.). Latgalian and Lithuanian (but also Belarusian not taken into account here), on the one hand, and Standard Russian, on the other, each show different kinds of deviations from the North-Russian and Finnic patterns. These deviations pertain not only to the functional domain, but also to selectional restrictions. In my data, many constructions that allow for the ip(g) in North Russian do not allow for the ip(g) in Baltic or Standard Russian, but not vice versa. The ip(g) is most productive in Finnic, where it became to some extent the unmarked option.
Despite somewhat closer relationships between North Russian and Finnic, typically for the category and the area of concern, many properties have their own hotbeds and their own areas of distribution. Thus, as regards property P5, it is clear that Russian/North Russian constitute the hotbed here, while South Finnic languages and Baltic Latgalian follow only to a small extent. Furthermore, such isoglosses as the exclusiveness of the bounded reading with the ip(g) in interaction with the aspectual and actional properties of the verb cut across this area. Here, Standard Russian, Lithuanian and Latgalian pattern together while Finnic differs by also allowing the imperfective (e.g. progressive) reading here, and North Russian takes an intermediate position by clearly preferring perfective verbs. Other isoglosses such as the prefixal quantifier, originally copied from Slavic, intervene with the ip(g) in a quite tricky manner, grouping Baltic and Slavic together as opposed to Finnic.
This paper has benefited greatly from the help provided by (alphabetically) Tuomas Huumo (Turku), Maria Koptjevskaja-Tamm (Stockholm), Liina Lindström (Tartu), Fedor Rozhanskiy (Tartu/Moscow) and the four anonymous reviewers. I cordially thank the editor of the jlc Robert Nicolaï for the work that he invested. All disclaimers apply.
abl – ablative, acc – accusative, act – active, adv – adverb, all – allative, aor – aorist, ip(g) – independent partitive genitive, dat – dative, delim – delimitative, ess – essive, excl – exclamative, fem – feminine, fut – future, gen – genitive, ines – inessive, inf – infinitive, instr – instrumental, invar – invariant, impf – imperfect, impv – imperative, ipfv – imperfective, ie – Indo-European, ip(g) – independent partitive (genitive) case, loc – locative, m – masculine, neg – negation, n – neuter, nom – nominative, np – nominal phrase, partc – participle, pass – passive, pl – plural, poss – possessive, prfv – perfective, pron – pronoun, prs – present, prt – particle, pp – prepositional phrase, ppp – passive past participle, pst – past, quant – quantifier, rfl – reflexive, sg – singular, subj – subject, telic – telic, tr – translative, voc – vocative, vp – verbal phrase.
Adamovičiūtė and Čekman 1984. Irena El’žbeta Adomavičjute, Valerij N. Čekman. Grammatičeskie lituanizmy v pol’skix periferijnyx govorax belorussko-litovskogo pogranič’ja. In RegierJanusz and WereniczWiaczesław (eds.) Studia nad polszczyzną kresową. Tom iiI 7–22. Wrocław / Warszawa / Kraków: Zakład narodowy im. ossolińskich. Wydawnictwo Polskiej Akademii Nauk.
ArkadievPeter M.2012. Participial complementation in Lithuanian. In GastVolker and DiesselHolger (eds.) Clause Linkage in Cross-Linguistic Perspective: Data-Driven Approaches to Cross-Clausal Syntax (Trends in Linguistics. Studies and Monographs) 285–334. Berlin / New York: Mouton de Gruyter.
BauerBrigitte. 2007. The definite article in Indo-European: emergence of a new grammatical category? In StarkElisabethLeissElisabeth and WernerAbraham (eds.) Nominal Determination: Typology Context Constraints and Historical Emergence103–139. Amsterdam: John Benjamins.
BorschevVladimirPaduchevaElena V.ParteeBarbara H.YakovG. Testelets and YanovichIgor2007. Russian Genitives, Non-Referentiality, and the Property-Type Hypothesis. In AntonenkoAndreiBailynJohn F. and BethinChristina (eds.) Formal Approaches to Slavic Linguistics: The Stony Brook Meeting 2007 (fasl 16). Ann Arbor: Michigan Slavic Publications48–67. url: https://udrive.oit.umass.edu/partee/fasl16FinalForPrinting.pdf.
ContiLuz and LuraghiSilvia. 2010. The Ancient Greek partitive genitive in typological. A paper presented at the international workshop “Partitives” organized by Silvia Luraghi and Tuomas Huumo hosted at the 40 of the sle in Vilnius 4 and 5 September 2010.
CorbettGreville. 1994. Systems of grammatical number in Slavonic, Slavonic and East European Review 72 (2): 201–217. A revised version of Systems of grammatical number in Slavonic In GilDavid (ed.) Studies in Number and Quantification. European Science Foundation Programme in Language Typology: Theme 7 Noun Phrase Structure: Working Paper no. vii/19: 1–17. Konstanz.
DahlEystein. 2009. Some semantic and pragmatic aspects of object alternation in Early Vedic. In BarðdalJóhanna and ChelliahShobhanna (eds.) The Role of Semantic Pragmatic and Discourse Factors in the Development of Case23–38. Amsterdam / Philadelphia: John Benjamins.
FilipHanna. 1989. Aspectual properties of the an-construction in German. In AbrahamWerner and JanssenTheo (eds.) Tempus – Aspekt - Modus. Die lexikalischen und grammatischen Formen in den germanischen Sprachen (Tense – Aspect - Mood. Lexical and Grammatical Forms in the Germanic Languages). Linguistische Arbeiten Band 237 259–292. Tübingen: Max Niemeyer Verlag.
von HeusingerKlaus and KaiserGeorg A.. 2007. Differential object marking and the lexical semantics of verbs in Spanish. In KaiserGeorg A. and LeonettiManuel (eds.) Proceedings of the Workshop “Definiteness Specificity and Animacy in Ibero-Romance Languages”85–110. Universität Konstanz (Fachbereich Sprachwissenschaft. Arbeitspapier 122).
Koptjevskaja-TammMaria. 2001. A piece of the cake and a cup of tea: partitive and pseudo-partitive nominal constructions in the Circum-Baltic languages. In DahlÖsten and MariaKoptjevskaja-Tamm (eds.) The Circum-Baltic Languages. Typology and Contact. Vol. 2. Grammar and Typology523–568. Amsterdam / Philadelphia: Benjamins.
Koptjevskaja-TammMaria. 2009. A lot of grammar with a good portion of lexicon: towards a typology of partitive and pseudo-partitive nominal constructions. In HelmbrechtJohannesYokoNishinaYong-MinShinSkopeteasStavros and VerhoevenElisabeth (eds.) Form and Function in Language Research 329–346. Berlin: Mouton de Gruyter.
KornfiltJaklin and von HeusingerKlaus. 2009. Specificity and partitivity in some Altaic Languages. In Ryosuke Shibagaki and Reiko Vermeulen (eds.) mit Working Papers in Linguistics (mwpl) 58. Proceedings of the 5th Workshop on Formal Altaic Linguistics (wafl 5) 19–40. Cambridge: mit.
Kuz’minaIrina B. and NemčenkoElena V.. 1964. K voprosu o konstrukcijax s fomoj imenitel’nogo padeža imeni pri perexodnyx glagolax i pri predikativnyx narečijax v russkix govorax. In Avanesov Ruben I. (ed.) Voprosy dialektologii vostočnoslavjanskix jazykov151–175. Moskva: Nauka.
MalyševaAnna. 2004. Roditel’nyj padež prjamogo dopolnenija ob”ekta pri glagolax nesoveršennogo vida v russkix letopisjax i sovremennyx arxangel’skix govorax. In KasatkinLeonid L. (ed.) Materialy i issledovanija po russkoj dialektologiiii(viii) 170–184. Moskva: Nauka.
MalyševaAnna. 2008. Tipy upotreblenija količestvennogo genetiva isčisljaemyx suščestvitel’nyx (na materiale sovremennyx arxangel’skix govorov i russkix letopisej). In KasatkinLeonid L. (ed.) Materialy i issledovanija po russkoj dialektologiiiiI(ix) V. Grammatika232–247. Moskva: Nauka.
MehligHans Robert. 2006. Glagol’nyj vid i vtoričnaja gomogenizacija oboznačaemoj situacii: K upotrebleniju delimitativnogo sposoba dejstvija v russkom jazyke. In Lehmann Volkmar Glagol’nyj vid i leksikografija: semantika i struktura slavjanskogo vida 4 235–76. München: Sagner.
MetslangHelle. 2001. On the developments of the Estonian aspect. The verbal particle ära. In: DahlÖsten and MariaKoptjevskaja-Tamm (eds.) Circum-Baltic Languages. Volume 2: Grammar and Typology443–481. [Studies in Language Companion Series]. Amsterdam / Philadelphia: John Benjamins.
NauNicole. 1996. Ein Beitrag zur Arealtypologie der Ostseeanrainersprachen. In BoretzkyNorbert (ed.) Areale Kontakte Dialekte Sprachen und ihre Dynamik in mehrsprachigen Situationen 51–67. Bochum-Essener Beiträge zur Sprachwandelforschung Bd. 24. Bochum: Brockmeyer.
NauNicole. 2014. Differential object marking in Latgalian. In Holvoet Axel and Nicole Nau (eds.) Grammatical Relations and Their Non-Canonical Encoding in Baltic. [Valency Argument Realization and Grammatical Relations in Baltic 1] 207–255. Amsterdam / Philadelphia: John Benjamins.
ParteeBarbara H.2008. Negation, intensionality, and aspect: Interaction with np semantics. In RothsteinSusan (ed.) Theoretical and Crosslinguistic Approaches to the Semantics of Aspect V. 110291–317. Amsterdam / Philadelphia: John Benjamins. (manuscript (final): https://udrive.oit.umass.edu/partee/ParteeInPressNegIntensAsp.pdf).
SchwyzerEduard and DebrunnerAlbert. 1950. Griechische Grammatik. Auf der Grundlage von Karl Brugmanns Griechische Grammatik. Zweiter Band. Syntax und syntaktische Stilistik. Vervollständigt und herausgegeben von A. Debrunner. München: C. H. Beck’sche Verlagsbuchhandlung.
SelkirkElisabeth O.1977. Some remarks on noun phrase structure. In CulicoverPeter W. ThomasWasow and AdrianAkmajianFormal Syntax: Papers from the mssb-uc Irvine Conference on the Formal Syntax of Natural LanguageNewport Beach California June 9–11 1976285–316. New York: Academic Press.
SeržantIlja A.2014a. Denotational properties of the independent partitive genitive in Lithuanian. In HolvoetAxel and NauNicole (eds.) Grammatical Relations and their Non-Canonical Encoding in Baltic257–299. [Valency Argument Realization and Grammatical Relations in Baltic 1]. Amsterdam / Philadelphia: John Benjamins.
SeržantIlja A.2014b. Independent (bare) partitive genitive in Russian and North Russian. In SeržantIlja A. and BjörnWiemer (eds.) Contemporary Approaches to Dialectology: The area of North Northwest Russian and Belarusian subdialects 270–329. Slavica Bergensia 13. Bergen: John Grieg AS.
SuihkonenPirkko. 2012: On aspect, aspectual domain and quantification in Finnish and Udmurt. In Suihkonen Pirkko BernardComrie and ValerySolovyev (eds.) Argument Structure and Grammatical Relations. A Crosslinguistic Typology325–354. slcs. Amsterdam / Philadelphia: John Benjamins.
TammAnne. 2006: Estonian transitive verbs and object case. In Butt Miriam and Tracy Holloway King (eds.) Proceedings of the lfg06 Conference 484–504. Universität Konstanz: csli Publications (http://csli-publications.stanford.edu/.
WiemerBjörnSeržantIlja A. and ErkerAksana. 2014. Convergence in the Baltic-Slavic contact zone. Triangulation approach. In Besters-DilgerJulianeDermarkarCynthiaPfänderStefan and RabusAchim (eds.) Congruence in Contact-Induced Language Change15–42. Berlin: Mouton de Gruyter.
1 It is not the aim of the paper to go into terminological discussion of whether the term partitive case or partitive genitive case or, e.g., the term partial used in Koptjevskaja-Tamm and Wälchli (2001) or some other term might be more appropriate here. It seems that as long as different facets of this case – which indeed not all comply with the semantic category partitivity – are sufficiently illustrated in the respective sections, the reader will not be misled by the very term. Given the functional versatility of this case it might be quite difficult to provide a single name for it so that it would cover the whole variety of meanings. At the same time, the fact that these different facets tend to cluster together is indicative of the existence of a hypercategory.
2 Intensional contexts are contexts that evoke concepts, not referents, and are not to be confused with intenTional contexts that typically form a part of intenSional contexts. They are opposed to the extensional approach to meaning, which attempts to correlate expressions in language with aspects of the world (Cruse, 2000: 21). It has been stressed in the literature that intensional verbs may typically have two readings: a specific or transparent reading (the speaker has a particular referent in mind as the object of intention) and an opaque, non-referential reading, i.e. with no existential presupposition (Quine, 1960: §32; Zimmermann, 1993). The latter has been argued to be, more precisely, a non-referential, existentially non-committal property-denoting reading (Borschev et al., 2007).
3 Only personal pronouns have a dedicated accusative case in some Finnic languages, e.g., in Finnish. I gloss the Finnic genitive (sg.) /nominative (pl.) /accusative (personal pronouns) case here as accusative following the typological tradition, cf., inter alia, Kiparsky (1998).
4 The nominative objects are used in constructions with no possibility to have an overt nominative subject.
5 It has been suggested that the Baltic and Slavic partitive genitive may arise from an ablative. This cannot be maintained for the following reasons. Morphologically all Baltic and Slavic genitive endings continue the Proto-Indo-European genitive endings except only for the masculine singular of the so-called o-declension. The latter ending, indeed, stems from the ie ablative. All other declensions (such as i-, u-, ū-, ā-, consonant-declensions) as well as the plural and dual endings, derive from the ie genitive and not from the ablative. It is, however, possible, but probably not verifiable, that the ie genitive ending itself has emerged from a previous ablative. In any event, this development cannot be related to the Finnic development of the partitive case from an ablative for diachronic reasons.
6 Exceptions are grammaticalized syntactic contexts that require the partitive (genitive) such as presence of a sentential negation. Another exception is the use of the ip(g) as the only vp-quantifier, cf. subsections 3.11, 3.12.
8 Note that the accusative marking would induce rather generic meaning as one of the reviewers kindly notes.
10 One of the reviewers suggests that there was first a change in meaning which, as a consequence, provided for the loosening of selectional restrictions. This is a tricky question, but it seems to me that it might have been exactly the other way around: at some point, kind expressions metaphorically entered the true partitive expression, because they have been treated as (super)sets. As a consequence, the overall meaning has extended including also the pseudo-partitive readings.
11 Quoted from Sands and Campbell (2001: 265).
12 Lees (2003) has compared two original and independent translations of St. Paul’s first letter to the Corinthians from Greek into Finnish (from 1992) and into Estonian (from 1989), using one of the latest translations of the New Testament.
13 In fact, it has been suggested that the ip(g) can induce an unbounded interpretation, especially with ingestion verbs in Ancient Greek and Vedic Sanskrit (Dahl, 2009: 37–41; Napoli, 2010). This is, however, not corroborated by the data presented. There might originally have been certain semantic considerations that created preferences for the ip(g) with a particular group of verbs in Ancient Greek or Vedic Sanskrit, as argued in Dahl (2009) and Napoli (2010). However, Napoli (2010) presents herself a number of co-occurrences of the ip(g) and aorist as well as the ip(g) and imperfect, both aorist and imperfect are grammatically marked in Ancient Greek for boundedness and unboundedness, respectively. The independence of the ip(g) from aspect/boundedness has been emphasized in Bauer (2007: 134) and Seržant (2012a: 133).
14 The contexts of most of the examples that are often cited in the literature (cf. Krysko, 2006: 179–185; Lopatina, 1998: 243; Malyševa, 2008: 240) do not unequivocally reflect the partial affectedness reading and a holistic reading is equally possible.
15 Cf. also Metslang (2001) on Estonian, who also adheres to a straightforward relationship between the partitive case and imperfective. However, what she means here is non-culmination, as will be clear from the following discussion.
16 Kiparsky (1998), who cites this example, adduces two other readings available: (iii) ‘He opened the window for a while.’ (cf. temporal transfer below) and (iv) ‘He opened the window again and again.’ Those are, however, not relevant at this point.
17 I adopt here the notion delimitative following, inter alia, Sasse (2002: 206), first introduced in Maslov (1959). Maslov refers to delimitatives as to aktionsart (Russ. ‘sposob dejstvija’), i.e. as pertaining to the domain of actionality and not to aspect sensu stricto. Delimitatives are typically derived from homogenous non-culminating predicates such as to walk or to sleep (cf. Mehlig, 2006 for this argument on Russian). However, as Mehlig (2006: 253ff) notes, there are a number of accomplishment verbs like pisat’ ‘to write’, pit’ ‘to drink’ or even otkryt’ ‘to open’ as well that can be conceptualized as homogenous, if the focus is on the activity taking place before the inherent culmination/endpoint.
18 The lack of definite boundaries might have been the reason for treating it as unbounded. Indeed, to some extent this reading behaves as unbounded: opening the window somewhat plus opening the window somewhat may also yield opening the window somewhat, tests on divisiveness will give analogical results. However, this reading is only restrictedly unbounded and above/below some level it stops being cumulative or divisive, respectively.
19 Note that verbs of other actional classes such as, e.g. stative and imperfective znat’ also allow ip(g) objects in North Russian. However, the ip(g) with these verbs has only np-internal functions with no impact on aspect and actionality.
20 The ip(g) may also be used with telic verbs, as has been discussed in subsection 3.10.2 above. I claim that the arbitrarily quantified ip(g) also imposes an arbitrary boundary on the whole event. The aspectual and actional value of the vp is thus perfective (temporal boundary/arbitrary boundary) and non-culminating – there is no commitment to exhaustiveness. The latter is removed by the ip(g) case-marking. Thus, Russian vy-pit’ (prfv-drink.inf) ‘to drink up’ is typically culminating (telic) as opposed to pit’ (drink.inf) ‘to drink’. Yet the vpvy-pit’ vody ‘to prfv-drink water.gen.sg’, with the object marked by the ip(g), is non-culminating (atelic), because there is no clear inherent endpoint implied; the action referred to is inherently cumulative. Furthermore, it has been claimed in the literature that Russian telic perfectives require definite objects. This is obviously not found with the ip(g) which is inherently indefinite. Crucial for the Russian perfectives is rather the presence of some boundary of the object’s referent, not necessarily its definiteness/inclusiveness.
21 Only with delimitatives.
23 Tveite (2004: 50–2) shows with corpus data that Larsson’s assumption about the accusative extending partitive in Livonian is not justified. Thus, partitive scores much higher than accusative on token frequency.
25 This is valid also with regard to the partitives headed by a quantifying expression in Slavic and Finnic (Brattico, 2011).
Lees (2003) has compared two original and independent translations of St. Paul’s first letter to the Corinthians from Greek into Finnish (from 1992) and into Estonian (from 1989) using one of the latest translations of the New Testament.
Cf. also Metslang (2001) on Estonian who also adheres to a straightforward relationship between the partitive case and imperfective. However what she means here is non-culmination as will be clear from the following discussion.
Kiparsky (1998) who cites this example adduces two other readings available: (iii) ‘He opened the window for a while.’ (cf. temporal transfer below) and (iv) ‘He opened the window again and again.’ Those are however not relevant at this point.