Contact-induced change in the languages of Europe: The rise and development of partitive cases and determiners in Finnic and Basque

2.1 The partitive case in Finnish and other Finnic languages

Even though it morphologically shares the distribution of case endings, the Finnic partitive is peculiar, as it does not fulfill the typical function of cases, i. e., to indicate the grammatical relation of an NP (Luraghi 2003: 61; cf. Blake 2000: 1). Rather, in Finnic languages the partitive encodes both subjects and direct objects, and expresses a meaning connected with open quantity, that is, indefiniteness or unboundedness with mass nouns or with count plurals (Luraghi and Kittilä 2014).

Compare (2) and (4) with (1) and (3).

Examples such as (1) and (2) are often discussed in the framework of differential object marking, usually disregarding co-occurring differential subject marking as shown in (3) and (4). In such occurrences, the partitive has a quantifying function, indicating an unbounded quantity, and is not connected with a particular grammatical relation. The quantifying function of the partitive can be activated with mass nouns, as in (2), or plural count nouns, as in (4). With the latter, it comes close to the function of an indefinite article but, while the partitive case can function as a marker of indefiniteness with plural count nouns, it remains different from an indefinite article, because crucially it cannot indicate indefiniteness with singular count nouns. ^[2] To be sure, the partitive case also has a number of other functions that we partly illustrate below, but we consider syntactic multi-functionality, i. e., the ability to be used quite independently of particular grammatical relations, to be its distinctive feature for the sake of comparison with other languages.

Besides occurrences such as (2) and (4), in Standard Finnish the partitive is regularly used under negation. In particular, negated objects and subjects of existential clauses, which typically combine an indefinite, non-topic subject argument with an unaccusative verb, are invariably marked with the partitive, regardless of verbal aspect, definiteness or quantification, as in (5) and (6).

In addition, atelic verbs take partitive objects. This happens with inherently atelic verbs (odottaa ‘wait’, seurata ‘follow’, rakastaa ‘love’, pelätä ‘fear’) or highly transitive, telic verbs construed atelically, as in (7). A verb such as tappaa ‘kill’ takes accusative objects. When it occurs with the partitive, unboundedness can be understood as a property of the object, which gives an indefinite reading, or of the event, which gives an imperfective reading, as shown in (8). ^[3]

Partitive subjects in Finnish are not limited to unaccusative verbs, as they can also occur with transitive verbs. An example is (9).

In such occurrences, according to Huumo (2018: 425, 437), the subject must be interpreted existentially: ‘There were many people waiting …’. Huumo (2018) conducted an in-depth study of partitive subjects of transitive verbs, and found that they are often accompanied by quantifiers, and, if plural, are always indefinite, even though they can co-occur with both indefinite and definite direct objects. The latter is the case in (10).

More occurrences of partitive subjects with transitive verbs are discussed in Sands and Campbell (2001); note however that acceptability judgements for their examples vary among speakers.

The conditions for partitive marking in Standard Finnish obtain, by and large, in the other Finnic languages, and can thus be reconstructed to Proto-Finnic (Ojajärvi 1950: 128–129; Laanest 1982: 295). Areas of variation concern the usage of the partitive for open quantity objects, which appears to be more restricted in East Finnic languages, such as Kven, Karelian and Vepse (Kettunen 1943: 46). South Vepse also shows partitive transitive subjects (Kettunen 1943: 50–51), as does Livonian (Denison 1957: 128). At least as far as Vepse is concerned, the roots of this phenomenon seem to lie in internal analogical generalizations (Ritter 1989: 45–46). The extent of variation in other Finnic languages, and the parameters governing that variation, is not precisely known as, aside from Finnish, detailed quantitative studies on partitive usage are available only for Estonian (Nemvalts 1996; Metslang 2012) and South Vepse (Ritter 1989)

2.2 Partitive and ablative in other Uralic languages

In order to make clear to what extent the usage of the partitive in Finnic, as sketched above, may or may not be conditioned by language contact, its historical development within the Uralic language family must be briefly sketched. The closest relative to Finnic within Uralic is Saami, which, with Finnic, developed from a specific Finnic-Saami proto-language immediately ancestral to Proto-Finnic and Proto-Saami. Within Uralic, the closest relative to Finnic-Saami appears to be Mordvin. The Finnic partitive derives from the Uralic ablative *-tA (Wickman 1955: 27). In Finnic, the partitive lost its local function: it has there been replaced by the present elative and ablative cases, formed by compound endings -s-ta (talo-sta ‘out of the house’) and -l-ta (pöydä-ltä ‘from (the surface of) the table’). The original local function of the partitive case is currently preserved only in spatial nouns such as al-ta ‘from below’, taka-a ‘from behind’.

Uralic ablative *-tA lies at the basis of the Saami plural accusative *-i-tA, which in Finnic is the underlying form of the plural partitive East Saami languages have retained the Uralic ablative in some restricted contexts such as comparative constructions, certain quantifiers and numerals, adpositions (Feist 2010: 223). Furthermore, in the earliest surviving texts from South Saami the ablative case marks direct objects with atelic verbs like ‘seek’, ‘follow’, etc. (Korhonen 1981: 216). As South Saami does not border any Finnic languages, this phenomenon can hardly be explained through language contact with Finnic, and may well be inherited.

South Saami also shows differential object marking (DOM) with plural objects, partially definiteness-based and partially quantity-based. Plural objects appear unmarked if they are indefinite or a closed set or a pair, but marked with the plural accusative (*-i-tA) if definite or if affected in an incremental, one-by-one manner (Magga and Magga 2012: 185–186; Korhonen 1981: 216), as in (11) and (12).

As argued by Korhonen (1981: 215), the diachronic development that gave rise to the Saami plural accusative is best understood as involving DOM, with the Uralic ablative *-tA used as the object marker of certain atelic verbs and, to some extent, with open quantity objects. This differential marking would later have been simplified with accusative (Uralic *-m) generalized in the singular and ablative/partitive (*-tA) generalized in the plural. This is a matter of some import for the historical background of the Finnic partitive, and we will return to it below (2.6).

In Mordvin, the Uralic ablative marker has developed into -do, which no longer has ablative function (it means neither separation nor origin or place). It is used with consumption verbs such as Erzya jarsams ‘eat’ with open quantity objects, and for the object of verbs of emotion (Bartens 1999: 94). Aside from this, it is used with quantifiers. In its original locative usage, the Mordvin ablative only survives in some highly idiomatic constructions (Bartens 1999: 94). Note that the Mordvin ablative is not used with other atelic verbs, nor does it signal imperfective aspect as in Finnic and historically also in Saami. In Erzya, the inessive case is used as an object case here (Bartens 1999: 96). The plural ablative in Mordvin is used only in the definite conjugation, formed by agglutination of a demonstrative pronoun to the plural stem (Bartens 1999: 82, 86–87), this means that semantically plural open quantity objects, being indefinite, are marked as singular in Mordvin (Bartens 1999: 94) and that the Finnic-Saami plural object marker *-i-tA has no cognate in Mordvin.

None of the other Uralic groups show the development of a partitive case out of the inherited ablative with the possible exception of West Mansi. In West Mansi dialects, a case based on the Ugric ablative *-nāl (unrelated to Uralic *-tA) is used as an object case marker -n, -nV. Partial ablative objects occur in other Mansi dialects (Liimola 1963: 43–44). ^[4] It is unclear to what extent such ablative arguments in other Mansi dialects are used as partitive objects: most of the examples mentioned by Liimola (1963: 43) show incremental theme objects, and the range of verbs that govern the ablative case appears to be restricted (‘eat’, ‘drink’, ‘buy’). In Mansi, the Uralic accusative *-m has disappeared in the West Mansi dialects which show grammaticalization of -n, -nV (Liimola 1963: 31). The latter suffix thus filled a morphosyntactic gap and might have done so without any intermediate partitive stage.

2.3 Partitives in other languages of Eurasia

As the hypothesis pursued here is that the Finnic and Basque partitives both result from Indo-European influence, it is a matter of some relevance that partitives have been mentioned in literature on Turkic (Pakendorf 2007: 142–152) and Mongolic (Poppe 1974: 151) as well. As parallels to Finnic, however, they are only apparent, and should rather be considered cases of DOM, as instantiated in Turkish. As argued in von Heusinger and Kornfilt (2017), the ablative case in Turkish occurs in partitive constructions, and can be used as object marker, provided that reference is made to a part of a previously specified whole. Consider Example (13).

The authors provide the following context for this sentence: “My mother always fills a big bowl with different pieces of apples, pears, and bananas. Yesterday evening I was intensively studying the different pieces of fruit, which were 8 apples, 10 pears and 4 bananas, and then …” (von Heusinger and Kornfilt 2017: 9)

Thus, as argued by these authors, partitive constructions indicate specificity in Turkic languages, as well as in Mongolian. Indefinite non-specific in Turkish trigger zero marking rather than accusative marking, as in (14).

The partitive in Sakha and Dolgan developed out of the Common Turkic locative case *-dA (Stachowski and Menz 1998; Tenišev 2002). This had both locative and ablative meanings in earlier Old Turkic (Erdal 2004: 173–175); it has been suggested that the partitive developed from the ablative meaning of the locative (Poppe 1959: 681). Its origin is thus the same as that of the Turkish ablative case used in partitive constructions as in shown above in (13). Examples discussed by Pakendorf are (15) and (16).

Notably, the Sakha partitive is restricted to partially affected mass noun objects and indefinite singular count nouns, as in (16). It is limited to the imperative mood (Pakendorf 2010: 728). It appears to be based on contact influence from Evenki, which sports a contrast between definite and indefinite accusative in certain moods (Pakendorf 2010: 727; see further Poppe [1959: 681]). The Sakha partitive likewise appears to signal indefiniteness rather than partiality, and as it only marks direct objects, it does not provide a parallel to the Finnic partitive.

A case slightly more similar to the Finnic one is found in Tofa, where a partitive (based on the Turkic locative) is used to mark partially affected mass nouns, or an unbound quantity of a mass noun in direct object position in the imperative mood, as in (17).

The Tofan partitive does not primarily mark indefinite objects as the Sakha partitive, and does not occur with singular count nouns. The use of Turkic *-dA for unbound quantities of mass noun objects is known elsewhere in Turkic, cf. (13).

In Khalaj, the ablative case appears to mark plural objects of open quantity as well as mass noun objects (Doerfer 1998: 91). As Tungusic influence can hardly be supported for Tofa or Khalaj, Johanson (2012: 2015) considers inheritance from Proto-Turkic as a possibility. In conclusion, Turkic partitives or partitive-ablatives occur in a much more restricted fashion than that of Finnic, are basically limited to specific objects, are often conditioned by verbal mood, and notably mark objects only.

2.4 The Balto-Slavic partitive-genitive

In Baltic and Slavic languages, the genitive is used as an object and subject marker in a manner that much more closely parallels Finnic. The genitive is used for open quantity direct objects both in Slavic (Koptjevskaja-Tamm and Wälchli 2001: 653) and in Baltic languages (Lithuanian and Latgalian; Holvoet and Nau 2014: 9–10); cf. (18) and (19). In contrast to the Finnic partitive, which primarily marks main arguments, the genitive in Baltic and Slavic retains a wide range of other uses, such as marking adnominal possessors. ^[5]

Open quantity, unbounded subjects may also be marked with the genitive in existential clauses and with unaccusative verbs, as in (20) and (21).

Moreover, the genitive marks negated objects in Russian, Polish and several other Slavic languages (Koptjevskaja-Tamm and Wälchli 2001: 653), in Lithuanian and Latgalian, though not in modern Latvian (Holvoet and Nau 2014: 7–9). These features appear to be common Balto-Slavic in origin: the near-absence of the partitive genitive in Standard Latvian is an innovation (Holvoet and Nau 2014: 7–9). In Slavic, the genitive of negated objects appears to be best represented in Polish, Russian (Seržant 2015: 357), Old Church Slavonic and Slovenian (Pirnat 2015: 4). Genitive subjects are somewhat sporadic outside of North Russian (Seržant 2015: 359), but they are attested in other Slavic languages (Koptjevskaja-Tamm and Wälchli 2001: 660). Genitive partial objects are attested in Old Church Slavonic with certain verbs (‘taste’, ‘take’) (Lunt 2001: 145) and genitive objects occur with verbs of perception, striving and attainment (Lunt 2001: 145).

Similar to the Finnic partitive, the partitive-genitive in Lithuanian and North Russian can signal temporal delimitation, cf. (22)–(24).

As Koptjevskaja-Tamm and Wälchli (2001: 657) point out, the range of intransitive verbs that can commonly or marginally occur with partitive subjects in existential or presentative constructions in Finnic is very broad. In Baltic and Russian, they occur in a more restricted fashion (Koptjevskaja-Tamm and Wälchli 2001: 659; Seržant 2015: 359), though more commonly in North Russian dialects influenced by Finnic (Koptjevskaja-Tamm and Wälchli 2001: 659–660), while in other Slavic languages they are even rarer (Koptjevskaja-Tamm and Wälchli 2001: 660). In Old Church Slavonic, genitive subjects may occur in negated existential clauses with copular byti (Lunt 2001: 146).

Historically, the Balto-Slavic genitive is based on a merger of the Indo-European genitive and ablative (Holvoet 2018: 2003). Its partitive functions have clear Indo-European roots (Section 2.5). For reasons of possible contact with Finnic, we anticipate here some remarks on the partitive genitive in Germanic, particularly Gothic. In Gothic, the genitive is used for partial objects as well as objects of experiential predicates, as in other ancient Indo-European languages, but it is also used generally with negated objects and negated subjects of existential clauses, as in (25).

There may be remnants of a genitive of negation in Old High German (Breitbarth et al. 2013: 151–152). In other Germanic languages, the genitive of negation does not occur, but the genitive is used for partial objects and the object of experiential predicates, e. g., in Old English (Mitchell 1985: 562–563). This latter usage is thus likely Proto-Germanic, in fact, Proto-Indo-European in origin. This is less clear for the genitive of negation as found in Gothic and, perhaps, Old High German: for Gothic at least, contacts with Balto-Slavic are also a possibility, however, compare (25) with the Greek example in (31) for a possible Indo-European origin of this construction.

2.5 The partitive genitive in ancient Indo-European languages

Far from being an isolate feature of Balto-Slavic, multi-functionality as a subject and object marker is typical of the partitive genitive ^[6] in many Indo-European languages. It occurs both in European and in non-European branches of the Indo-European language family. Examples (26) and (27) are from Indo-Iranian (Reichelt 1909: 257; Dahl 2014).

The partitive genitive functions as subject in (26) and as direct object in (27), indicates an unbounded quantity, and, as we have to do with plural count nouns, indefiniteness. Similar to Finnish, the partitive genitive subject in Avestan does not agree with the verb, which remains in the singular (Example (26)). ^[7] This is not the case in all Indo-European languages: in Ancient Greek, partitive genitive plural subjects could either agree or not agree in number with the verb (see Conti and Luraghi 2014 for a thorough discussion and examples).

The quantifying function of the partitive genitive was inherited in Indo-European languages. Notably, in Balto-Slavic and Ancient Greek the genitive merged with the ablative, so in principle one might consider the ablative as the source for the partitive meaning. However, in languages in which the ablative did not merge with the genitive the quantifying function was a feature of the genitive (not of the ablative). This is shown by Indo-Iranian, as in (26) and (27) and Latin, as in (28).

The partitive-genitive has a wide extension in Ancient Greek. As argued in Conti and Luraghi (2014), it could replace other cases in any function except third arguments of verbs of giving. Similar to Finnish, the partitive genitive could alternate with the accusative based on telicity, and indicate that an action was not carried out completely, as in (29) and (30), to be compared with (7). ^[8]

In Ancient Greek, the partitive genitive is only sporadically used under negation, and in such case, only in the function of subject, as in (31).

In all Indo-European languages, low transitivity predicates often take genitive direct objects. As already mentioned for Germanic (Section 2.4), this is true for example for experiential verbs, such as verbs of memory (e. g., Latin memini, Ancient Greek mimnēskō), verbs of emotion (Ancient Greek éramai ‘love’, ‘desire’, Russian bojat’sja ‘fear’) and perception verbs (Ancient Greek akoúō Sanskrit śru- ‘hear’).

2.6 Discussion

The Finnic partitive has been argued to be mainly the result of Baltic influence by Larsson (1983, 2001), and Bjarnadóttir and de Smit (2013). Larsson’s hypothesis has not met with general approval, as noted by Leinonen (2015: 197), and its ramifications for the timing of Finnic(-Saami)-Baltic contacts can even be said to have been neglected in recent literature, as we will argue below. Recently, it has been criticized by Seržant (2015: 404). While Seržant (2015: 393) does allow for Baltic influence in case of the Finnic partitive for negated existential subjects, he believes the partitive of negation in general to rest on an internal development in Finnic (Seržant 2015: 392) and argues that the interaction between partitive and aspectuality spread from Finnic to Balto-Slavic instead (Seržant 2015: 387–388). Seržant furthermore argues that other features of the Finnic partitive are indigenous within Finnic, such as its usage for partial objects (2015: 351) and for objects of experiential predicates (2015: 358). Seržant thus concludes that “[…] 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 […]” (Seržant 2015: 404).

Whereas it is true that the usage of the partitive-genitive was extended in Balto-Slavic when compared to other Indo-European languages, and that contacts with Finnic may account for this, we would nonetheless agree with Larsson (1983, 2001) that the main direction of influence was from Balto-Slavic to Finnic. Our arguments are the following: 1) The Balto-Slavic partitive genitive has clear Indo-European roots, even if partitivity developed to the highest degree in Balto-Slavic. Contrarily, the Finnic partitive has no Uralic cognates outside of (historically) Saami and (marginally) Mordvin. 2) A Balto-Slavic origin for the Finnic partitive can be harmonized with what we know of Finnic-Baltic contacts in general and their chronology. 3) Though the partitive is the only Finnic morphosyntactic feature of possible Balto-Slavic origin that has been researched in detail, there may be more such features.

To deal with the latest point first, Bjarnadóttir and de Smit (2013: 49–50) argue that, parallel to the grammaticalization of the Uralic ablative as an object marker, the Uralic accusative *-m, originally used for definite objects only, was extended to indefinite objects in Saami and Finnic. This development could easily be explained by contact with Baltic (which sports *-m with both definite and indefinite objects). De Smit (2015: 249–252) argues that a number of changes in the Finnic verb system may be due to the Baltic or Balto-Slavic as well. Proto-Uralic sported no voice system, however, by the time of Proto-Finnic, a morphological passive as well as a set of passive participial suffixes were introduced and integrated into a new tense/aspect system. This change closely matches Balto-Slavic and Baltic with which Proto-Finnic was in contact, furthermore, there are striking similarities between individual Finnic and Baltic passive constructions, such as the use of genitive agents (de Smit 2015: 260–264). These features, and the whole notion of Baltic influence on Finnic morphosyntax, need more research, but there is definitely a potential to integrate the hypothesis of a Baltic-origin partitive in a wider hypothesis of morphosyntactic Balticisms in Finnic. Added to this should be Larsson’s (2001: 247) argument that a very large number of loanwords were borrowed from Baltic to Finnic, but lexical borrowing in the opposite direction is very rare. It is of course possible for lexical influence and substratum influence to have opposite directions, but contact-induced grammatical or phonological change is not necessarily substratal: grammatical and phonological features may ride piggyback on lexical borrowings. In his skeptical take, Seržant often implicitly assumes that the inherited origin of a construction in Finnic is incompatible with an explanation based on Baltic contacts. For example, the usage of the Uralic ablative in object marking in Saami and Mordvin (Seržant 2015: 348) is argued to signify that “ a question emerges whether language contact still may sufficiently be argued for” (Seržant 2015: 349). This brings us to a problem that has generally been neglected in research literature: the timing of Baltic morphosyntactic and loanword contacts. The partitive may indeed be an inherited feature in Finnic (Seržant 2015: 248): in fact, the grammaticalization of the Saami plural accusative is best explained by assuming that the partitive was used to mark plural indefinite objects aside from incremental themes. However, this does not mean it did not arise through Baltic or Balto-Slavic influence at a language stage that preceded Finnic, i. e., Proto-Finnic-Saami. A number of early Baltic loanwords in Finnic have counterparts in Saami. Aikio (2012: 73–76) regards the Baltic loanwords in Saami as diffused through Finnic. However, as Larsson (2001: 242) points out, there is no linguistic evidence for this hypothesis. Baltic loanwords such as Finnish halla ‘overnight frost’, Saami suoldni ‘dew, haze, mist’ (from Proto-Baltic šalna ‘frost’) can clearly be reconstructed to a common Finno-Saamic form, hence must be assumed to have been borrowed during this period. Given that there are some Baltic loanwords in Saami without a counterpart in Finnic (Aikio 2012: 74), Finnic-Baltic contacts in the area of morphosyntax could well have begun in the period of the Finnic-Saami, rather than the Finnic, protolanguage.

Actually, there are more than thirty Baltic loanwords in Mordvin, too (Grünthal 2012), which result from direct contacts between (the ancestors of) the Mordvins and the Balts at the eastern edge of the Baltic linguistic area (Grünthal 2012: 299). Given obvious Slavic but also Turkic (Róna-Tas 1989: 766–768) influence on Mordvin, one should not overstate the historical significance of ablative-marked incremental objects in Mordvin. Their presence may have been conditioned or supported by a variety of contact languages: recall that ablative-marked incremental objects are common in Turkic languages, whose locative/ablative (*-dA) is also fortuitously very similar to the Mordvin ablative suffix.

When accounting for the chronology of language contact involved, one must keep in mind that the Germanic and Balto-Slavic loanword layers in Finnic and Saami represent but stages in a continuum of language contacts. Specifically, there is a layer of late Northwest Indo-European or Pre-Germanic and Pre-Baltic loanwords in Finnic and Saami (Kallio 2012: 227–228). Kallio (2012: 227) connects this loanword layer to the Corded Ware or Battle Axe period of 3200–2300 bce, which on the Indo-European side would represent the ancestors of Germanic, Balto-Slavic, Celtic and Italic. The Uralic recipient language would be a form of Proto-Uralic that had already areally spread but not linguistically diverged (Kallio 2006: 10, 17). In other words, there is evidence that a form of Indo-European directly ancestral to Germanic and Balto-Slavic was in contact with the West Uralic ancestor of Finnic, Saami and Mordvin. Given the extensive use of the partitive-genitive in the ancient Indo-European languages, it is possible that even the earliest stages of the development of the partitive in Uralic (as incremental theme object in Mordvin) have been conditioned by Indo-European influence (for an alternative scenario, see Koptjevskaja-Tamm and Wälchli 2001: 664).

The usage of the partitive for plural open quantity objects must have emerged during the Finno-Saami stage, perhaps through an intermediate stage of incrementality, which may be reflected in South Saami. The rise of partitive subjects should be reconstructed for Proto-Finnic. Given the Indo-European roots of the equivalent phenomenon in Balto-Slavic, a Balto-Slavic or Proto-Baltic origin of Finnic partitive subjects seems quite unproblematic, even though the extension of partitive subjects in Finnic is wider than its original extension in Indo-European languages. Concerning the partitive under negation, Seržant (2015: 392) argues that “this rule is a common Eastern Circum-Baltic innovation, not inherited from any of the ancestor languages.” Granting that evidence for the genitive of negation in other Indo-European languages is sketchy (Pirnat 2015: 20), the emergence of the genitive of negation in Balto-Slavic can nonetheless be described as being based on the well-established Indo-European partitive-genitive (Pirnat 2015: 25–26), whereas we lack such a basis on the Uralic side. Whereas it is true that the Finnic partitive is used more widely than the Baltic partitive-genitive, which is historically on the wane (Leinonen 2015: 198), this does not indicate the directionality of contact influence. There is no law in diachronic linguistics that precludes the possibility of a borrowed feature extending its range of usage in a replica language, and Larjavaara (1991: 402–403) argues this to have happened precisely with the aspectual usage of the partitive in Finnic, though Seržant (2015: 360) seems somewhat reluctant to accept this possibility.

We thus argue, with Larsson (2001: 247), that the scenario of Balto-Slavic influence on Finnic-Saami and Finnic is strongly preferable to that of a Sprachbund-like phenomenon arising from multilateral contacts. This position is supported by the weight of a number of arguments not in themselves decisive: the Indo-European roots of the phenomenon on the Balto-Slavic side which are not matched by any equivalent Uralic roots on the Finnic side, and the (unproven) possibility of more Baltic-influenced morphosyntactic changes in Proto-Finnic taken together with the (proven) presence of Baltic loanwords in Finnic. The basic argument is this: compared to the other Uralic languages, Proto-Finnic developed a number of features (grammatical voice, an analytical perfect tense, a more thorough distinction between subject and object case) that moved it typologically closer to surrounding Indo-European languages such as Balto-Slavic and Germanic. But we cannot state that Balto-Slavic moved into a more ‘Uralic’ direction as compared to its Indo-European precursor. ^[9] This is particularly the case if Mordvin, rather than Finnic, is taken as exemplifying Uralic morphosyntax. Distinctive features such as definiteness-based primary argument marking, object cross-reference on the verb and absence of voice opposition with participles are all inherited from Uralic and must be assumed for the earliest stages of Proto-Finnic-Saami as well (de Smit 2015: 248–250), but they have no equivalent in Balto-Slavic.

A question then remains: on what basis did speakers of Proto-Finnic-Saami identify the Uralic ablative and the Balto-Slavic genitive? In their core functions, the cases were after all very different: the Uralic ablative was an adverbial case marking location, the Balto-Slavic genitive primarily marked adnominal possession. An answer provided by Koptjevskaja-Tamm (2001) might be the usage of the Uralic ablative in constructions with nominal and numeral quantifiers, which is well-established in Mordvin (Bartens 1999: 94, 119, 121). This has, as Koptjevskaja-Tamm (2001: 540–541) remarks, parallels in Balto-Slavic and elsewhere in Indo-European. Koptjevskaja-Tamm and Wälchli (2001: 664–666) argue that its presence in quantifier constructions (which obviously occurred in subject and object positions) may have provided a basis for the grammaticalization of the partitive as an object case in Finnic through dropping the quantifier. The use of the genitive in pseudopartitive constructions in Indo-European is archaic (Koptjevskaja-Tamm 2001: 554), and Koptjevskaja-Tamm (2001: 562–563) considers the usage of partitives in (pseudo-)partitive constructions in Finnic and Saami to be indigenous, while allowing for Balto-Slavic contact in the case of numeral constructions. If this is correct, (pseudo-)partitive constructions may have provided a basis for the interlingual identification of the Uralic ablative and the Balto-Slavic genitive.

3.1 The Basque partitive

Besides Finnic languages, Basque also has a partitive case that is diachronically connected with an earlier ablative (Ariztimuño 2014). In Basque grammatical descriptions, the partitive is often described as a determiner (Trask 2003: 124; de Rijk 2008), as it partly shares the distribution of determiners. ^[10]

Examples (32)–(35) illustrate the widely held view that “[t]he partitive marker can be argued to be the negative form of the existential interpretation (in absolutive case) of the Basque definite article (D) [-a(k)]” (Etxeberria 2014b: 310). This interpretation was defended earlier by Trask (2003: 124): “The partitive is a polarity item, and it occurs chiefly in polarity contexts, in positions in which an absolutive would otherwise occur”. This implies that the partitive encodes not only direct objects, but also subjects, as in (35). Lack of association with a specific grammatical relation is typical of partitive markers (Luraghi and Kittilä 2014: 17, 20).

Nonetheless, typical partitive constructions in Basque (with some degree of dialectal variation) are not limited to negative contexts (Trask 2003: 124–125). The partitive form may be found in existential sentences (36), especially when an adjective accompanies the noun, and, when used with non-existential predicates, it may introduce an emphatic existential interpretation (37).

Moreover, the Basque partitive also occurs in interrogative clauses (38) and in subordinate clauses with non-assertive modality, especially conditionals in Modern Basque (39). At earlier stages of the language the partitive also occurred in other types of clause, cf. (40), a purpose clause.

3.2 From ablative to partitive

As already mentioned, the Basque partitive originated from an ablative. The Modern Basque ablative has the ending -tik, which differs from the partitive -ik/-rik, but historically the ablative has been characterized by varying endings: both -rean and -(r)ik are attested in older texts, from the sixteenth century onwards. The form in -rean is documented in western dialects until the seventeenth century, whereas in the east the ablative morpheme was only -(r)ik. This was also used in the west until it was replaced by -tik (Santazilia 2013: 262–263). The etymology of -(r)ik is not clear: although several proposals have been made to derive this marker from reconstructed verbal bases (see Ariztimuño [2014: 339–342] for an overview), they remain rather speculative.

The original ablative meaning is still present in several constructions. Some quite idiomatic expressions are shown in (41). They are being replaced in Modern Basque by alternative constructions containing the new ablative form in -tik: instead of kalerik kale ‘from street to street’, speakers now tend to prefer kale batetik bestera ‘from one street to another’.

The Basque partitive is also used in superlative constructions, as in (42), in which it can also be indicative of an ablative origin (‘the nicest from (the set of) toys’ > ‘the nicest toy’).

The ablative origin also underlies adverbial partitive constructions denoting a resultant state (Etxepare 2003: 552), a sort of functional parallel to the Latin ablative absolute or other absolute constructions in ancient Indo-European languages, cf. (43).

In the grammaticalization process that can be posited for the development of this kind of construction, the point of departure is the ablative: *‘from/since seen it’ > ‘having seen it’ (Ariztimuño 2014: 336–337). Such an evolution from an ablative meaning to near past is attested in other languages too (Heine and Kuteva 2002: 34).

Another usage of the partitive, though not very common, can be found in a type of complex quantifier construction that includes words like asko ‘much’, gutxi ‘few’, ezer ‘something’ and inor ‘someone’ (the two latter words meaning ‘nothing’ and ‘nobody’, respectively, under negation). Crucially, such constructions are typically used in negative contexts, see (44) and (45), although sporadically they may also be found in affirmative constructions, as shown in (46).

One of the usual thanking formulae in Basque ( eskerrik asko ‘many thanks’ < lit. ‘many from/of thanks’) is a fossilized instance of this partitive construction, ultimately derivable from the ablative case.

From a paradigmatic standpoint, it is obvious that the Basque partitive does not share the morphological properties of case morphemes. It has just one form, which is number-indifferent, in contrast to other cases within the nominal paradigm, which can be attached to singular, plural and even indefinite bases. Consequently, the degree of integration of the partitive in the paradigm is so low that it is better characterized as a determiner, an extra-paradigmatic or pseudo-inflectional marker (Trask 1997: 93; Santazilia 2013: 273). On the other hand, as Ariztimuño (2014: 324) rightly points out, the partitive differs from determiners, especially the definite article, in that it does not allow any other case morpheme to be attached to it (see argi-a-ren ‘light.sg-def-gen’ vs. argi-rik ‘light-par’, with no option of incorporating any other case marker). Still, the low degree of integration of the partitive in the nominal paradigm may have certain diachronic implications. Even though it derives from an ablative, intra-paradigmatic form, its further development makes us suspicious as to the language-internal nature of its evolution, since the partitive exhibits peculiar traits that are to a certain extent reminiscent of the behavior of determiners in the Romance languages. We turn now to this issue.

3.3 The emergence of the Romance partitive article

A parallel to the Basque partitive determiner can be drawn with the partitive article of the Romance languages, such as French du/de la/des. This latter item developed out of a former marker of a case function, the ablative preposition de, which also turned into the marker of the ablative and genitive relation in Romance. The preposition merged with the definite article in part of the Romance languages, and gave rise to a partitive determiner (Carlier 2007) as in (47). Example (48) from Late Latin shows the onset of the development (see Luraghi 2013 for more examples and discussion).

The chronology of the Romance and the Basque developments, assuming – as is probably the case – that the rise of a definite article in Basque goes back no farther than the Medieval period (on the reasons for this, see Section 3.1), allows viewing the latter as a contact-induced change as well. It also needs to be remarked that the development of the partitive article at its onset also involved Ibero-Romance (see Carlier and Lamiroy 2014: 502–504), as shown in Example (49).

When functioning as a determiner, de plus definite article no longer shares the distribution of other prepositions, as it does not indicate a grammatical relation (adnominal modification, the genitive function) or the semantic role (generally source) of a noun. Compare the determiner use of di plus definite article with its use as preposition in Modern Italian.

The prepositional use of di is kept distinct from its use as a determiner by its distribution. Example (50) shows, among other things, that di plus definite article in its determiner use can also co-occur with other primary prepositions, while two primary prepositions in their proper function never co-occur (Luraghi 2013; Luraghi and Kittilä 2014: 24). This was already the case in sixteenth century French, as shown in (47). Example (51) shows the adnominal use of a prepositional phrase with di, which corresponds to the Latin adnominal genitive.

3.4 Discussion

We must first acknowledge that our data do not meet all the conditions required by Thomason (2001: 93–94) and Poplack and Levey (2010: 410) to make an entirely solid case for contact-induced change. Specifically, we have highly limited access to the linguistic situation prior to the alleged contact between Basque and Romance, and this prevents us from drawing far-reaching conclusions. Also, we can see no significant differences in the presence of the partitive determiner among dialects that could have experienced a varying degree of influence from neighboring Romance languages. Notably, however, the extent to which partitive forms are used may vary, with the dialects in contact with French and Gascon showing a more extensive use. The Low Navarrese variety of Mixe (in Basque, Amikuze) analyzed by Haase (1992) in his study of the influence of French and Gascon upon Basque provides some examples. Though to a more limited extent than Standard French, Gascon varieties also sport a kind of partitive marker, featuring de plus the definite article, which occurs when an adjective accompanies a noun as in (52). This marker appears especially in the region of Bearne, which is in contact with some Souletin varieties of Basque (Allières 1992: 811), but probably also in other areas (for the general absence of a partitive determiner in Gascon, see Rohlfs [1977: 178–179]).

In Mixe Basque, along with other functions that are common to other dialects, the partitive/ablative in -rik also encodes the agent in passive constructions (Haase 1992: 132). This functional extension of the morpheme, with no parallels in the southern dialects (the Basque spoken in Spain), can be regarded as proof of a contact-induced change: Gascon uses de (alongside per) for the agent of passive constructions (Haase 1992: 132; Heine and Kuteva 2003: 551; Ross 2007: 126).

In this way, some of the evidence at hand appears to suggest that the effects of contact have indeed been responsible for the evolution of the Basque partitive. Additional arguments can support this hypothesis. First, as discussed in Section 3.2, the partitive shows, unless the other case markers in Basque, a very low degree of structural integration in the inflectional system, even though it derives historically from the ablative case. This status is a consequence of a deparadigmaticization process (Norde 2009: 131), which instantiates secondary degrammaticalization and can be described as a ‘discharge’ from an inflectional paradigm. ^[11] This innovation certainly constitutes an exceptional development within the Basque nominal morphology. Second, the diachronic pathway leading from an ablative to a partitive is a shared feature that may link Basque and Romance (Ross 2007: 125). Third, the functional distribution of the partitive has much in common, despite some differences, in Basque and Romance (recall the use of the partitive beyond exclusively negative constructions). Notably, the possibility of a contact-induced change, which is historically plausible on sociolinguistic grounds, is reinforced by other structural developments that point in the same direction. These include the functional replacement of the instrumental by the comitative case (likely under the pressure of the instrumental-comitative polysemy in Romance, see Haase [1992: 67]; Heine and Kuteva [2003: 542–543], Heine and Kuteva [2006: 248]) and the emergence of DOM with the dative case replacing the absolutive in certain varieties (Odria 2014). Perhaps most importantly, the development of a definite article out of a demonstrative pronoun (-a < *har ‘that’), in a way similar to the change that took place in Late Latin and Romance (Trask 1997: 199), can support the hypothesis of a contact-induced development of the partitive.

This last innovation, along with the emergence of the indefinite article and of the partitive marker, seems to suggest the creation of a determiner system at least partly modelled after the Romance languages. The grammaticalization of the definite article in Basque gave rise to a complete declensional paradigm opposed to the indefinite or determinerless one. According to Manterola (2009, 2015: cx, 249–251), the main inflectional endings of the nominal paradigm were formed through an agglutination process in which the inflected forms of the distal demonstrative were involved. Notably, this exclusively Basque development relies on an earlier grammaticalization of demonstratives that echoes that of Romance demonstratives (even though, as is well known, the rise of articles out of demonstratives is a rather common pathway of grammaticalization). In Aquitanian, “the more-or-less direct ancestor of Basque” (Trask 1995: 87, 1997: 182, 402), which is attested in inscriptions dating from the first to third century ce, there is no trace of the definite article (Gorrochategui 1984, forthc.). Accordingly, the classic account (see Manterola [2015: 241] for a summary) situates its appearance in the Middle Ages, well after the time of the Basque-Latin language contacts and even the first Basque-Romance ones.

On the other hand, the possibly even later emergence of the Basque indefinite article bat from the numeral ‘one’, a fairly common source crosslinguistically (Heine and Kuteva 2002) has been considered a rather recent case of replica grammaticalization by Haase (1992: 59–61; see further Heine and Kuteva 2003: 556) ^[12] and of contact-induced grammaticalization by Heine and Kuteva (2005: 247–248). Once again, the change is likely modelled after Romance (cf. French un(e) and Spanish un(a), both ‘one’ and indefinite article), even though it has not reached the same degree of grammaticalization (the use of the indefinite article in Basque is much more restricted than the use of corresponding Romance articles; Trask 2003: 122). In this perspective, the very development of a plural indefinite form like batzu (from bat ‘one’ and the collective suffix -zu; the final -k in the modern batzuk is a recent addition) may be considered, following Michelena (1987 [1971]: 148), as a replication of the Spanish plural unos/unas. Be that as it may, what seems clear is that the Basque system of determiners, including definite and indefinite articles as well as the partitive marker, exhibits a structure rather similar to that developed (previously, in all likelihood) by the surrounding Romance languages.

In view of all these facts, it seems fair to conclude that in the rise of the Basque partitive determiner the structural influence of the neighboring Romance languages, specifically French and Spanish, and to a lesser extent Gascon, should be regarded as a relevant, perhaps even determinant factor.