Causatives in the languages of the Volga-Kama Region

1 Goals, methods, data sources

The Uralic and Turkic languages of the Volga-Kama Region of European Russia are, due to the extensive structural convergence between them, frequently treated as a Sprachbund or language convergence area (e.g., Helimski 2003: 159). One feature shared by these languages is a strong position of morphological causatives.

Causative constructions, as defined by Song (2001), denote complex situations consisting of two component events: 1) the causing event, and 2) the caused event. The linguistic representations of that macro situation always involve two participants of the event. The causer does or initiates something in order for the causee to “carr[y] out an action or undergo[] a change of condition or state as a result of the causer’s action” (Song 2001: 257). Both Uralic and Turkic languages show a preference for transitivization (generally by means of causative suffixes) while detransitivization plays a comparatively small role, especially in contrast to Russian. The present paper focuses on morphosyntactic features of causative constructions in three Uralic (Mari, Udmurt, and Komi-Permyak) and two Turkic (Chuvash and Tatar) languages of the area.

Due to the generally similar classification of the languages under consideration, points of investigation that might yield data relevant to contact-linguistic surveys are found on a smaller scale and pertain to the minutiae of causative constructions in these languages. For each language or variety under consideration, the following core research questions deserve attention:

1. Is morphological causation productive irrespective of verbal transitivity features (see Section 5)?

2. Can more than one causative suffix be attached to a stem (see Section 5)?

3. What is the argument structure of a causative construction: how are the causee and the direct object of the caused action marked respectively (see Section 6)?

4. Are quasi-causative constructions productive, which consist of a causative derivation of a verb denoting some sort of physical or mental state or change and a single argument marked as the direct object, e.g., ‘(it) makes me cold’ → ‘I am cold’ (see Section 7); are there any lexical restrictions on this category?

Song divides causative constructions into three major types depending on their realization: lexical, morphological, and periphrastic (Song 2001: 260). In the lexical class, the difference between a non-causative and causative form is expressed by lexical means, i.e., suppletion, as in the case of to eat ↔ to feed in English.

In morphological causation, verbal affixes are used, as in Hungarian in (1).

In periphrastic causative constructions the predicate of the cause and the predicate of the effect are in different clauses. Additionally, it is the caused event and the NP referring to the causer that is foregrounded as opposed to the backgrounded event of the effect and the causee. The third criterion of periphrastic causative constructions is that the verb of the causing event cannot have any specific meaning (cf. Song 2013a). Periphrastic causative constructions using the auxiliary ‘make’ are productive in English as illustrated in (2).

In non-periphrastic (i.e., lexical and morphological) causative constructions the predicate of the causing action and the caused effect belong to the same clause. It is possible that the predicate consists of more than one verb, but the causative construction must be monoclausal. Additionally, the NP denoting the causer occupies a more prominent position than the NP depicting the causee, i.e., it is the grammatical subject. Finally, the marker of causation (be it an affix or a verb) should not have any additional meaning beyond the denotation of bare causation (cf. Song 2013b).

2 The study of causative constructions

This section aims to give a brief overview on how the structural and functional properties of causative constructions have been described and categorized in recent decades.

Nedyalkov and Silnitsky (1973) provide a cross-linguistic categorization of causatives where they define the causative situation (CS) as a macro event consisting of the micro events: the causing event (called the antecedent) and the caused event (referred there as the consequent). The relation between non-causative and causative verbs can be marked by various grammatical means. Based on the semantics of the verbs, they classify causative constructions into the following types:

1. those where a causative verb is derived from a non-causative one using morphological derivation (e.g., Bashkir: hyn- ‘to become broken’ → hyn-dyr ‘to make broken’, Nedyalkov and Silnitsky 1973: 2).

2. those where a non-causative verb is derived from a causative one using morphological derivation through a detransitivizing (or anticausative) morpheme (e.g., Russian lomat’ ‘to break (something)’ → lomat’sja ‘to become broken’ Nedyalkov and Silnitsky 1973: 2).

3. those where there are non-directed oppositions including different subtypes. In this class, the opposition between non-causative and causative verbs is ambiguous as neither verb in a pair can be considered as the base of derivation. This applies to correlative root opposition (e.g., Hungarian fordul ‘to turn (intr.)’ ↔ fordít ‘to make turn’) as well as to ambitransitive verbs (e.g., English to turn, to hang).

4. those where the causative relation is expressed by suppletion (for instance to kill and to die in English).

Nedyalkov and Silnitsky (1973: 8) describe numerous universal tendencies in the encoding of causation in their study. Firstly and unsurprisingly, individual languages can apply more than one of the strategies detailed above. Furthermore, the detransitivizing strategy (point b) is cross-linguistically less common than the transitivizing one. Additionally, they found that intransitive verbs are more likely to take a morphological marker to denote causation than transitive verbs. If a language uses morphological causatives with transitive verbs, it also uses it with intransitive verbs.

In his pioneering work on morphological causatives, Comrie (1976) assumes that the underlying structure of a causative sentence consists of a matrix clause and an embedded clause. The subject of the matrix clause indicates the causer while the subject of the embedded clause denotes the causee and depending on the argument structure of the verb can have an embedded direct object, embedded indirect object, and/or an embedded oblique constituent. Using Turkish as an example, he illustrates how the marking of the causee depends on this argument structure of the verb: the causee appears as a direct object (i.e., in the accusative, 3a) if it is intransitive, as an indirect object (i.e., in the dative, 3b) if it is transitive, and as an oblique element (i.e., in an adpositional phrase, 3c) if it is ditransitive:

He creates the following hierarchy as an abstraction upon this basis, stating that the causee’s realization shifts from left to right to the leftmost position that is not already occupied.

Subject > Direct object > Indirect object > Other oblique constituents [Comrie 1976: 263]

Song is critical of assuming universality of this hierarchy, pointing out that only few languages behave as this hierarchy would predict (Song 2001: 265). Notably, numerous languages allow for both the causee and the direct object of a transitive verb to be marked as direct objects, e.g., Udmurt as shown in example (4) and numerous other languages in our sample (see Section 6).

Comrie (1976: 276–296) classifies this as a case of syntactic doubling, which he characterizes as conventional for oblique arguments (Comrie 1976: 276) but increasing less so for indirect objects (Comrie 1976: 277) and direct (Comrie 1976: 284) objects, and essentially excluded for subjects (Comrie 1976: 294–295). Song notes that if doubling is permitted in simple non-causative clauses in a language, it is also permitted in morphological causative clauses (Song 2001: 267–268). Within our dataset, Permic languages show an inverse situation (see also Section 6): here the direct and indirect object of ditransitive verbs always have different case marking (see (5)), but double accusative marking can be found in causative constructions, see (4).

Song notes that ditransitive verbs are less likely to be causativized than transitive verbs, and transitives less likely than intransitive ones (Song 2001: 266).

Turning our viewpoint to periphrastic or syntactic causation, the fundamental difference between this type and morphological causatives is that the predicate of cause and that of effect are separate lexical verbs (Song 2001: 261). Kulikov (2001: 886–887) differentiates between syntactic causatives and non-fused biclausal ones, stating that in syntactic types “the causative morpheme is a free form, typically a verb meaning ‘cause’, ‘make’, ‘let’, ‘give’”, while biclausal sentences “refer to causative situations but do not represent cohesive units.”

Dixon’s (2000) definition of causation is based on the domain of valence-change being one of the main universal tendency to valence-increasing operations besides the usage of applicatives:

… a causative construction involves the specification of an additional argument, a causer, onto a basic clause. A causer refers to someone or something (which can be an event or state) that initiates or controls the activity. This is the defining property of the syntactic–semantic function A (transitive subject). That is, if a causative construction is formed by derivation, it will involve the addition of a new argument in A function (the causer). In the causative of an intransitive clause the original S argument (the causee) will almost always go into O function in the new transitive clause (with the causer being A). In the causative of a transitive, the causer always becomes A; then the original A (the causee) and/or O arguments generally have their syntactic functions reassigned. (Dixon 2000: 30–31)

Dixon (2000: 30–31) introduces a category he calls “Two verbs in one predicate” and divides it into two types. The first type are serial verb constructions, understood according to Aikhenvald’s definition as “a sequence of verbs which act together as a single predicate, without any overt marker of coordination, subordination, or syntactic dependency of any other sort” (Aikhenvald 2006: 1); as this type is completely unknown in the area under consideration in this paper, it will not be investigated further. In a second type, illustrated by French in example (6), the predicate includes two verbs: a finite and a non-finite one which must immediately follow one another. The causer NP takes the role of the A, while the causee NP is expressed by an oblique:

He continues to define periphrastic constructions as those where the verb of causation and the effect are in separate clauses (Dixon 2000: 35). In some languages the causee can preserve its original function, taking the role of A in the subordinate clause.

In contrast to this, in English periphrastic causatives the causee is coded as the object of the transitive causative verb:

Thus, the causee can receive marking according to its function of the subordinate clause (example (7) – Macushi) or its function in the main clause (example (8) – English), and in some languages for both functions (see Dixon 2000: 36).

Causative constructions contribute to the foundation of the valence orientation typology by Nichols et al. (2004). In this typology the expression of pairs of concepts, one plain and one induced, is investigated, e.g., to fear ↔ to frighten, to die ↔ to kill. Four basic types can be found cross-linguistically as regards dominant strategy of marking these pairs. It should be noted that languages usually employ more than one encoding strategy but yet can show a preference to one of these types (Nichols et al. 2004: 182):

1. In transitivizing languages, the plain verb tends to be unmarked while the induced verb is marked (augmented), oftentimes making use of dedicated causatives. This type is the most widespread across languages. Grammatical marking of causation can be regular and transparent (see Turkish in 3 as well as Tables 2 and 3 in Section 4) and they often have an OV order and can lack inflectional passives or other valence-decreasing operations. Uralic and Turkic languages, including those under investigation in this paper, fall into this category.

2. In detransitivizing languages (also referred to as anticausative, see e.g., Kulikov 2001), it is often the induced verb that is unmarked, e.g., Russian lomat’ ‘to break (something)’ → lomat’sja ‘to become broken’ (Nichols et al. 2004: 198); detransitivizing derivations often are employed alongside suppletion. This class is smaller but covers most of Indo-European. Languages in this group might have irregular intransitive paradigms and higher morphological complexity. They are usually accusative languages and rarely ergative.

3. In neutral languages, oftentimes both the plain and induced verb are derived; the verbal root has no inherent transitivity. Languages of this type usually employ auxiliaries and/or phrasal verbs (Nichols et al. 2004: 198).

4. In the indeterminate type, verbs are ambitransitive: there are no overt differences between the plain and induced verbs; the transitivity of a certain clause can be understood from a broader context (see also Nedyalkov and Silnitsky 1973; Comrie 1976; Song 1996).

3 The study of structural convergence in the Volga-Kama Region

The Turkic and Uralic languages of the Volga and Ural regions of the Russian Federation, while genealogically diverse (coming from different language families, and within the language families from distant branches), share a striking number of structural features. Based on these, they are frequently subsumed in the so-called Volga-Kama Sprachbund (e.g., Helimski 2003: 159) or language area. As is typical for a purported Sprachbund, one can distinguish between the core of the Sprachbund in which structural convergence is strong, and a periphery in which structural convergence is increasingly weak.

Figure 1 shows the political subdivisions of the Russian Federation associated with the relevant speaker communities. It must be noted that these political boundaries only roughly match settlement areas of the speaker communities to some extent; the geographical scopes of the individual languages are wide and overlapping. The subdivisions shown are:

Figure 1:

Titular republics and okrugs of the languages of the Volga-Kama Sprachbund (based on commons.wikimedia.org/wiki/File:Russia_-_blank_map_(2008-01).svg).

With the possible exception of Tatar, the largest minority language of Russia, all languages under consideration can, in spite of their large number of speakers (see Table 1), be considered endangered due to lacking transmission to younger generations (cf. Zamyatin et al. 2012).

Genealogically speaking, Tatar and Bashkir are two closely related Volga Kipchak languages (Berta 1998: 283; Schönig 2007: 182), while Chuvash differs greatly from all other Turkic languages, being the only extant member of the Bolgar Turkic branch of the family (Kornfilt 2018: 519; Schönig 1997: 121). As regards Uralic languages, Komi, Komi-Permyak (historically treated as a dialect of Komi), and Udmurt are comparatively closely related Permic languages (Riese 1998), while Mari is generally considered a branch of its own by contemporary scholars (Kulonen 2002: 108; Vilkuna 1998: 173).

Not only does the degree of convergence differ between languages conventionally placed within this convergence area (with Mari showing more Turkic features than Udmurt, and Udmurt showing more Turkic features than Mordvin and Komi), but also between different varieties of the languages spoken in the region:

1. Mari was subject to Bolgar influence before it broke up into its contemporary dialects. Bolgar influence is found in all varieties of Mari (Hesselbäck 2005; Saarinen 1997a: 195, 1997b: 393); Hill Mari, spoken on the right bank of the Volga in immediate proximity to Chuvashia continued to be under Chuvash influence over the centuries and shows younger Chuvash loan elements (Ivanov 1981: 87), and considerably stronger Russian influence. Meanwhile, a stronger influence of the Kipchak languages can be detected in the varieties of Mari spoken on the left bank of the Volga, especially in the varieties spoken by the Eastern Mari diaspora in Bashkortostan (Bereczki 1984: 311; Bereczki 1992–1994: I: 26; Hesselbäck 2005: 94; Moisio and Saarinen 2008: 134), which emerged as a result of the eastward migration of Maris in the 17th century (Lallukka 2003).

2. Proto-Permic, the ancestor of Komi and Udmurt, was subject to Bolgar Turkic influence (Róna-Tas 1998: 760). After the breakup of Proto-Permic around the 9–10th centuries (Bereczki 2003: 27), Komi was no longer subject to Turkic influence but rather came under Russian influence. Turkic lexemes found in Komi-Permyak can be explained through borrowing from Udmurt (Róna-Tas 1998: 760) or Russian dialects. In the case of Udmurt, younger Tatar influence is stronger in the Southern dialects (Kel’makov 1975: 95), especially the peripheral dialects (Csúcs 1998: 277). A notable exception to this trend is Beserman, a variety spoken in northern Udmurtia. It has been proposed that Beserman speakers have a Turkic ethnic background (Róna-Tas 1988: 765); their variety shows distinct Turkic influence. This means that Southern Udmurt and Beserman can be assumed to have two radically different types of Turkic adstrates: Southern Udmurt has a Tatar superstrate going back to Tatar’s previous status as the prestige language in the region, while Beserman has a Tatar substrate going back to the speaker community’s switch from a Turkic to a Permic language. The implications of these different contact linguistic circumstances have not been systematically studied.

3. In addition to genealogical distance, a Mari substrate is assumed to be one of the reasons for the great structural difference between Chuvash and other Turkic languages (Agyagási 1998: 668). Chuvash dialects are generally divided into the northern Viryal dialects and southern Antari dialects; Viryal dialects, spoken in close proximity to Hill Mari, show especially strong Mari influence (Clark 1998: 451; Johanson 2009: 245).

Another language that must be taken into consideration in areal surveys of the region is Russian, which today is the dominant language of intercultural communication throughout the region and exerts a strong influence on all languages under consideration. It should be noted however that these circumstances are comparatively recent, with Russian only becoming dominant in the central Volga-Kama Region in the second half of the twentieth century (Kangasmaa-Minn 1998: 220). For the Komi languages, spoken comparatively far in the north, Russian has been the dominant contact language for centuries.

As regards convergence between languages of the region that cannot be explained through inheritance, lexical borrowing between specific languages/branches has been subject to extensive research for over a century (Csúcs 1990; Isanbayev 1989; Räsänen 1920, 1923; Wichmann 1903); the diffusion of different loanword layers has been analyzed and compared (e.g., Saarinen 1997a, 2010). In recent decades, areal phenomena in the domain of phonology, morphology and increasingly also syntax and structural borrowings such as code copies have received increasing attention (e.g., Bradley 2016; Hesselbäck 2005; Johanson 2000; Saarinen 1997b; Wintschalek 1993).

Even in regard to recent and methodologically sound research work, criticism has been raised and room for improvement has been suggested. The differing methodological and terminological frameworks in different branches of linguistics (e.g., Uralic studies and Turkic studies) can be an impediment to unbiased comparisons, as comparable structures might be studied and described in vastly differing manners (cf. Widmer 2004). Furthermore, in spite of the great genealogical and structural distance between Chuvash and Tatar, different types of Turkic contact influence in Uralic languages are often poorly distinguished (cf. Bradley 2016: 82 about the attribution of a clear Chuvash contact phenomenon in Mari to Tatar). There is also oftentimes poor differentiation between phenomena found only in the Volga-Kama Region (i.e., true Sprachbund features) and general features borrowed into Uralic languages that can be found throughout the Turkic language family (cf. Hesselbäck 1996). There is comparatively more research into the borrowing of forms (i.e., transfer of a lexeme or affix from one language to another) than into the borrowing of functions (i.e., overlap in the range of functions in which a lexeme or affix is used), as discussed in Bereczki (2002). Finally, there is comparatively little research into the microvariation of contact phenomena, considering the unequal spread of adstrates between varieties of the languages of the region (see above).

In the paper at hand, we aim to examine one small domain of morphosyntax in the languages of the region, taking these dangers into consideration: causative constructions, and the exact range of possibilities these have in the individual languages of the region and the lexical restrictions in place. Our focus lies on Mari, Tatar, Chuvash, and two of the Permic languages (Udmurt and Komi-Permyak), with Komi-Zyrian, Bashkir, and Mordvin currently not included in our survey.

4 Valence changing operations in the Volga-Kama Region

Causation can be seen as a valence-changing operation: causative constructions are transitivizing constructions in which an additional constituent is added. In contrast, in detransitivizing constructions (e.g. reflexive constructions), the valence of a verb is reduced by one.

All languages of the region have a broad inventory of valence-changing derivational suffixes (cf. Landmann 2014a: 114–118 for Tatar, 2014b: 109–112 for Chuvash, Kalinina et al. 2006 and Bartens 2000: 288–289 for the Permic languages in general, Tánczos 2015 for Udmurt, Batalova 2002: 53 for Komi-Permyak, Riese et al. 2019: 391–395 for Mari), both valence-increasing (causativizing, transitivizing) and valence-decreasing (decausativizing, detranzitivizing). These suffixes are oftentimes so productive that they are labeled as markers for a reflexive or causative “voice” in Russian-language publications (e.g., Pengitov et al. 1961: 161–164 for Mari, Cypanov 2005 for Komi-Permyak), though this classification is problematic, partly exactly because of their productivity: there is the possibility of recursive application of causative suffixes in double causatives (see Section 6), and different valence-changing suffixes can co-occur, see (9).

Despite there being both ample transitiving and detransitiving morphology, the relative role these mechanisms play in a language deserves investigation through the lens of valence orientation (see also Section 2). A useful tool for the classification of a language is the fixed list of 18 verb pairs found in Nichols and Grünthal (2016: 30), such as fear ↔ frighten, die ↔ kill. For each language under investigation, it is determined how the causative alternation is realized for these 18 pairs to determine general tendencies of a language. As shown in Tables 2 and 3 on the example of Mari and Tatar (data from Riese et al. 2014 and Asylgarayev et al. 2007), both Uralic and Turkic languages of the region have a strong tendency towards transitivization. In both cases, however, this seems to be typical of the language families as a whole, but is in contrast to strongly detransitiving Russian (cf. Nichols et al. 2004: 197–198).

As regards their basic typological classification, the languages of the region are all in a similar position: They have both productive detransitiving and transitiving morphology but have a tendential preference towards transitivization (cf. Mikola 1999 for Uralic); morphological means are the dominant means by which causation is expressed; they are members of language families in which a preference towards transitivization is the standard; they are all subject to influence from highly detransitivizing Russian (cf. Nichols et al. 2004: 154).

5 Causative morphosyntax

This present section focuses on verbal transitivity and we aim to answer two questions set in Section 1: the productivity of intransitive and transitive morphological causation on one hand and the possibility of double causatives, i.e. the usage of multiple causative suffixes on the other.

1. Is morphological causation productive irrespective of verbal transitivity features?

2. Can more than one causative suffix be attached to a stem?

As has been shown in previous works, morphological causatives are common in the languages of the present study (Dolovai 2006; Mikola 1999). All languages under consideration use causative suffixes that seem to show superficial similarities:

1. Mari: -kt, -t/-d, -ar, -Dar ^[2]

2. Udmurt: -t

3. Komi-Permyak: -(ö)t, dialectally -d

4. Chuvash: -t, -Ar, -(t)tAr

5. Tatar: -t, -Er (-Är), -DEr (-DÄr), -GEr (-GÄr), -GEz (-GÄz), -sÄt, -Ez

In Mari, -Dar was borrowed from Turkic (cf. Alhoniemi 1985: 165) and is primarily (but not exclusively) used on borrowed stems: umylo- ‘to understand’ (Chuvash stem, Moisio and Saarinen 2008: 903) → umyl tar e- ‘to explain’, jom- ‘to become lost’ (Uralic stem, Moisio and Saarinen 2008: 184) → jom dar e- ‘to lose’. In spite of the similarity in both form and function, we are not assuming a shared ancestry for Uralic and Turkic causative -t (cf. Dolovai 2006: 153; Róna-Tas 1998: 75).

6 Argument structure of causative constructions

This section aims to answer research question (iii): What is the argument structure of a causative construction; how are the causee and the direct object of the caused action marked respectively? Can the causee and direct object of the caused action be marked by the same grammatical means?

All languages of the Volga-Kama Region have rich case morphology and have at least a nominative, genitive, dative, and accusative case. It should be noted though that in Chuvash, the accusative and dative cases have merged into one case (Landmann 2014b: 14) and that all languages under consideration except for Mari show differential object marking and allow for unmarked indefinite objects (cf. Landmann 2014b: 14 for Chuvash, 2014a: 15 for Tatar, Csúcs 1998: 283 for Udmurt, Klumpp 2009 for Komi-Permyak).

Causation increases the number of arguments within the clause. The marking of the resulting increased number of constituents is language-dependent (cf. Comrie 1976). Generally, the subject NP of a non-causative intransitive clause (20a) becomes the direct object (20b), while the direct object of a transitive verb (20c) keeps its marking and the causee is marked as the indirect object or with an oblique case (20d) (Song 2001: 263–268).

Some languages differ from the pattern mentioned above since they allow double direct objects in causative constructions (cf. Comrie 1976 for instance). This property can be a useful feature to establish differences among the languages of the area. As shown in (21c), Mari conventionally does not utilize double accusative marking.

However, grammars report double accusative marking in dialects (Alhoniemi 1993: 52), and isolated examples of this phenomenon can be found in the (literary) corpus (see Section 1), as illustrated in (22).

The exact distribution of this pattern in Mari dialects requires further investigation. Given that parallel encoding of the object and causee is precluded in Tatar, but present in Chuvash (presumably as the accusative and dative cases have merged), a point of examination could be whether double accusatives have a stronger position in Mari varieties that have been subject to stronger Chuvash influence.

In Udmurt, double accusative marking is used (Tánczos 2015: 101–102; Dolovai 2006), cf. (23).

Double accusative marking in Udmurt is restricted to causative constructions. It can be avoided in causative constructions as well as the instrumental case can in restricted contexts also be used to encode the causee (Tánczos 2015: 107–108, example (24)), but this implies factitive semantics, i.e. that the causer has less control over the action.

In Komi-Permyak, when transitive verbs are used in (periphrastic) causative constructions, two accusative-marked constituents appear in the clause, see (25).

In Chuvash, dative-accusative syncretism hampers interpretation: indirect and definite direct objects receive the same case marking. Word order plays a role in distinguishing the constituents in causative constructions: the object tends to immediately precede the verbal form and is in turn preceded by the causee, see (26).

As indefinite direct objects are unmarked, they can be distinguished from the (always marked) causee based on morphology, see (27).

When causative forms are formed from ditransitive verbs, parallel case encoding occurs again, as in (28).

A periphrastic construction can be used here as well, as shown in (29). Note three arguments receiving the same case marking.

In contrast, Tatar adheres to Comrie’s case hierarchy and does not allow double accusatives; the causee is marked with the dative case as in (30a) and (30b). In (30a), the definite object is in the accusative, while the indefinite object in (30b) is unmarked.

When causative clauses are formed from ditransitive verbs as in (31a), the causee is marked with the ablative case; native speakers might prefer the periphrastic construction shown in (31b) to morphological causation, though the case marking is the same in the morphological and periphrastic causative structure.

In summation and to answer research question (iii), of the languages under consideration only Tatar unambiguously adheres to Comrie’s case hierarchy (see Section 2). The interpretation of Chuvash data is difficult as the distinction between accusative and dative forms was lost, i.e., direct and indirect objects are not morphologically distinguished. Mari generally seems to adhere to Comrie’s case hierarchy, though examples of direct object doubling, i.e., double accusatives, seem to be widespread. Udmurt clearly allows for double accusatives in causative constructions. As the morphological causation of transitive verbs is not productive in Komi-Permyak, this question is less applicable here. However, in periphrastic causative structures, two accusative-marked constituents can co-occur in a clause.

7 Quasi-causative constructions

This section pertains to research question (iv), the productivity and lexical restrictions of quasi-causative constructions.

The so-called quasi-causative construction consists of a causative derivation of a verb denoting some sort of physical or mental state or change and a single argument marked as the direct object, i. e. with the accusative in the languages under consideration (F. Gulyás 2016: 171–175). This structure is well-known in certain Finno-Ugric languages, e.g., Finnish (see (32)), and are described as impersonal constructions in these (Stipa 1962).

Quasi-causative constructions can be analyzed in two ways: one can consider the argument of the verb a non-canonical subject (cf. Sands and Campbell 2001: 288–289 for Finnish) or the clause can be treated as a subjectless one. However, it is clear from the translations of examples 32–38 that the argument marked by the accusative (dative-accusative in Chuvash and partitive in Finnish) bears the semantic role of experiencer. Therefore, the number of verbs that can appear in this construction type is restricted: they all denote physical or mental states or changes in such states.

Quasi-causative constructions can be found in other Uralic languages throughout the region, see examples (34)–(35).

Our data show some similar instances from the neighboring Turkic languages, see (36) and (37). This is notable as these structures do not seem to be commonplace in Turkic; see below.

These constructions can also be found in Uralic languages outside of the region, e.g., Finnish (see (32) above). There is also a superficial similarity to an impersonal construction found in Russian (see (38)), though Russian, a strongly detransitivizing language (cf. Nichols et al. 2004: 154), does not employ causative suffixes.

In summation and to answer research question (iv), quasi-causative constructions are commonplace in all languages under consideration. However, their areal and genealogical distribution requires further investigation. To the best of our knowledge, they cannot be found in more distantly related Turkic and Slavic languages such as Turkish, Uzbek, and Slovak, but our data does not at this point enable us to make conclusions on an areal component to the spread of this structure.

8 Summary

In the previous chapters we illustrated some morphosyntactic properties of causatives in the Uralic and Turkic languages of the Volga-Kama region and the manner in which these languages differ and resemble each other in this domain. We summarize our findings in Table 4.

With the exception of Komi-Permyak, all languages of the region productively use causative suffixes on transitive stems. Also, all languages under consideration but Komi-Permyak allow for causative derivations to receive an additional causative suffix.

Case marking of the constituents (especially the causee and the direct object of a transitive verb) in causative constructions varies from language to language in our sample. While the two Permic languages clearly allow for double accusatives, they are completely absent from Tatar. Standard Mari also seems to lack double accusatives, though counterexamples can be found; it seems plausible that some dialectal variance can be found in this domain. In Chuvash, examples can be found in which the causee and the direct object are both in the dative-accusative case, but syncretism of the accusative and dative cases impedes the interpretation of these cases.

We can see that the Uralic and Turkic languages of the Volga-Kama region have the so-called quasi-causative construction. Since this clause type is not reported from Turkic languages outside the linguistic area, the question arises whether it can be an instance of code-copying from Uralic to Turkic, but further research is needed. In Uralic, quasi-causatives are found in Permic and Mari but are absent from Ugric languages (Hungarian, Mansi, Khanty, cf. F. Gulyás 2016). To make the picture more complicated, Finnish employs quasi-causative in a productive manner (Sands and Campbell 2001). Another possible explanation is that this clause type is due to Russian influence, but this cannot account for the Finnish data. The lacking causative suffixes in Russian constructions of this type are a further complicating factor.

Further research into these questions would require both a wider and more narrow scope: it is both necessary to examine the structural convergence within the area through the lens of the language families as a whole to form better hypotheses on the origin of structural similarities such as the usage of quasi-causative constructions in the region, and to examine the dialectal distribution of competing structures within a language (most prominently different strategies of constituent marking in Mari: if double accusative marking is permissible) to determine if differences between varieties correlate with differing historical language contacts.

Abbreviations

1

first person

2

second person

3

third person

abl

ablative

acc

accusative

caus

causative

com

comitative

dat

dative

detrans

detransitivization

erg

ergative

frust

frustrative

fut

future

gen

genitive

inf

infinitive

inst

instrumental

nec

necessitive

obj

objective conjugation

part

partitive

pass

passive

pfx

verbal prefix

pl

plural

prep

preposition

prs

present tense

pst

past tense

ptcp

participle

px

possessive suffix

sg

singular