Interaction between grammatical phenomena and information structure in Northern Mansi

全文:

Introduction

The aim of this article is to provide an overview of grammatical phenomena influenced by information structure in Northern Mansi (< Ob-Ugric < Ugric < Finno-Ugric < Uralic¹). Ob-Ugric languages present a unique case of grammatical systems heavily governed by information structure² [Klumpp, Skribnik 2022]. It is reported that in Khanty and Mansi varieties, voice [Skribnik 2001; Filchenko 2012; Virtanen 2015; 2023; É. Kiss 2019; Urmanchieva, Plungian 2021; Muravyev, Zhornik 2024], object agreement [Nikolaeva 1999; Koshkareva 2002; Skribnik 2004; Virtanen 2015; É. Kiss 2019; Klumpp 2023], syntax of ditransitive constructions [Virtanen 2015; Bíró, Sipőcz 2017], and use of possessive markers in non-possessive functions [Janda 2015; 2019] are determined at least in part by the information-structural status of the core participants.

A recent overview of information structure in Uralic [Klumpp, Skribnik 2022] lists the following phenomena related to information structure in Northern Mansi: possessive marking (3sg in non-possessive function), object agreement, passive as a “heavy-duty topicalizer”, dative shift, pro-drop (3^rd person). The main research question with regard to these categories is to determine how speakers make the choice between different categories (e.g. subject vs. object conjugation for describing the same situation) and what factors influence this choice.

In this paper I will, on the one hand, provide an overview of existing literature on information structure in Northern Mansi and Mansi in general. On the other hand, I will describe some of the phenomena connected to discourse structure in Northern Mansi that have not been previously studied as well as suggest a broader approach encompassing both information structure and a wider range of discourse-related parameters. The space allocated to each phenomenon in this article reflects the significance of their impact on the whole discourse system as well as the availability and thoroughness of previous research on the topic.

The structure of this article is as follows: In Section 1, I briefly introduce the Northern Mansi language. In Section 2, I discuss object agreement, passive voice, and variation of participant coding in ditransitive constructions. In Section 3, I describe phenomena that have been less researched, such as non-possessive uses of 3sg possessive marker and prosody, or not studied at all, such as zero-reference and particles ta and ti. In conclusion, I present a summary of the grammatical phenomena that interact with information structure in Northern Mansi and propose directions for future research in this domain.

1. The Northern Mansi language

The Northern Mansi language is the only living Mansi variety. Together, Khanty and Mansi languages form the Ob-Ugric subgroup of the Ugric branch of the Finno-Ugric language group within the Uralic language family. Previously, three other Mansi varieties existed: Eastern Mansi, Western Mansi, and Southern Mansi. Each of these languages had an individual history of language contact, and they had significant phonetic and grammatical differences (e.g. in the number of cases), leading to limited mutual intelligibility. There are different views on the exact number of Mansi languages. Some researchers distinguish only two or three based on lexicostatistical data, see [Koryakov 2017: 94; Koryakov et al. 2022: 74–75; Fedotova 2021].

Northern Mansi is spoken in Russia, in two neighboring regions of Western Siberia: Khanty-Mansi Automonous Okrug (Yugra) and Sverdlovsk Oblast’. According to the latest 2020 All-Russian Population Census³, 1346 people claimed Mansi⁴ language proficiency and 1008 reported Mansi language use. Native speakers and local researchers⁵ estimate that the real number of speakers is unlikely to be higher than a thousand, possibly, it is even lower. Northern Mansi is considered a critically endangered language, see, e.g., [Skribnik, Koshkaryova 1996; Bíró, Sipőcz 2009].

The Northern Mansi language is divided into several dialectal varieties: Upper Lozva, Sosva, Sygva, and Ob’. In this article, I will provide examples from two Northern Mansi varieties: Upper Lozva and Sosva. Information structure and the grammatical phenomena in question are quite similar in different Northern Mansi varieties (although there is variation across Mansi languages, see, e.g., a study on Western Mansi passive in comparison with other Mansi varieties in [Urmanchieva, Plungian 2021]), so the exact choice of dialect is not crucial for this paper. The main source of examples for this paper is Valery Chernetsov’s archive [Ryndina et. al 2022], containing Northern Mansi narratives from 1925–1938. Apart from that, examples from my personal field data⁶, studies on Mansi syntax and information structure, as well as published folklore collections [Rombandeeva 2005; Popova, Rombandeeva 2010; Kumaeva 2015] are used. The exact source is cited next to the translation of each example, except for the elicited ones, and the spelling of the original source is preserved.

2. Argument expression

2.1. General information

Ob-Ugric languages have an argument coding system which includes differential object agreement, passive voice, and variation of participant coding in ditransitive constructions. Existing works assume that these phenomena are governed to a significant degree by information structure and that different coding options represent various information-structural configurations. In Sections 2.2–2.3, I will explore previous research in this area and discuss factors responsible for the choice of argument expression strategies.

Object agreement is expressed morphologically in a dedicated subject-object agreement para digm (2), distinct from the subject paradigm (1).

(1) am xūl puw-s-um

1sg fish catch-pst-1sg

‘I was fishing / I caught a fish.’

(2) am ti xūl ta puw-əs-l-um

1sg this fish ptcl catch-pst-sg.o-1sg.s

‘I finally caught this fish.’

In a passive construction, the O-participant⁷ is promoted to subject and controls verb agreement, the verb receives the passive marker -we, and the A-participant is optionally expressed in the lative case, cf.:

(3) am ti-xōtal alpǝl kw̄al-s-um kwāl-tt-uwe-s-um ruɕ

1sg this-day morning get.up-pst-1sg get.up-caus-pass-pst-1sg Russian

pīɣ-na

youngster-lat

‘Today I woke up in the morning, the Russian guy woke me up.’ [ULMC: AEJ_DZh_SP_310118_myday]

Passive in Ob-Ugric is very productive, even intransitive verbs can be passivized and participants other than those expressed as the direct object in an active construction can be promoted to the subject position, as in (4), where a participant with the role of location is promoted to subject.

(4) a. xājtnut nēɣl-ap-as mān ēl-i-pāl-uw-t

wolf appear-mom-pst[3sg] 1pl front-attr-side-poss.1pl-loc

‘A wolf appeared in front of us.’

1. mān xājtnut-n nēɣl-ap-awe-s-uw

1pl wolf-lat appear-mom-pass-pst-1pl

‘A wolf appeared in front of us’ (lit. “We were appeared by a wolf”). [ULMC: AAA_DZh_SP_300118_wolfs]

The third coding alternation involved in this system is the so-called “dative shift” in ditransitive constructions, where there is variation in participant encoding, see [Bíró, Sipőcz 2017]. One option is to mark the recipient as a direct object, with the theme receiving instrumental case marking; the other is to mark the theme as a direct object, and the recipient as an indirect object in the lative case. This also interacts with the choice of subject vs. object agreement: in the first case both subject and object agreement can be used (in both cases S-IO-DO and S-DO-IO word orders are possible), see (5a)–(5b) from [Nikolaeva 2001], while in the latter case the object agreement is usually preferred, as in elicited (5c).

(5) a. am naŋǝn-n sūp jūnt-ēɣ-ǝm [Nikolaeva 2001: 24]

1sg 2sg.obl⁸-lat shirt sew-npst-1sg

1. am sūp naŋǝn-n jūnt-i-l-um [Nikolaeva 2001: 24]

1sg shirt 2sg.obl-lat sew-npst-sg.o-1sg.s

1. am naŋǝn sūp-ǝl jūnt-i-l-um

1sg 2sg.obl shirt-ins sew-npst-sg.o-1sg.s

‘I sew a shirt for you.’

Ditransitive verbs can also occur in a passive construction, and in this case the recipient is promoted to subject and the theme bears oblique case marking; the agent is usually omitted, see:

(6) n’āl-ǝl wār-we-s jōwt-ǝl wār-we-s

arrow-ins make-pass-pst[3sg] bow-ins make-pass-pst[3sg]

‘An arrow and a bow were made for him.’ [Kálmán 1976: 68]

Thus, in the case of constructions with ditransitive verbs we have at least four syntactic options of participant encoding. As we will see in 2.2, previous research indicated that the choice of a particular construction in each case is determined at least in part by pragmatics.

2.2. Previous research

Recent research claims that the coding system outlined above is largely pragmatically driven. This view arose around the 1960–70s. The Ob-Ugric voice has attracted attention of researchers since the first half of the 20^th century, see, e.g., [Bouda 1936]. The first accounts of Ob-Ugric and Mansi passive voice [Lavotha 1960; Honti 1982; Schiefer 1985] tried to approach this phenomenon from a semantic standpoint and determined different types of constructions depending on semantic roles of the core participants, most notably, see [Kulonen 1989]. In the studies of Evdokiya Rombandeeva, a native Mansi speaker and linguist, we find the first ideas that the choice of voice and agreement may be affected by definiteness of the constituents and the emphasis put on them, see [Rombandeeva 1979]. The early 2000s marked the beginning of the information-structural turn in Uralic studies, starting with the works of Irina Nikolaeva [1999; 2001], where she introduced the notion of secondary topic and tied passive and object agreement to the information structure in Khanty. This approach was developed by Elena Skribnik [2001; 2004] specifically for Mansi, as well as by Natal’ya Koshkareva [2002] for Khanty. This line of analysis ties agreement, voice, and variation of participant coding in ditransitive constructions into a single system, which allows for a more comprehensive description of argument expression in Ob-Ugric.

According to this line of research, argument coding in Ob-Ugric is closely tied to information structure, and different coding options represent various information-structural configurations. An active construction is used when the main topic of the utterance corresponds to the A-participant. Subject agreement is used when the O-participant is in focus, while the object agreement is triggered when the O-participant bears the information-structural role of secondary topic in Nikolaeva’s terms. Passive voice is used when the O-participant, not the A-participant, is the most topical constituent in a clause; this category is viewed as the main topicalizing device in Ob-Ugric. When analyzed more broadly, beyond the scope of a single clause, passive can be viewed as a means of maintaining the current discourse topic by promoting the topical participant to subject when it is not agentive, as in (7).

(7) {The heroes were hunting in the forest on their own, then they met Mis-hum⁹ and he decided to bring them to his village.}

ta꞊kōs poxt-as-iɣ misxum ta꞊min-i ta꞊tot-awe-ɣ

ptcl꞊conc hold-pst-3du Mishum ptcl꞊go-npst[3sg] ptcl꞊carry-pass-3du

‘They held tight (onto Mishum), Mishum is going, and they are being carried (to his village).’ [Chernetsov’s archive 50, text 2]

Variation in ditransitive constructions is explained in the same terms. The coding can change depending on the alternations of information structure, as in the following example from [Kálmán 1976: 68] cited in [Bíró, Sipőcz 2017], which demonstrates three consecutive utterances in a narrative. In (8a), there is no topical object, thus subject agreement is used. In (8b), the recipient is topical and promoted to direct object, which triggers object agreement. In (8c), the main hero (the mythical ēkwa pīɣriɕ, who is asking for an arrow and a bow) is the most topical constituent; thus, he is promoted to subject via passive voice.

(8) a. akwēkw anǝm-n n’āl wār-en jōwt wār-en

aunt 1sg.obl-lat arrow make-imp.2sg bow make-imp.2sg

‘Auntie, make me an arrow, make me a bow.’

1. n’āl-ǝl wār-i-l-ǝm jowt-ǝl wār-i-l-ǝm

arrow-ins make-npst-sg.o-1sg.s bow-ins make-npst-sg.o-1sg.s

‘I will make you an arrow, I will make you a bow.’

1. ja-t’i n’āl-ǝl wār-we-s jowt-ǝl wār-we-s

well-ptcl2 arrow-ins make-pass-pst[3sg] bow-ins make-pass-pst[3sg]

‘Well, an arrow and a bow were made for him.’ [Kálmán 1976: 68]

After these theoretical developments, numerous works on various Khanty and Mansi languages have appeared, where the authors analyze the phenomena of active vs. passive voice, object vs. subject agreement and variation in ditransitive constructions. Apart from Irina Nikolaeva’s work on Obdorsk Khanty [Nikolaeva 2001; Dalrymple, Nikolaeva 2012], the main research in this domain for Khanty varieties was carried out by Andrey Filchenko [2006; 2007; 2012] based on Eastern Khanty data. Studies on Mansi languages are even more numerous: Susanna Virtanen [Virtanen 2015] tied all the aforementioned phenomena into a unified system for Eastern Mansi, Anna Urmanchieva and Vladimir Plungian analyzed passive in Western Mansi based on archival data gathered by Artturi Kannisto in 1901–1905 [Urmanchieva, Plungian 2021]. Bernadett Bíró and Katalin Sipőcz performed extensive research on ditranstive constructions in Ob-Ugric in general [Sipőcz 2015] as well as specifically in Mansi with main focus on Northern Mansi [Bíró, Sipőcz 2017]. Finally, Katalin É. Kiss analyzed voice and agreement phenomena in Ob-Ugric from a generative standpoint, see, e.g., [É. Kiss 2019] for an analysis of the passive as a means of demoting the [-topic] highest argument and promoting a [+topic] internal argument.

These works sought to apply the information-structural approach with topic-promotion as the main mechanism behind the phenomena in question to data from individual Ob-Ugric varieties or to describe the interaction between various parts of the argument coding system in more detail. In recent years it became apparent that, even though the information-structural approach has immensely deepened the understanding of argument coding in Ob-Ugric, there are numerous exceptions to the topic-promotion rule, at least in some Ob-Ugric varieties, specifically in Northern Mansi. For example, (9) is the first sentence of the text, where there is no topical constituent whatsoever, and still, passive is used. This might be explained by the need to demote an unknown or irrelevant A-participant in a generic context. However, such contexts are not described in the literature which considers topic-promotion to be the main function of the passive as existing exceptions.

(9) sāŋkwəltap jīw-əl wār-awe

sankwyltap wood-abl make-pass[npst.3sg]

‘A sankwyltap¹⁰ is made of wood.’ [ULMC: ASN_DZh_SP_310717_sankwiltap]

In (10) we see an example from a similar type of text, where generic actions during different holidays are discussed. The “wedding” is mentioned for the first time and is still promoted to subject via passivization.

(10) tuwəl os pūri pūri wār-awe

then add wedding wedding make-pass[npst.3sg]

[Also a holiday is celebrated when a child is born.] ‘Then also a wedding is celebrated.’ [ULMC: BTP_DZh_140217_prazdniki]

Even though existing literature did not suggest any differences in the use of passive voice among various genres, we might suspect that texts with generic instructions function differently than narratives, and most of the examples in the existing literature come from narrative texts. However, non-topical subjects of passive constructions are found in narrative texts as well. Example (11) contains the first clauses of a narrative text, and in the third clause we find passive, although the subject “sable” is also mentioned for the first time and cannot be topical, whereas the omitted agent is topical, as the two forest men who chased down the sable are the main heroes of the text.

(11) kit wōr xum ōl-ē-ɣ wōr-n min-as-iɣ akw mā-t

two forest man live-npst-3du forest-lat go-pst-3du one place-loc

n’oxs xāŋxt-uwe-s

sable chase.down-pass-pst[3sg]

‘Two men live in the forest. They went to the forest. In one place they chased down a sable.’ [They killed the sable and went further. In one place they chased down another sable.] [Chernetsov’s archive 50, text 2]

Such examples are not that rare, and they suggest that the information-structural status of the core participants is a very significant but not the only parameter influencing the choice of argument coding in Ob-Ugric. In the most recent works, attempts to identify other parameters were undertaken, in [Muravyev, Zhornik 2023; 2024] for passive in Ob-Ugric, in [Muravyev 2023] for passive in Northern Khanty, in [Klumpp 2023] for differential object marking in West Mansi¹¹. We present this new line of analysis in the next section.

2.3. Parameters complementing the information-structural approach

The recent approach to studying argument coding in Ob-Ugric outlined in [Klumpp 2023], [Muravyev 2023], and [Muravyev, Zhornik 2024] considers information structure to be an important factor but also stresses the need to include other parameters for a more comprehensive analysis. This is in line with recent typological developments, such as in [Just 2024], where differential A- and O-indexing is analyzed cross-linguistically with regard to identifiability, animacy, and topicality of the core arguments. For Finno-Ugric direct objects, a similar set of parameters was outlined in [Serdobolskaya, Toldova 2012] and [Toldova, Serdobolskaya 2017].

In [Klumpp 2023], we find a broader set of parameters, including (but not limited to) possessor marking, givenness, animacy, the syntactic function of the antecedent, distance from it, and word order. Applying frequency analysis, the author concludes, for example, that dative-lative marking in West Mansi appears on given objects referred to with a full noun phrase and occurs significantly more often with referents of high animacy and referents with direct givenness (and not, e.g., givenness via a given possessor). Such conclusions are not possible within a purely information-structural analysis, and a broader set of parameters might account for the “exceptions” we encountered earlier. In [Muravyev 2023], a similar set of parameters is used for analyzing the choice of voice in Northern Khanty. The author concludes that at least givenness and animacy, apart from topicality, influence the use of passive voice, and that other parameters may be added in the future to enable more comprehensive studies of these phenomena in Ob-Ugric.

In [Muravyev, Zhornik 2023], we made a first attempt of creating a parametric voice model in Ob-Ugric, where we introduced the parameters of animacy, definiteness, and topicality (discourse topic as described in [Givón (ed.) 1983]) for both A- and O-participants, as well as volitionality for the A-participant and affectedness for the O-participant. However, it was lacking some important parameters accounting for the use of voice in Ob-Ugric. In [Muravyev, Zhornik 2024], we modified this approach by adding a more fine-grained set of parameters related to the referential status of the core participants, see parameters in [Gundel et al. 1993], as well as definiteness according to [Roberts 2003], and specificity according to [von Heusinger 2002]. We concluded that the use of passive is driven not by information structure alone but by a complex interaction of various properties of both core arguments. This approach can explain cases of A-demotion when A is specific indefinite, non-specific indefinite, generic, or definite new and P is not topical; hence, it has no apparent reason to be promoted. We also discovered that different parameters have various significance for the A- and O-participants, for example, animacy is a more significant parameter for the A-participant than for the O-participant. Based on the combination of all relevant factors, we proposed the following hierarchy of passivization for Ob-Ugric (the higher a participant is on the hierarchy, the more likely it is to be promoted in case of P or demoted in case of A):

A-demotion cline: Inanimate > Non-specific or generic > Specific indefinite > Definite non-topical > Topical > Speech act participant

P-promotion cline: Speech act participant > Topical > Episode-central > Definite > Weak definite > Specific indefinite > Non-specific or generic

This hierarchy is a preliminary attempt to unify within one scheme all the relevant parameters, namely person, animacy, topicality, definiteness, and specificity, resulting in a combination of the prominence hierarchy (see prominence hierarchies¹² in [Aissen 2003]) and the referential hierarchy¹³ (see [Gundel et al. 1993]), which also accounts for the fact that some parameters are more relevant for one participant than the other.

Moreover, we noticed that text genre plays a significant role in argument coding. In (9) repeated below as (12), we encountered an example from a procedural text, containing instructions on how to create a traditional Mansi instrument.

(12) sāŋkwəltap jīw-əl wār-awe

sankwyltap wood-abl make-pass[npst.3sg]

‘A sankwyltap is made of wood.’ [ULMC: ASN_DZh_SP_310717_sankwiltap]

In this text, there is no definite A-participant, and the O-participant is generic, as well as the actions described in the text, and the discourse structure is significantly different from the one we encounter in narratives. Our recent research suggests that in narratives such factors as episode boundaries, narrative roles of the participants (protagonist, antagonist, etc.), and other plot characteristics may influence argument coding. However, this is a topic for future studies.

Thus, the current goal of further research on argument coding in Ob-Ugric is to determine the set of factors that influence the choice of voice and agreement and to provide a comprehensive description of the mechanism discussed in this section. This type of research requires a unified annotation for Ob-Ugric narrative texts, which would then facilitate a qualitative analysis. Since text type influences argument coding, it would seem fruitful to zoom in on the use of these phenomena in narratives, as they are the most abundant type of material for Northern Mansi and Ob-Ugric in general, and then expand the analysis to other genres, such as dialogues and procedural texts.

3. Less-researched phenomena

Phenomena described in this and the next section have received much less attention than Ob-Ugric argument coding, and they probably exert less influence on the whole linguistic system. Nevertheless, they are important for understanding the general discourse structure of Northern Mansi and are mentioned in the overview [Klumpp, Skribnik 2022]. These categories include non-possessive uses of the 3sg possessive marker, zero-reference, discourse particles, and prosody. With regard to prosody, an analysis of audio data is required, hence I am not able to provide any insight into that domain as of yet. Some initial research results specifically on the prosodic marking of focal constituents in Northern Mansi were outlined in [Pokrovskaya 2022], where it was concluded that the focal constituent is marked by pitch rising on the stressed syllable of the corresponding word. Apart from that, no research on the interaction between Northern Mansi information structure and prosody has been carried out.

3.1. Non-possessive functions of poss.3sg

Northern Mansi, as other Uralic languages, exhibits a set of possessive markers for each person and number combination. They can be used to express the usual range of possessive semantics (13). However, the poss.3sg marker -(t)e is also used in so-called “non-possessive” contexts, see [Janda 2015], where the presence of the possessive marker is determined by the information-structural context and no possessive relation is present (14):

(13) ı̯aɣ-āɣi-te kēr sāli-ūı̯ χum-ı̮ɣ ɔ̄ńꞩ-i-t-e

elder-girl-poss.3sg iron reindeer-animal man-trans have-npst-sg.o-3sg.s

‘His sister has the Iron Wolf as her husband.’ [Chernetsov’s archive 50, text 1]

(14) {The Iron Wolf wants to eat the main hero. The main hero was in the bathhouse.}

kon ta꞊kwal-ap-ɒs, sar lāꞩal,

outside ptcl꞊go.out-mom-pst[3sg] hort slow

uı̯-e nupəl lāw-i

animal-poss.3sg towards say-npst[3sg]

‘He went outside, “Wait a moment”, he says to the beast.’ [Chernetsov’s archive 50, text 1]

This function of possessive suffixes is attested in most Uralic languages (see [Janda et al. 2022; Fraurud 2001; Nikolaeva 2003; Simonenko 2014]). However, the range of contexts where such possessives are used as well as which person and number markers can be used in a non-possessive function differ, see, e.g., [Khomchenkova 2022] for Hill Mari or [Serdobolskaya et al. 2017] for Beserman Udmurt. In Northern Khanty, the poss.2sg marker can be used outside of possessive contexts [Mikhailov 2023; 2024], while in Northern Mansi such uses are only possible for the poss.3sg marker, see [Janda 2015]. Previous research referred to these non-possessive uses as definiteness marking, see [Gerland 2014]. In [Janda 2015; 2019], however, it is concluded that the Northern Mansi poss.3sg marker is a referential device used for information structure purposes. In some other Uralic languages (e.g. Hill Mari, see [Khomchenkova 2022]) possessive markers are used to denote selection from a set, and in Northern Mansi we also find such contexts, see (15) adapted from [Janda 2015], where in (15a) and in the text before this sentence two men are considered as a set, and in (15b) only one member of this set is being referred to:

(15) a. kittiɣ ta꞊min-as-iɣ

apart ptcl꞊go-pst-3du

‘They went their own ways’.

1. mān’ɕi jāniɣ ōjka-te noms-i

Mansi big man-poss.3sg think-npst[3sg]

‘The big Mansi man is thinking.’ [Janda 2015: 254]

It seems important to further research the connection between non-possessive uses of the poss.3sg marker and the information structural status of the constituents, as well as determine other parameters relevant for the use of poss.3sg.

3.2. Zero-reference to 3^rd-person participants

In Northern Mansi narratives, full noun phrases and even pronouns are quite rarely used for referring to a participant, and zero-reference is common (see [Klumpp, Skribnik 2022: 1033] for a discussion of this phenomenon in Uralic under the notion of “pro-drop”). Full noun phrases are usually used to introduce a participant for the first time (16), to reactivate it when it has not been mentioned for a long time (17), or to distinguish between participants when several of them perform actions in the current episode (18).

(16) {As the first sentence of the text:}

ı̯a an χum ı̯aɣ-āɣi-tēn-təl ɔ̄l-èı̮ɣ

so one man elder-girl-poss.3du-ins live-npst.3du

‘So one man lives with his sister.’ [Chernetsov’s archive 50, text 1]

(17) {The Iron Wolf was mentioned once over 100 clauses ago.}

taχ kēr sāli ūı̯-n ta꞊al-awe-m taχ

fut iron reindeer animal-lat ptcl꞊kill-pass[npst]-1sg fut

‘(It seems that) the Iron Wolf will kill me.’ [Chernetsov’s archive 50, text 1]

(18) {The animal companions are trying to remove a fang which killed their master from his ear and change turns.¹⁴ }

1. sɔ̄rup ɔ̄ı̯ka ta꞊sipɣ-i

moose man ptcl꞊suck-npst[3sg]

‘The moose is sucking (the fang out).’

1. l’eu̯ ı̱ꞩwer sipɣ-i

lion animal suck-npst[3sg]

‘The lion is sucking (the fang out).’

1. tuwəl raꞩ ɔ̄ı̯ka sipɣ-uŋkwe pat-s

then lynx man suck-inf begin-pst[3sg]

‘Then the lynx started to suck (the fang out).’ [Chernetsov’s archive 50, text 1]

The 3^rd person pronouns are used even rarer, usually to put emphasis on the participant, e.g. to contrast it with another participant, see (19).

(19) {The hero is washing in the bathhouse to delay being eaten by the Iron Wolf. A raven comes to tell him that his animal companions are coming to save him, and he needs to buy a little more time.}

χulaχ-e ta꞊min-as tau̯ ta꞊lou̯t-χat-i

raven-poss.3sg ptcl꞊go-pst[3sg] 3sg ptcl꞊wash-refl-npst[3sg]

‘The raven left, and he (the main hero) continues to wash himself.’ [Chernetsov’s archive 50, text 1]

Thus, most of the topical and activated participants in a Northern Mansi narrative are referred to with zero-reference. In the literature, it is usually suggested that topical constituents enable such reduced referential means. However, in Northern Mansi both participants of a transitive clause are frequently referred to with zero-reference, and the participants are only encoded within the verb form via subject and object marking, as in (20).

(20) kolas ńakw-əl ta꞊titt-i-t-e ti꞊titt-ǝs-t-e ti꞊pusm-ǝs

flour dust-ins ptcl꞊feed-npst-sg.o-3sg.s ptcl2꞊feed-pst-sg.o-3sg.s ptcl2꞊recover-pst[3sg]

‘(The brother) is feeding (his sister) with flour dust, (he) fed (her), (she) recovered (from her illness).’ [Chernetsov’s archive 50, text 1]

In Northern Mansi, zero-reference is a means of tracking a highly activated or topical referent and it is intertwined with the system of voice and agreement (see also [Toldova, Volkova 2021] for Northern Khanty), whereby, for example, a topical constituent can be promoted to subject in passive construction and also be referred to with zero anaphora to signify its highly activated status, see (21).

(21) {The main hero was killed by his sister.}

ı̯aɣ-aɣi-tè-n ta꞊tot-wè-s kuꞩńèꞩ pālt

elder-girl-poss.3sg-lat ptcl꞊bring-pass-pst[3sg] blacksmith to

‘His sister brought him to the blacksmith (to put him inside a barrel).’ [Chernetsov’s archive 50, text 1]

Thus, it is important to research the connection between the choice of voice and agreement and referential means used to refer to the participants.

3.3. Particles ta and ti

Another phenomenon in Northern Mansi possibly pertaining to information structure includes some of the particles. Northern Mansi has a large number of various particles, and it has been suggested that some of them attach to focal constituents, see examples (22) and (23) (cited in [Klumpp, Skribnik 2022: 1021]), where the conditional particle ꞊ke is attached to the elements under focus:

(22) am꞊ke sārt tāɣint-i naŋ mājt-ǝn am lēwat-ēɣ-ǝm

1sg꞊cond first fill-npst[3sg] 2sg liver-poss.2sg 1sg pull.out-npst-1sg

‘If MINE (yarnwinder) will be pulled first, I will pull out your liver.’ [Kannisto, Liimola 1951: 205]

(23) juw joxt-ēɣ-ǝn akʷ-ǝn mat-ǝs꞊ke

to.home come-npst-2sg aunt-poss.2sg age-pst[3sg]꞊cond

manǝr wār-i-l-ǝn

what do-npst-sg.o-2sg.s

‘When you come home, if your aunt has become old, what will you do (with her)?’ [Chernetsov’s archive 44, text 5]

However, in this case only the position of the particle is determined by the information structure, not its presence or absence. In my research, I encountered other particles that may be dependent on certain information-structural, or at least broader discourse properties of a clause or a chunk of text where it appears. Two particles exhibit this behavior: ta, which is very frequent, and ti, which is significantly rarer. These particles are diachronically connected to the demonstrative pronouns ta ‘that’ and ti ‘this’. However, the pronouns and particles can be discerned by their position. The demonstratives appear as dependents in a noun phrase, while the particles usually appear as proclitics before verbs, as in (24), and they even can be inserted between a preverb and a verb, see (25).

(24) ti l’ɔ̄ŋχ χɔ̄sit ti꞊tot-ima

this way along ptcl2꞊bring-mir.pass[3sg]

‘He was brought along this way.’ [Chernetsov’s archive 50, text 1]

(25) tuwǝl tān xot꞊ta꞊tip-s-ǝt

then 3pl off꞊ptcl꞊get.lost-pst-3pl

‘Then they got lost.’

Particles are barely mentioned in existing literature on Northern Mansi: grammars [Rombandeeva 1973] and [Riese 2001] only provide a short list labeling them as “specifying”. In the latest overview [Bakró-Nagy et al. 2022], they are not mentioned at all. Several pages in [Saynakhova 1966: 179–185] are dedicated to the particles ta and ti. However, examples given in the dissertation include the demonstrative pronouns ta ‘that’ and ti ‘this’ and do not provide a clear picture of the functions of the particles.

My analysis showed that the particle ta is extremely frequent in narrative texts: in the narratives from Valery Chernetsov’s archive (containing mostly Mansi texts from 1925–1938), ta is present on average in 25 % of clauses¹⁵. One of the more frequent contexts where it appears includes two consecutive clauses, which are usually identical except for the presence of ta in the first one, see (26) and (27).

(26) a. aku̯ ı̯ot ı̯uu̯ ta꞊min-as-ı̮ɣ

one com to.home ptcl꞊go-pst-3du

‘(They) went home together.’

1. ı̯uu̯ min-as-ı̮ɣ, ı̯un χūl-s-ı̮ɣ

to.home go-pst-3du home stay.overnight-pst-3du

‘Went home, spend the night at home.’ [Chernetsov’s archive 50, text 1]

(27) a. ı̯uu̯꞊ta꞊ꞩalt-s-ət

home꞊ptcl꞊enter-pst-3pl

‘(They) went inside.’

1. ı̯uu̯ ꞩalt-s-ət, uı̯-an-e kolkan-na ı̯ol꞊ta꞊χuı̯-lı̮ɣt-as-ət

inside enter-pst-3pl animal-pl-poss.3sg floor-lat down꞊ptcl꞊lie.down-mom-pst-3pl

‘(They) went inside, his animals laid down on the floor.’ [Chernetsov’s archive 50, text 1]

These instances can be classified as recapitulative bridging linkage constructions [Guérin, Aiton 2019], which include verbatim or very close repetition of the preceding clause before a new chunk of narrative. In this case, ta marks the beginning of a bridging construction, which might signal a turn in the narrative or a beginning of a new episode. However, this is only one type of context where ta can appear and further research on the role of this particle is needed. Rarely, such repetition is possible without the particle ta, but in most cases it is present. It seems that in cases where the particle is absent, we find a more stative, non-perfective situation, see (28).

(28) a. ı̯aɣ-āɣi-te ı̯un ɔ̄l-i

elder-daughter-poss.3sg at.home be-npst[3sg]

‘His older sister is at home.’

1. ı̯un ɔ̄l-i, akw ēr-t kēr sāli-ūı̯

at.home be-npst[3sg] one moment-loc Iron reindeer-animal

ꞩarit-ime āraꞩ-n tuu̯ min-ās

burn-ptcp.res fire-lat there go-pst[3sg]

‘(She) is at home, at one point she went to the fire where the Iron Wolf was burnt.’ [Chernetsov’s archive 50, text 1]

The particle ti is much rarer, ranging approximately between 5–7 % of the clauses, e.g. in a text “The tale about the Iron Wolf” from Valery Chernetsov’s archive among 390 clauses there are 122 instances of ta and 24 instances of ti. The functions of ti (see (29)) are even less clear than those of ta and require further investigation.

(29) {The hero’s sister got sick and sent her brother to find a special type of flour to heal her.}

ań ı̯aɣ-āɣi-te iŋ aɣm-əŋ, kolas ńakw-əl

now elder-girl-poss.3sg still illness-prop flour dust-ins

ta꞊titt-i-t-e, ti꞊titt-ǝs-t-e, ti꞊pusm-əs

ptcl꞊feed-npst-sg.o-3sg.s ptcl2꞊feed-pst-sg.o-3sg.s ptcl2꞊recover-pst[3sg]

‘Now his sister is still sick, he is feeding her with flour dust, he fed her, she recovered.’ [Chernetsov’s archive 50, text 1]

Further analysis of both of these particles is needed to figure out their functions and types of contexts where they are used. The next step of analysis could be to figure out whether these particles specifically interact with other phenomena discussed in this article.

Conclusion

Northern Mansi, as other Ob-Ugric languages, has a multitude of grammatical phenomena that are at least partially conditioned by information structure. This peculiarity has attracted the attention of many researchers over the years and most of their efforts were aimed at investigating the argument coding system and its connection to information structure proper, that is, to the notions of topic, focus and, in some instances, givenness. Recent research shows two possible pathways of development in this area. First, categories related to argument expression, such as choice of agreement, voice and participant coding in ditransitive constructions might be to some degree influenced by other factors, such as the referential status or animacy of the participants. It seems beneficial to combine them with the information-structural status of the constituents into a single comprehensive analysis. Secondly, it is important to understand that the argument coding system does not function independently from other phenomena in the language. Some of these phenomena, e.g. referential means, non-possessive uses of possessive markers and discourse particles have been reported to be influenced by information structure as well, see [Klumpp, Skribnik 2022]. At first, more research on specific phenomena is needed to understand their functions and the parameters affecting their choice more clearly. The next step could be to unify all the phenomena mentioned above, including the argument coding system, into a single analysis, where the discourse structure of Northern Mansi would be analysed as a whole comprehensive system. A significant challenge at this stage is dealing with differences among text genres in Northern Mansi. Narratives should be analyzed differently from texts containing generic descriptions and instruction. However, the general direction of such research looks promising and a step-by-step approach to gradually covering more ground within the domain of Northern Mansi information structure and discourse might be successful.

ABBREVIATIONS

1, 2, 3 — 1^st, 2^nd, 3^rd person

abl — ablative

add — additive

attr — attributive

caus — causative

com — comitative

conc — concessive

cond — conditional

du — dual

fut — future

hort — hortative

imp — imperative

inf — infinitive

ins — instrumental

lat — lative

loc — locative

mir — mirative

mom — momentative

npst — non-past

o — object

obl — oblique

pass — passive

pl — plural

poss — possessive

prop — proprietive

pst — past

ptcl — particle ta

ptcl2 — particle ti

ptcp.res — resultative participle

refl — reflexive

s — subject

sg — singular

trans — translative

¹ This genealogical affiliation represents the traditional views on the existence of the Ob-Ugric and Ugric branches, as in [Honti 1998], see, however, [Skribnik, Laakso 2022] for a discussion of the validity of this grouping.

² In this article, I understand information structure in terms of [Krifka 2007] as local management of common ground, which includes the notions of topic, focus, and givenness, mostly analyzed at the level of a single utterance. According to [Lambrecht 1994: 131, 213], topic is defined as “whatever the propo sition is about”, while focus is “the semantic component of a ⟨…⟩ proposition whereby the assertion differs from the presupposition”. In contrast, I use the term “discourse structure” to refer to a broader system of speech/text organization, including global common ground, narrative strategies, discourse topicality in terms of [Givón (ed.) 1983], etc.

³ Available online at: https://rosstat.gov.ru/vpn/2020 (Volume 5, Table 5).

⁴ In all Russian documents and sources Northern Mansi is referred to simply as Mansi.

⁵ The results of surveys carried out by the Ob-Ugric Institute of Applied Research and Development in Khanty-Mansiysk can be found at: https://ouipiir.ru/content/мониторинг.

⁶ Texts, as well as limited elicitation, were gathered during my fieldwork on the Upper Lozva dialect in the villages of Ushma and Treskol’je of the Ivdel’ District of the Sverdlovsk Oblast’, Russia, in 2017–2021. Examples taken from these texts are marked with ULMC (Upper Lozva Mansi Corpus).

⁷ Here I introduce the notions of A- and O-participant, which refers to the syntactic roles of the constituents in a construction with two core arguments. The A-participant is either a subject of an active construction or a demoted participant (usually the agent) in a passive construction. The O-participant is either a direct object in an active construction or a subject in a passive construction.

⁸ In Northern Mansi, there is no accusative case markers, and the “accusative” forms of personal pronouns are formed by attaching a possessive marker of the same person and number to the stem of the personal pronoun, e.g. the “accusative” form of taw [3sg] would be taw-e [3sg-poss.3sg]. All other marked cases are formed by attaching the case marker to the “accusative” form of the personal pronoun, which we gloss as an oblique form.

⁹ Mis-hum is a Mansi mythological creature, a man living in the forest.

¹⁰ A Mansi musical instrument.

¹¹ West Mansi is unique among the Mansi languages in that it has a category of dative-lative objects, which are described in [Klumpp 2023]. Northern Mansi lacks differential object marking because the accusative case is absent in the language, and the dative-lative case is not used to mark direct objects. Thus, all objects in Northern Mansi bear the unmarked “nominative” form, except for personal pronouns which have an “accusative form”, and the differential object marking is carried out only by subject vs. object agreement. Eastern and Southern Mansi possess DOM via nominative vs. accusative marking.

¹² Aissen [2003] proposes two prominence hierarchies:

1. Animacy scale: Human > Animate > Inanimate
2. Definiteness scale: Personal pronoun > Proper name > Definite NP > Indefinite specific NP > Non-specific NP

¹³ In [Gundel et al.1993], the following givenness hierarchy is proposed to represent the cognitive statuses relevant for the choice of referring expressions in discourse:

in focus > activated > familiar > uniquely identifiable > referential > type identifiable

¹⁴ The whole story is published with Russian glossing and translation in [Zhornik 2020].

¹⁵ Ten narratives from Valery Chernetsov’s archive (ranging from 300 to 900 clauses) were split into clauses, for each text I counted the total amount of clauses and the amount of clauses containing ta and calculated the percentage of clauses with the particle for each text and the average percentage for the ten texts.

作者简介

Daria Zhornik

编辑信件的主要联系方式.
Email: daria.zhornik@yandex.ru
俄罗斯联邦, Moscow; Munich, Germany

参考